what is a visual representation of data

What Is Data Visualization: Brief Theory, Useful Tips and Awesome Examples

• Share on Facebook
• Share on Twitter

By Al Boicheva

in Insights , Inspiration

3 years ago

Viewed 11,082 times

Spread the word about this article:

Updated: June 23, 2022

To create data visualization in order to present your data is no longer just a nice to have skill. Now, the skill to effectively sort and communicate your data through charts is a must-have for any business in any field that deals with data. Data visualization helps businesses quickly make sense of complex data and start making decisions based on that data. This is why today we’ll talk about what is data visualization. We’ll discuss how and why does it work, what type of charts to choose in what cases, how to create effective charts, and, of course, end with beautiful examples.

So let’s jump right in. As usual, don’t hesitate to fast-travel to a particular section of your interest.

Article overview: 1. What Does Data Visualization Mean? 2. How Does it Work? 3. When to Use it? 4. Why Use it? 5. Types of Data Visualization 6. Data Visualization VS Infographics: 5 Main Differences 7. How to Create Effective Data Visualization?: 5 Useful Tips 8. Examples of Data Visualization

1. What is Data Visualization?

Data Visualization is a graphic representation of data that aims to communicate numerous heavy data in an efficient way that is easier to grasp and understand . In a way, data visualization is the mapping between the original data and graphic elements that determine how the attributes of these elements vary. The visualization is usually made by the use of charts, lines, or points, bars, and maps.

• Data Viz is a branch of Descriptive statistics but it requires both design, computer, and statistical skills.
• Aesthetics and functionality go hand in hand to communicate complex statistics in an intuitive way.
• Data Viz tools and technologies are essential for making data-driven decisions.
• It’s a fine balance between form and functionality.
• Every STEM field benefits from understanding data.

2. How Does it Work?

If we can see it, our brains can internalize and reflect on it. This is why it’s much easier and more effective to make sense of a chart and see trends than to read a massive document that would take a lot of time and focus to rationalize. We wouldn’t want to repeat the cliche that humans are visual creatures, but it’s a fact that visualization is much more effective and comprehensive.

In a way, we can say that data Viz is a form of storytelling with the purpose to help us make decisions based on data. Such data might include:

• Tracking sales
• Identifying trends
• Identifying changes
• Monitoring goals
• Monitoring results
• Combining data

3. When to Use it?

Data visualization is useful for companies that deal with lots of data on a daily basis. It’s essential to have your data and trends instantly visible. Better than scrolling through colossal spreadsheets. When the trends stand out instantly this also helps your clients or viewers to understand them instead of getting lost in the clutter of numbers.

With that being said, Data Viz is suitable for:

• Annual reports
• Presentations
• Social media micronarratives
• Informational brochures
• Trend-trafficking
• Candlestick chart for financial analysis
• Determining routes

Common cases when data visualization sees use are in sales, marketing, healthcare, science, finances, politics, and logistics.

4. Why Use it?

Short answer: decision making. Data Visualization comes with the undeniable benefits of quickly recognizing patterns and interpret data. More specifically, it is an invaluable tool to determine the following cases.

• Identifying correlations between the relationship of variables.
• Getting market insights about audience behavior.
• Determining value vs risk metrics.
• Monitoring trends over time.
• Examining rates and potential through frequency.
• Ability to react to changes.

5. Types of Data Visualization

As you probably already guessed, Data Viz is much more than simple pie charts and graphs styled in a visually appealing way. The methods that this branch uses to visualize statistics include a series of effective types.

Map visualization is a great method to analyze and display geographically related information and present it accurately via maps. This intuitive way aims to distribute data by region. Since maps can be 2D or 3D, static or dynamic, there are numerous combinations one can use in order to create a Data Viz map.

COVID-19 Spending Data Visualization POGO by George Railean

The most common ones, however, are:

• Regional Maps: Classic maps that display countries, cities, or districts. They often represent data in different colors for different characteristics in each region.
• Line Maps: They usually contain space and time and are ideal for routing, especially for driving or taxi routes in the area due to their analysis of specific scenes.
• Point Maps: These maps distribute data of geographic information. They are ideal for businesses to pinpoint the exact locations of their buildings in a region.
• Heat Maps: They indicate the weight of a geographical area based on a specific property. For example, a heat map may distribute the saturation of infected people by area.

Charts present data in the form of graphs, diagrams, and tables. They are often confused with graphs since graphs are indeed a subcategory of charts. However, there is a small difference: graphs show the mathematical relationship between groups of data and is only one of the chart methods to represent data.

Infographic Data Visualization by Madeline VanRemmen

With that out of the way, let’s talk about the most basic types of charts in data visualization.

They use a series of bars that illustrate data development.  They are ideal for lighter data and follow trends of no more than three variables or else, the bars become cluttered and hard to comprehend. Ideal for year-on-year comparisons and monthly breakdowns.

These familiar circular graphs divide data into portions. The bigger the slice, the bigger the portion. They are ideal for depicting sections of a whole and their sum must always be 100%. Avoid pie charts when you need to show data development over time or lack a value for any of the portions. Doughnut charts have the same use as pie charts.

They use a line or more than one lines that show development over time. It allows tracking multiple variables at the same time. A great example is tracking product sales by a brand over the years. Area charts have the same use as line charts.

Scatter Plot

These charts allow you to see patterns through data visualization. They have an x-axis and a y-axis for two different values. For example, if your x-axis contains information about car prices while the y-axis is about salaries, the positive or negative relationship will tell you about what a person’s car tells about their salary.

Unlike the charts we just discussed, tables show data in almost a raw format. They are ideal when your data is hard to present visually and aim to show specific numerical data that one is supposed to read rather than visualize.

Data Visualisation | To bee or not to bee by Aishwarya Anand Singh

For example, charts are perfect to display data about a particular illness over a time period in a particular area, but a table comes to better use when you also need to understand specifics such as causes, outcomes, relapses, a period of treatment, and so on.

6. Data Visualization VS Infographics

5 main differences.

They are not that different as both visually represent data. It is often you search for infographics and find images titled Data Visualization and the other way around. In many cases, however, these titles aren’t misleading. Why is that?

• Data visualization is made of just one element. It could be a map, a chart, or a table. Infographics , on the other hand, often include multiple Data Viz elements.
• Unlike data visualizations that can be simple or extremely complex and heavy, infographics are simple and target wider audiences. The latter is usually comprehensible even to people outside of the field of research the infographic represents.
• Interestingly enough, data Viz doesn’t offer narratives and conclusions, it’s a tool and basis for reaching those. While infographics, in most cases offer a story and a narrative. For example, a data visualization map may have the title “Air pollution saturation by region”, while an infographic with the same data would go “Areas A and B are the most polluted in Country C”.
• Data visualizations can be made in Excel or use other tools that automatically generate the design unless they are set for presentation or publishing. The aesthetics of infographics , however, are of great importance and the designs must be appealing to wider audiences.
• In terms of interaction, data visualizations often offer interactive charts, especially in an online form. Infographics, on the other hand, rarely have interaction and are usually static images.

While on topic, you could also be interested to check out these 50 engaging infographic examples that make complex data look great.

7. Tips to Create Effective Data Visualization

The process is naturally similar to creating Infographics and it revolves around understanding your data and audience. To be more precise, these are the main steps and best practices when it comes to preparing an effective visualization of data for your viewers to instantly understand.

1. Do Your Homework

Preparation is half the work already done. Before you even start visualizing data, you have to be sure you understand that data to the last detail.

Knowing your audience is undeniable another important part of the homework, as different audiences process information differently. Who are the people you’re visualizing data for? How do they process visual data? Is it enough to hand them a single pie chart or you’ll need a more in-depth visual report?

The third part of preparing is to determine exactly what you want to communicate to the audience. What kind of information you’re visualizing and does it reflect your goal?

And last, think about how much data you’ll be working with and take it into account.

2. Choose the Right Type of Chart

In a previous section, we listed the basic chart types that find use in data visualization. To determine best which one suits your work, there are a few things to consider.

• How many variables will you have in a chart?
• How many items will you place for each of your variables?
• What will be the relation between the values (time period, comparison, distributions, etc.)

With that being said, a pie chart would be ideal if you need to present what portions of a whole takes each item. For example, you can use it to showcase what percent of the market share takes a particular product. Pie charts, however, are unsuitable for distributions, comparisons, and following trends through time periods. Bar graphs, scatter plots,s and line graphs are much more effective in those cases.

Another example is how to use time in your charts. It’s way more accurate to use a horizontal axis because time should run left to right. It’s way more visually intuitive.

3. Sort your Data

Start with removing every piece of data that does not add value and is basically excess for the chart. Sometimes, you have to work with a huge amount of data which will inevitably make your chart pretty complex and hard to read. Don’t hesitate to split your information into two or more charts. If that won’t work for you, you could use highlights or change the entire type of chart with something that would fit better.

Tip: When you use bar charts and columns for comparison, sort the information in an ascending or a descending way by value instead of alphabetical order.

4. Use Colors to Your Advantage

In every form of visualization, colors are your best friend and the most powerful tool. They create contrasts, accents, and emphasis and lead the eye intuitively. Even here, color theory is important.

When you design your chart, make sure you don’t use more than 5 or 6 colors. Anything more than that will make your graph overwhelming and hard to read for your viewers. However, color intensity is a different thing that you can use to your advantage. For example, when you compare the same concept in different periods of time, you could sort your data from the lightest shade of your chosen color to its darker one. It creates a strong visual progression, proper to your timeline.

Things to consider when you choose colors:

• Different colors for different categories.
• A consistent color palette for all charts in a series that you will later compare.
• It’s appropriate to use color blind-friendly palettes.

5. Get Inspired

Always put your inspiration to work when you want to be at the top of your game. Look through examples, infographics, and other people’s work and see what works best for each type of data you need to implement.

This Twitter account Data Visualization Society is a great way to start. In the meantime, we’ll also handpick some amazing examples that will get you in the mood to start creating the visuals for your data.

8. Examples for Data Visualization

As another art form, Data Viz is a fertile ground for some amazing well-designed graphs that prove that data is beautiful. Now let’s check out some.

Dark Souls III Experience Data

We start with Meng Hsiao Wei’s personal project presenting his experience with playing Dark Souls 3. It’s a perfect example that infographics and data visualization are tools for personal designs as well. The research is pretty massive yet very professionally sorted into different types of charts for the different concepts. All data visualizations are made with the same color palette and look great in infographics.

My dark souls 3 playing data by Meng Hsiao Wei

Greatest Movies of all Time

Katie Silver has compiled a list of the 100 greatest movies of all time based on critics and crowd reviews. The visualization shows key data points for every movie such as year of release, oscar nominations and wins, budget, gross, IMDB score, genre, filming location, setting of the film, and production studio. All movies are ordered by the release date.

100 Greatest Movies Data Visualization by Katie Silver

The Most Violent Cities

Federica Fragapane shows data for the 50 most violent cities in the world in 2017. The items are arranged on a vertical axis based on population and ordered along the horizontal axis according to the homicide rate.

The Most Violent Cities by Federica Fragapane

Family Businesses as Data

These data visualizations and illustrations were made by Valerio Pellegrini for Perspectives Magazine. They show a pie chart with sector breakdown as well as a scatter plot for contribution for employment.

PERSPECTIVES MAGAZINE – Family Businesses by Valerio Pellegrini

Orbit Map of the Solar System

The map shows data on the orbits of more than 18000 asteroids in the solar system. Each asteroid is shown at its position on New Years’ Eve 1999, colored by type of asteroid.

An Orbit Map of the Solar System by Eleanor Lutz

The Semantics Of Headlines

Katja Flükiger has a take on how headlines tell the story. The data visualization aims to communicate how much is the selling influencing the telling. The project was completed at Maryland Institute College of Art to visualize references to immigration and color-coding the value judgments implied by word choice and context.

The Semantics of Headlines by Katja Flükiger

Moon and Earthquakes

This data visualization works on answering whether the moon is responsible for earthquakes. The chart features the time and intensity of earthquakes in response to the phase and orbit location of the moon.

Moon and Earthquakes by Aishwarya Anand Singh

Dawn of the Nanosats

The visualization shows the satellites launched from 2003 to 2015. The graph represents the type of institutions focused on projects as well as the nations that financed them. On the left, it is shown the number of launches per year and satellite applications.

WIRED UK – Dawn of the by Nanosats by Valerio Pellegrini

Final Words

Data visualization is not only a form of science but also a form of art. Its purpose is to help businesses in any field quickly make sense of complex data and start making decisions based on that data. To make your graphs efficient and easy to read, it’s all about knowing your data and audience. This way you’ll be able to choose the right type of chart and use visual techniques to your advantage.

You may also be interested in some of these related articles:

• Infographics for Marketing: How to Grab and Hold the Attention
• 12 Animated Infographics That Will Engage Your Mind from Start to Finish
• 50 Engaging Infographic Examples That Make Complex Ideas Look Great
• Good Color Combinations That Go Beyond Trends: Inspirational Examples and Ideas

Add some character to your visuals

Cartoon Characters, Design Bundles, Illustrations, Backgrounds and more...

Like us on Facebook

Subscribe to our newsletter

Be the first to know what’s new in the world of graphic design and illustrations.

• [email protected]

Browse High Quality Vector Graphics

E.g.: businessman, lion, girl…

Related Articles

How to turn yourself into animated cartoon in zoom.us, need powerpoint backgrounds the best places to check out [+ freebies], 35 modern clean website design examples of flattering simplicity, inspirational small business website examples that do it right, 10 of the best logo design software options [free and paid], check out our infographics bundle with 500+ infographic templates:, enjoyed this article.

Don’t forget to share!

• Comments (2)

Al Boicheva

Al is an illustrator at GraphicMama with out-of-the-box thinking and a passion for anything creative. In her free time, you will see her drooling over tattoo art, Manga, and horror movies.

Thousands of vector graphics for your projects.

Hey! You made it all the way to the bottom!

Here are some other articles we think you may like:

The Basics of Character Design Process: Insights and Examples

by Al Boicheva

Freepik vs Vecteezy: Which one to choose?

by Iveta Pavlova

How to Find The Right WordPress Development Company for You

by Lyudmil Enchev

Looking for Design Bundles or Cartoon Characters?

A source of high-quality vector graphics offering a huge variety of premade character designs, graphic design bundles, Adobe Character Animator puppets, and more.

Data visualization is the representation of data through use of common graphics, such as charts, plots, infographics and even animations. These visual displays of information communicate complex data relationships and data-driven insights in a way that is easy to understand.

Data visualization can be utilized for a variety of purposes, and it’s important to note that is not only reserved for use by data teams. Management also leverages it to convey organizational structure and hierarchy while data analysts and data scientists use it to discover and explain patterns and trends.  Harvard Business Review  (link resides outside ibm.com) categorizes data visualization into four key purposes: idea generation, idea illustration, visual discovery, and everyday dataviz. We’ll delve deeper into these below:

Idea generation

Data visualization is commonly used to spur idea generation across teams. They are frequently leveraged during brainstorming or  Design Thinking  sessions at the start of a project by supporting the collection of different perspectives and highlighting the common concerns of the collective. While these visualizations are usually unpolished and unrefined, they help set the foundation within the project to ensure that the team is aligned on the problem that they’re looking to address for key stakeholders.

Idea illustration

Data visualization for idea illustration assists in conveying an idea, such as a tactic or process. It is commonly used in learning settings, such as tutorials, certification courses, centers of excellence, but it can also be used to represent organization structures or processes, facilitating communication between the right individuals for specific tasks. Project managers frequently use Gantt charts and waterfall charts to illustrate  workflows .  Data modeling  also uses abstraction to represent and better understand data flow within an enterprise’s information system, making it easier for developers, business analysts, data architects, and others to understand the relationships in a database or data warehouse.

Visual discovery

Visual discovery and every day data viz are more closely aligned with data teams. While visual discovery helps data analysts, data scientists, and other data professionals identify patterns and trends within a dataset, every day data viz supports the subsequent storytelling after a new insight has been found.

Data visualization

Data visualization is a critical step in the data science process, helping teams and individuals convey data more effectively to colleagues and decision makers. Teams that manage reporting systems typically leverage defined template views to monitor performance. However, data visualization isn’t limited to performance dashboards. For example, while  text mining  an analyst may use a word cloud to to capture key concepts, trends, and hidden relationships within this unstructured data. Alternatively, they may utilize a graph structure to illustrate relationships between entities in a knowledge graph. There are a number of ways to represent different types of data, and it’s important to remember that it is a skillset that should extend beyond your core analytics team.

Use this model selection framework to choose the most appropriate model while balancing your performance requirements with cost, risks and deployment needs.

Register for the ebook on generative AI

The earliest form of data visualization can be traced back the Egyptians in the pre-17th century, largely used to assist in navigation. As time progressed, people leveraged data visualizations for broader applications, such as in economic, social, health disciplines. Perhaps most notably, Edward Tufte published  The Visual Display of Quantitative Information  (link resides outside ibm.com), which illustrated that individuals could utilize data visualization to present data in a more effective manner. His book continues to stand the test of time, especially as companies turn to dashboards to report their performance metrics in real-time. Dashboards are effective data visualization tools for tracking and visualizing data from multiple data sources, providing visibility into the effects of specific behaviors by a team or an adjacent one on performance. Dashboards include common visualization techniques, such as:

• Tables: This consists of rows and columns used to compare variables. Tables can show a great deal of information in a structured way, but they can also overwhelm users that are simply looking for high-level trends.
• Pie charts and stacked bar charts:  These graphs are divided into sections that represent parts of a whole. They provide a simple way to organize data and compare the size of each component to one other.
• Line charts and area charts:  These visuals show change in one or more quantities by plotting a series of data points over time and are frequently used within predictive analytics. Line graphs utilize lines to demonstrate these changes while area charts connect data points with line segments, stacking variables on top of one another and using color to distinguish between variables.
• Histograms: This graph plots a distribution of numbers using a bar chart (with no spaces between the bars), representing the quantity of data that falls within a particular range. This visual makes it easy for an end user to identify outliers within a given dataset.
• Scatter plots: These visuals are beneficial in reveling the relationship between two variables, and they are commonly used within regression data analysis. However, these can sometimes be confused with bubble charts, which are used to visualize three variables via the x-axis, the y-axis, and the size of the bubble.
• Heat maps:  These graphical representation displays are helpful in visualizing behavioral data by location. This can be a location on a map, or even a webpage.
• Tree maps, which display hierarchical data as a set of nested shapes, typically rectangles. Treemaps are great for comparing the proportions between categories via their area size.

Access to data visualization tools has never been easier. Open source libraries, such as D3.js, provide a way for analysts to present data in an interactive way, allowing them to engage a broader audience with new data. Some of the most popular open source visualization libraries include:

• D3.js: It is a front-end JavaScript library for producing dynamic, interactive data visualizations in web browsers.  D3.js  (link resides outside ibm.com) uses HTML, CSS, and SVG to create visual representations of data that can be viewed on any browser. It also provides features for interactions and animations.
• ECharts:  A powerful charting and visualization library that offers an easy way to add intuitive, interactive, and highly customizable charts to products, research papers, presentations, etc.  Echarts  (link resides outside ibm.com) is based in JavaScript and ZRender, a lightweight canvas library.
• Vega:   Vega  (link resides outside ibm.com) defines itself as “visualization grammar,” providing support to customize visualizations across large datasets which are accessible from the web.
• deck.gl: It is part of Uber's open source visualization framework suite.  deck.gl  (link resides outside ibm.com) is a framework, which is used for  exploratory data analysis  on big data. It helps build high-performance GPU-powered visualization on the web.

With so many data visualization tools readily available, there has also been a rise in ineffective information visualization. Visual communication should be simple and deliberate to ensure that your data visualization helps your target audience arrive at your intended insight or conclusion. The following best practices can help ensure your data visualization is useful and clear:

Set the context: It’s important to provide general background information to ground the audience around why this particular data point is important. For example, if e-mail open rates were underperforming, we may want to illustrate how a company’s open rate compares to the overall industry, demonstrating that the company has a problem within this marketing channel. To drive an action, the audience needs to understand how current performance compares to something tangible, like a goal, benchmark, or other key performance indicators (KPIs).

Know your audience(s): Think about who your visualization is designed for and then make sure your data visualization fits their needs. What is that person trying to accomplish? What kind of questions do they care about? Does your visualization address their concerns? You’ll want the data that you provide to motivate people to act within their scope of their role. If you’re unsure if the visualization is clear, present it to one or two people within your target audience to get feedback, allowing you to make additional edits prior to a large presentation.

Choose an effective visual:  Specific visuals are designed for specific types of datasets. For instance, scatter plots display the relationship between two variables well, while line graphs display time series data well. Ensure that the visual actually assists the audience in understanding your main takeaway. Misalignment of charts and data can result in the opposite, confusing your audience further versus providing clarity.

Keep it simple:  Data visualization tools can make it easy to add all sorts of information to your visual. However, just because you can, it doesn’t mean that you should! In data visualization, you want to be very deliberate about the additional information that you add to focus user attention. For example, do you need data labels on every bar in your bar chart? Perhaps you only need one or two to help illustrate your point. Do you need a variety of colors to communicate your idea? Are you using colors that are accessible to a wide range of audiences (e.g. accounting for color blind audiences)? Design your data visualization for maximum impact by eliminating information that may distract your target audience.

An AI-infused integrated planning solution that helps you transcend the limits of manual planning.

Build, run and manage AI models. Prepare data and build models on any cloud using open source code or visual modeling. Predict and optimize your outcomes.

Unlock the value of enterprise data and build an insight-driven organization that delivers business advantage with IBM Consulting.                                   

Your trusted Watson co-pilot for smarter analytics and confident decisions.

Use features within IBM Watson® Studio that help you visualize and gain insights into your data, then cleanse and transform your data to build high-quality predictive models.

Data Refinery makes it easy to explore, prepare, and deliver data that people across your organization can trust.

Learn how to use Apache Superset (a modern, enterprise-ready business intelligence web application) with Netezza database to uncover the story behind the data.

Predict outcomes with flexible AI-infused forecasting and analyze what-if scenarios in real-time. IBM Planning Analytics is an integrated business planning solution that turns raw data into actionable insights. Deploy as you need, on-premises or on cloud.

We use essential cookies to make Venngage work. By clicking “Accept All Cookies”, you agree to the storing of cookies on your device to enhance site navigation, analyze site usage, and assist in our marketing efforts.

Manage Cookies

Cookies and similar technologies collect certain information about how you’re using our website. Some of them are essential, and without them you wouldn’t be able to use Venngage. But others are optional, and you get to choose whether we use them or not.

Strictly Necessary Cookies

These cookies are always on, as they’re essential for making Venngage work, and making it safe. Without these cookies, services you’ve asked for can’t be provided.

Show cookie providers

• Google Login

Functionality Cookies

These cookies help us provide enhanced functionality and personalisation, and remember your settings. They may be set by us or by third party providers.

Performance Cookies

These cookies help us analyze how many people are using Venngage, where they come from and how they're using it. If you opt out of these cookies, we can’t get feedback to make Venngage better for you and all our users.

• Google Analytics

Targeting Cookies

These cookies are set by our advertising partners to track your activity and show you relevant Venngage ads on other sites as you browse the internet.

• Google Tag Manager

Infographics

• Daily Infographics
• Popular Templates
• Accessibility
• Graphic Design
• Graphs and Charts
• Data Visualization
• Human Resources
• Beginner Guides

Blog Graphic Design What is Data Visualization? (Definition, Examples, Best Practices)

What is Data Visualization? (Definition, Examples, Best Practices)

Written by: Midori Nediger Jun 05, 2020

Words don’t always paint the clearest picture. Raw data doesn’t always tell the most compelling story. 

The human mind is very receptive to visual information. That’s why data visualization is a powerful tool for communication.    

But if “data visualization” sounds tricky and technical don’t worry—it doesn’t have to be. 

This guide will explain the fundamentals of data visualization in a way that anyone can understand. Included are a ton of examples of different types of data visualizations and when to use them for your reports, presentations, marketing, and more.

Table of Contents

• What is data visualization?

What is data visualization used for?

Types of data visualizations.

• How to present data visually  (for businesses, marketers, nonprofits, and education)
• Data visualization examples

Data visualization is used everywhere. 

Businesses use data visualization for reporting, forecasting, and marketing. 

CREATE THIS REPORT TEMPLATE

Nonprofits use data visualizations to put stories and faces to numbers. 

Source:  Bill and Melinda Gates Foundation

Scholars and scientists use data visualization to illustrate concepts and reinforce their arguments.

CREATE THIS MIND MAP TEMPLATE

Reporters use data visualization to show trends and contextualize stories. 

While data visualizations can make your work more professional, they can also be a lot of fun.

What is data visualization? A simple definition of data visualization:

Data visualization is the visual presentation of data or information. The goal of data visualization is to communicate data or information clearly and effectively to readers. Typically, data is visualized in the form of a chart , infographic , diagram or map. 

The field of data visualization combines both art and data science. While a data visualization can be creative and pleasing to look at, it should also be functional in its visual communication of the data. 

Data, especially a lot of data, can be difficult to wrap your head around. Data visualization can help both you and your audience interpret and understand data. 

Data visualizations often use elements of visual storytelling to communicate a message supported by the data. 

There are many situations where you would want to present data visually. 

Data visualization can be used for:

• Making data engaging and easily digestible
• Identifying trends and outliers within a set of data
• Telling a story found within the data
• Reinforcing an argument or opinion
• Highlighting the important parts of a set of data

Let’s look at some examples for each use case.

1. Make data digestible and easy to understand

Often, a large set of numbers can make us go cross-eyed. It can be difficult to find the significance behind rows of data. 

Data visualization allows us to frame the data differently by using illustrations, charts, descriptive text, and engaging design. Visualization also allows us to group and organize data based on categories and themes, which can make it easier to break down into understandable chunks. 

Related : How to Use Data Visualization in Your Infographics

For example, this infographic breaks down the concept of neuroplasticity in an approachable way:

Source: NICABM

The same goes for complex, specialized concepts. It can often be difficult to break down the information in a way that non-specialists will understand. But an infographic that organizes the information, with visuals, can demystify concepts for novice readers.

CREATE THIS INFOGRAPHIC TEMPLATE

NEW! Introducing: Marketing Statistics Report 2022

It’s 2022 already. Marketers, are you still using data from pre-COVID times?

Don’t make decisions based on outdated data that no longer applies. It’s time you keep yourself informed of the latest marketing statistics and trends during the past two years, and learn how COVID-19 has affected marketing efforts in different industries — with this FREE marketing statistics report put together by Venngage and HubSpot .

The report uses data gathered from over 100,000 customers of HubSpot CRM. In addition to that, you’ll also know about the trends in using visuals in content marketing and the impacts of the pandemic on visual content, from 200+ marketers all over the world interviewed by Venngage.

GET YOUR FREE COPY

2. Identify trends and outliers

If you were to sift through raw data manually, it could take ages to notice patterns, trends or outlying data. But by using data visualization tools like charts, you can sort through a lot of data quickly. 

Even better, charts enable you to pick up on trends a lot quicker than you would sifting through numbers.

For example, here’s a simple chart generated by Google Search Console that shows the change in Google searches for “toilet paper”. As you can see, in March 2020 there was a huge increase in searches for toilet paper:

Source: How to Use SEO Data to Fuel Your Content Marketing Strategy in 2020

This chart shows an outlier in the general trend for toilet paper-related Google searches. The reason for the outlier? The outbreak of COVID-19 in North America. With a simple data visualization, we’ve been able to highlight an outlier and hint at a story behind the data. 

Uploading your data into charts, to create these kinds of visuals is easy. While working on your design in the editor, select a chart from the left panel. Open the chart and find the green IMPORT button under the DATA tab. Then upload the CSV file and your chart automatically visualizes the information. 

3. Tell a story within the data

Numbers on their own don’t tend to evoke an emotional response. But data visualization can tell a story that gives significance to the data. 

Designers use techniques like color theory , illustrations, design style and visual cues to appeal to the emotions of readers, put faces to numbers, and introduce a narrative to the data. 

Related : How to Tell a Story With Data (A Guide for Beginners)

For example, here’s an infographic created by World Vision. In the infographics, numbers are visualized using illustrations of cups. While comparing numbers might impress readers, reinforcing those numbers with illustrations helps to make an even greater impact. 

Source: World Vision

Meanwhile, this infographic uses data to draw attention to an often overlooked issue:

Read More:  The Coronavirus Pandemic and the Refugee Crisis

4. Reinforce an argument or opinion

When it comes to convincing people your opinion is right, they often have to see it to believe it. An effective infographic or chart can make your argument more robust and reinforce your creativity. 

For example, you can use a comparison infographic to compare sides of an argument, different theories, product/service options, pros and cons, and more. Especially if you’re blending data types.

5. Highlight an important point in a set of data

Sometimes we use data visualizations to make it easier for readers to explore the data and come to their own conclusions. But often, we use data visualizations to tell a story, make a particular argument, or encourage readers to come to a specific conclusion. 

Designers use visual cues to direct the eye to different places on a page. Visual cues are shapes, symbols, and colors that point to a specific part of the data visualization, or that make a specific part stand out.

For example, in this data visualization, contrasting colors are used to emphasize the difference in the amount of waste sent to landfills versus recycled waste:

Here’s another example. This time, a red circle and an arrow are used to highlight points on the chart where the numbers show a drop: 

Highlighting specific data points helps your data visualization tell a compelling story.

6. Make books, blog posts, reports and videos more engaging

At Venngage, we use data visualization to make our blog posts more engaging for readers. When we write a blog post or share a post on social media, we like to summarize key points from our content using infographics. 

The added benefit of creating engaging visuals like infographics is that it has enabled our site to be featured in publications like The Wall Street Journal , Mashable , Business Insider , The Huffington Post and more. 

That’s because data visualizations are different from a lot of other types of content people consume on a daily basis. They make your brain work. They combine concrete facts and numbers with impactful visual elements. They make complex concepts easier to grasp. 

Here’s an example of an infographic we made that got a lot of media buzz:

Read the Blog Post: Every Betrayal Ever in Game of Thrones

We created this infographic because a bunch of people on our team are big Game of Thrones fans and we wanted to create a visual that would help other fans follow the show. Because we approached a topic that a lot of people cared about in an original way, the infographic got picked up by a bunch of media sites. 

Whether you’re a website looking to promote your content, a journalist looking for an original angle, or a creative building your portfolio, data visualizations can be an effective way to get people’s attention.

Data visualizations can come in many different forms. People are always coming up with new and creative ways to present data visually. 

Generally speaking, data visualizations usually fall under these main categories:

An infographic is a collection of imagery, charts, and minimal text that gives an easy-to-understand overview of a topic. 

While infographics can take many forms, they can typically be categorized by these infographic types:

• Statistical infographics
• Informational infographics
• Timeline infographics
• Process infographics
• Geographic infographics
• Comparison infographics
• Hierarchical infographics
• List infographics
• Resume infographics

Read More: What is an Infographic? Examples, Templates & Design Tips

Charts 

In the simplest terms, a chart is a graphical representation of data. Charts use visual symbols like line, bars, dots, slices, and icons to represent data points. 

Some of the most common types of charts are:

• Bar graphs /charts
• Line charts
• Bubble charts
• Stacked bar charts
• Word clouds
• Pictographs
• Area charts
• Scatter plot charts
• Multi-series charts

The question that inevitably follows is: what type of chart should I use to visualize my data? Does it matter?

Short answer: yes, it matters. Choosing a type of chart that doesn’t work with your data can end up misrepresenting and skewing your data. 

For example: if you’ve been in the data viz biz for a while, then you may have heard some of the controversy surrounding pie charts. A rookie mistake that people often make is using a pie chart when a bar chart would work better. 

Pie charts display portions of a whole. A pie chart works when you want to compare proportions that are substantially different. Like this:

CREATE THIS CHART TEMPLATE

But when your proportions are similar, a pie chart can make it difficult to tell which slice is bigger than the other. That’s why, in most other cases, a bar chart is a safer bet.

Here is a cheat sheet to help you pick the right type of chart for your data:

Want to make better charts? Make engaging charts with Venngage’s Chart Maker .

Related : How to Choose the Best Types of Charts For Your Data

Similar to a chart, a diagram is a visual representation of information. Diagrams can be both two-dimensional and three-dimensional. 

Some of the most common types of diagrams are:

• Venn diagrams
• Tree diagrams
• SWOT analysis
• Fishbone diagrams
• Use case diagrams

Diagrams are used for mapping out processes, helping with decision making, identifying root causes, connecting ideas, and planning out projects.

CREATE THIS DIAGRAM TEMPLATE

Want to make a diagram ? Create a Venn diagram and other visuals using our free Venn Diagram Maker .

A map is a visual representation of an area of land. Maps show physical features of land like regions, landscapes, cities, roads, and bodies of water. 

Source: National Geographic

A common type of map you have probably come across in your travels is a choropleth map . Choropleth maps use different shades and colors to indicate average quantities. 

For example, a population density map uses varying shades to show the difference in population numbers from region to region:

Create your own map for free with Venngage’s Map Maker .

How to present data visually (data visualization best practices)

While good data visualization will communicate data or information clearly and effectively, bad data visualization will do the opposite. Here are some practical tips for how businesses and organizations can use data visualization to communicate information more effectively. 

Not a designer? No problem. Venngage’s Graph Maker  will help you create better graphs in minutes.

1. Avoid distorting the data

This may be the most important point in this whole blog post. While data visualizations are an opportunity to show off your creative design chops, function should never be sacrificed for fashion. 

The chart styles, colors, shapes, and sizing you use all play a role in how the data is interpreted. If you want to present your data accurately and ethically, then you need to take care to ensure that your data visualization does not present the data falsely. 

There are a number of different ways data can be distorted in a chart. Some common ways data can be distorted are:

• Making the baselines something other than 0 to make numbers seem bigger or smaller than they are – this is called “truncating” a graph
• Compressing or expanding the scale of the Y-axis to make a line or bar seem bigger or smaller than it should be
• Cherry picking data so that only the data points you want to include are on a graph (i.e. only telling part of the story)
• Using the wrong type of chart, graph or diagram for your data
• Going against standard, expected data visualization conventions

Because people use data visualizations to reinforce their opinions, you should always read data visualizations with a critical eye. Often enough, writers may be using data visualization to skew the data in a way that supports their opinions, but that may not be entirely truthful.

Read More: 5 Ways Writers Use Graphs To Mislead You

Want to create an engaging line graph? Use Venngage’s Line Graph Maker to create your own in minutes.

2. Avoid cluttering up your design with “chartjunk”

When it comes to best practices for data visualization, we should turn to one of the grandfather’s of data visualization: Edward Tufte. He coined the term “ chartjunk ”, which refers to the use of unnecessary or confusing design elements that skews or obscures the data in a chart. 

Here’s an example of a data visualization that suffers from chartjunk:

Source: ExcelUser

In this example, the image of the coin is distracting for readers trying to interpret the data. Note how the fonts are tiny – almost unreadable. Mistakes like this are common when a designers tries to put style before function. 

Read More : The Worst Infographics of 2020 (With Lessons for 2021)

3. Tell a story with your data

Data visualizations like infographics give you the space to combine data and narrative structure in one page. Visuals like icons and bold fonts let you highlight important statistics and facts.

For example, you could customize this data visualization infographic template to show the benefit of using your product or service (and post it on social media):

USE THIS TEMPLATE

  This data visualization relies heavily on text and icons to tell the story of its data:

This type of infographic is perfect for those who aren’t as comfortable with charts and graphs. It’s also a great way to showcase original research, get social shares and build brand awareness.

4. Combine different types of data visualizations

While you may choose to keep your data visualization simple, combining multiple types of charts and diagrams can help tell a more rounded story.

Don’t be afraid to combine charts, pictograms and diagrams into one infographic. The result will be a data visualization infographic that is engaging and rich in visual data.

Design Tip: This data visualization infographic would be perfect for nonprofits to customize and include in an email newsletter to increase awareness (and donations).

Or take this data visualization that also combines multiple types of charts, pictograms, and images to engage readers. It could work well in a presentation or report on customer research, customer service scores, quarterly performance and much more:

Design Tip: This infographic could work well in a presentation or report on customer research, customer service scores, quarterly performance and much more.

Make your own bar graph in minutes with our free Bar Graph Maker .

5. Use icons to emphasize important points

Icons are perfect for attracting the eye when scanning a page. (Remember: use visual cues!)

If there are specific data points that you want readers to pay attention to, placing an icon beside it will make it more noticeable:

Design Tip: This infographic template would work well on social media to encourage shares and brand awareness.

You can also pair icons with headers to indicate the beginning of a new section.

Meanwhile, this infographic uses icons like bullet points to emphasize and illustrate important points. 

Design Tip: This infographic would make a great sales piece to promote your course or other service.  

6. Use bold fonts to make text information engaging

A challenge people often face when setting out to visualize information is knowing how much text to include. After all, the point of data visualization is that it presents information visually, rather than a page of text. 

Even if you have a lot of text information, you can still create present data visually. Use bold, interesting fonts to make your data exciting. Just make sure that, above all else, your text is still easy to read.

This data visualization uses different fonts for the headers and body text that are bold but clear. This helps integrate the text into the design and emphasizes particular points:

Design Tip: Nonprofits could use this data visualization infographic in a newsletter or on social media to build awareness, but any business could use it to explain the need for their product or service. 

As a general rule of thumb, stick to no more than three different font types in one infographic.

This infographic uses one font for headers, another font for body text, and a third font for accent text. 

Read More: How to Choose Fonts For Your Designs (With Examples)

Design Tip: Venngage has a library of fonts to choose from. If you can’t find the icon you’re looking for , you can always request they be added. Our online editor has a chat box with 24/7 customer support.

7. Use colors strategically in your design

In design, colors are as functional as they are fashionable. You can use colors to emphasize points, categorize information, show movement or progression, and more. 

For example, this chart uses color to categorize data:

Design Tip : This pie chart can actually be customized in many ways. Human resources could provide a monthly update of people hired by department, nonprofits could show a breakdown of how they spent donations and real estate agents could show the average price of homes sold by neighbourhood.

You can also use light colored text and icons on dark backgrounds to make them stand out. Consider the mood that you want to convey with your infographic and pick colors that will reflect that mood. You can also use contrasting colors from your brand color palette.

This infographic template uses a bold combination of pinks and purples to give the data impact:

Read More: How to Pick Colors to Captivate Readers and Communicate Effectively

8. Show how parts make up a whole

It can be difficult to break a big topic down into smaller parts. Data visualization can make it a lot easier for people to conceptualize how parts make up a whole.

Using one focus visual, diagram or chart can convey parts of a whole more effectively than a text list can. Look at how this infographic neatly visualizes how marketers use blogging as part of their strategy:

Design Tip: Human resources could use this graphic to show the results of a company survey. Or consultants could promote their services by showing their success rates.

Or look at how this infographic template uses one focus visual to illustrate the nutritional makeup of a banana:

CREATE THIS FLYER TEMPLATE

9. Focus on one amazing statistic

If you are preparing a presentation, it’s best not to try and cram too many visuals into one slide. Instead, focus on one awe-inspiring statistic and make that the focus of your slide.

Use one focus visual to give the statistic even more impact. Smaller visuals like this are ideal for sharing on social media, like in this example:

Design Tip: You can easily swap out the icon above (of Ontario, Canada) using Venngage’s drag-and-drop online editor and its in-editor library of icons. Click on the template above to get started.

This template also focuses on one key statistic and offers some supporting information in the bar on the side:

10. Optimize your data visualization for mobile

Complex, information-packed infographics are great for spicing up reports, blog posts, handouts, and more. But they’re not always the best for mobile viewing. 

To optimize your data visualization for mobile viewing, use one focus chart or icon and big, legible font. You can create a series of mobile-optimized infographics to share multiple data points in a super original and attention-grabbing way.

For example, this infographic uses concise text and one chart to cut to the core message behind the data:

CREATE THIS SOCIAL MEDIA TEMPLATE

Some amazing data visualization examples

Here are some of the best data visualization examples I’ve come across in my years writing about data viz. 

Evolution of Marketing Infographic

Graphic Design Trends Infographic

Stop Shark Finning Nonprofit Infographic

Source: Ripetungi

Coronavirus Impact on Environment Data Visualization

What Disney Characters Tell Us About Color Theory

World’s Deadliest Animal Infographic

Source: Bill and Melinda Gates Foundation

The Secret Recipe For a Viral Creepypasta

Read More: Creepypasta Study: The Secret Recipe For a Viral Horror Story

The Hero’s Journey Infographic

Read More: What Your 6 Favorite Movies Have in Common

Emotional Self Care Guide Infographic

Source: Carley Schweet

Want to look at more amazing data visualization? Read More: 50+ Infographic Ideas, Examples & Templates for 2020 (For Marketers, Nonprofits, Schools, Healthcare Workers, and more)

Discover popular designs

Infographic maker

Brochure maker

White paper online

Newsletter creator

Flyer maker

Timeline maker

Letterhead maker

Mind map maker

Ebook maker

What is Data Visualization? Definition, Examples, Best Practices

This guide provides an introduction to data visualization, including real-world examples, best practices and editable templates.

• Share on LinkedIn
• Share through Email
• Print this page
• Bookmark this page

Resource Details

June 5, 2020

Data visualization is the visual presentation of data or information. The goal of data visualization is to communicate data or information clearly and effectively to readers. Typically, data is visualized in the form of a chart, infographic, diagram, or map.

The field of data visualization combines both art and data science. While data visualization can be creative and pleasing to look at, it should also be functional in its visual communication of the data.

This resource explains the fundamentals of data visualization, including examples of different types of data visualizations and when and how to use them to illustrate findings and insights.

Do you have feedback on this resource?

Thank you for your feedback as we strive to curate and publish resources to help social impact organizations succeed with data.

Send us a note

Explore More

Related Guides & Resources

Getting started with data visualization.

Visualizing data is one of the most effective ways of communicating data. This can take many forms from live digital data dashboards to static charts shared in social media channels.

Communicating Results: Design your Visualization

Let’s get visual: nonprofit data visualization.

The Ultimate Guide to Data Visualization

Data visualization is important because it breaks down complex data and extracts meaningful insights in a more digestible way. Displaying the data in a more engaging way helps audiences make sense of the information with a higher chance of retention. But with a variety of charts and graphs, how can you tell which is best for your specific content and audience?

Consider this your ultimate guide to data visualization. We’re breaking down popular charts and graphs and explaining the differences between each so that you can choose the best slide for your story. 

Charts vs. graphs

We know that numbers don’t lie and are a strong way to back up your story, but that doesn’t always mean they’re easy to understand. By packaging up complex numbers and metrics in visually appealing graphics you’re telling your audience exactly what they need to know without having to rack their brain to comprehend it. Graphs and charts are important in your presentation because they take your supporting statistics, and story, and make them more relatable. 

Charts present data or complex information through tables, infographics , and diagrams, while graphs show a connection between two or more sets of data.

A histogram is a visual representation of the distribution of data. The graph itself consists of a set of rectangles— each rectangle represents a range of values (called a "bin"), while the height corresponds to the numbers of the data that fall within that range.

Histograms are oftentimes used to visualize the frequency distribution of continuous data. Things such as measurements of height, weight, or time can all be organized in the graph. They can also be used to display the distribution of discrete data, like the number of shoes sold in a shoe department during any given period of time.

Histograms are a useful tool for analyzing data, as they allow you to quickly see the shape of the data distribution, the location of the central tendency (the mean or median), and the full spread of the data. They’re a great chart that can also reveal any changes in the data, making it easier to digest.

Need to add a little visual interest to your business presentation? A bar graph slide can display your data easily and effectively. Whether you use a vertical bar graph or horizontal bar graph, a bar graph gives you options to help simplify and present complex data, ensuring you get your point across.

Use it to track long-term changes.

Vertical bar graphs are great for comparing different groups that change over a long period of time. Small or short-term changes may not be as obvious in bar graph form.

Don’t be afraid to play with design .

You can use one bar graph template slide to display a lot of information, as long as you differentiate between data sets. Use colors, spacing, and labels to make the differences obvious.

Use a horizontal graph when necessary.

If your data labels are long, a horizontal bar graph may be easier to read and organize than a vertical bar graph. 

Don’t use a horizontal graph to track time.

A vertical bar graph makes more sense when graphing data over time, since the x-axis is usually read from left to right.

Histograms vs. bar graphs

While a histogram is similar to a bar graph, it groups numbers into ranges and displays data in a different way.

Bar graphs are used to represent categorical data, where each bar represents a different category with a height or length proportional to the associated value. The categories of a bar graph don’t overlap, and the bars are usually separated by a gap to differentiate from one another. Bar graphs are ideal when you need to compare the data of different categories.

On the other hand, histograms divide data into a set of intervals or "bins". The bars of a histogram are typically adjacent to each other, with no gaps, as the bins are continuous and can overlap. Histograms are used to visualize the shape, center, and spread of a distribution of numerical data.

A pie chart is a circular graph (hence the name ‘pie’) that’s used to show or compare different segments — or ‘slices’ — of data. Each slice represents a proportion that relates to the whole. When added up, each slice should equal the total. Pie charts are best used for showcasing part-to-whole relationships. In other words, if you have different parts or percentages of a whole, using a pie chart is likely the way to go. Just make sure the total sum equals 100%, or the chart won’t make a lot of sense or convey the message you want it to. Essentially, any type of content or data that can be broken down into comparative categories is suitable to use. Revenue, demographics, market shares, survey results — these are just a few examples of the type of content to use in a pie chart. However, you don’t want to display more than six categories of data or the pie chart can be difficult to read and compare the relative size of slices. 

Donut Charts

A donut chart is almost identical to a pie chart, but the center is cut out (hence the name ‘donut’). Donut charts are also used to show proportions of categories that make up the whole, but the center can also be used to display data. Like pie charts, donut charts can be used to display different data points that total 100%. These are also best used to compare a handful of categories at-a-glance and how they relate to the whole. The same type of content you’d use for a pie chart can also work for a donut chart. However, with donut charts, you have room for fewer categories than pie charts — anywhere from 2 to 5. That’s because you want your audience to be able to quickly tell the difference between arc lengths, which can help tell a more compelling story and get your point across more efficiently. 

Pie charts vs. donut charts

You may notice that a donut chart and a pie chart look almost identical . While a donut chart is essentially the same as a pie chart in function, with its center cut out, the “slices” in a donut chart are sometimes more clearly defined than in a pie chart.

When deciding between a pie chart or a donut chart for your presentation, make sure the data you’re using is for comparison analysis only. Pie and donut charts are usually limited to just that — comparing the differences between categories. The easiest way to decide which one to use? 

The number of categories you’re comparing. If you have more than 4 or 5 categories, go with a pie chart. If you have between 2 and 4 categories, go with a donut chart. Another way to choose? If you have an extra data point to convey (e.g. all of your categories equal an increase in total revenue), use a donut chart so you can take advantage of the space in the middle.

Comparison charts

As its name implies, a comparison chart or comparison graph draws a comparison between two or more items across different parameters. You might use a comparison chart to look at similarities and differences between items, weigh multiple products or services in order to choose one, or present a lot of data in an easy-to-read format.

For a visually interesting twist on a plain bar chart, add a data comparison slide to your presentation. Our data comparison template is similar to a bar graph, using bars of varying lengths to display measured data. The data comparison template, however, displays percentages instead of exact numbers. One of the best things about using Beautiful.ai’s data comparison slide? You can customize it for your presentation. Create a horizontal or vertical slide, remove or add grid lines, play with its design, and more.

Gantt charts

A Gantt chart , named after its early 20th century inventor Henry Gantt, is a birds-eye view of a project. It visually organizes tasks displayed over time. Gantt charts are incredibly useful tools that work for projects and groups of all sizes. 

It’s a type of bar chart that you would use to show the start and finish dates of several elements of a project such as what the project tasks are, who is working on each task, how long each task will take, and how tasks group together, overlap, and link with each other. The left side of a Gantt chart lists each task in a project by name. Running along the top of the chart from left to right is a timeline. Depending on the demands and details of your project, the timeline may be broken down by quarter, month, week, or even day.

Project management can be complex, so it’s important to keep your chart simple by using a color scheme with cool colors like blues or greens. You can color code items thematically or by department or person, or even highlight a single task with a contrasting color to call attention to it. You can also choose to highlight important tasks using icons or use images for other annotations. This will make your chart easier to read and more visually appealing. 

Additional tips for creating an effective Gantt chart slide .

Use different colors

How many colors you use and how you assign them is up to you. You might choose one color to represent a specific team or department so that you can see who is responsible for which tasks on your chart, for example. 

Set milestones

Don’t forget to set milestones where they make sense: deadlines required by clients or customers, when a new department takes over the next phase of the project, or when a long list of tasks is completed. 

Label your tasks

When used with a deliberate color scheme, labeling your tasks with its project owner will prevent confusion and make roles clear to everyone. 

Jordan Turner

Jordan is a Bay Area writer, social media manager, and content strategist.

Recommended Articles

7 examples of how bar graphs help tell a story, how to use a flowchart, new data visualization: data import & linking, how to find a graph or chart that makes your data look good.

What Is Data Visualization and Why Is It Important? A Complete Introduction

They say a picture is worth a thousand words, and this is especially true for data analytics.

Data visualization is all about presenting data in a visual format, using charts, graphs, and maps to tell a meaningful story. It’s a crucial step in the data analysis process—and a technique (or art form!) that all areas of business can benefit from.

In this guide, we’ll tell you everything you need to know about data visualization (also known as data viz). We’ll explain what it is, why it matters, some of the most common types, as well as the tools you can use to create them.

This guide is ideal for anyone who wants to present, communicate, and share data-driven insights.

If you’d like to learn more data analytics skills, try this free data short course .

• What is data visualization?
• Why is data visualization important?
• When should you visualize your data? 
• Different types of data visualization and when to use them
• Top data visualization tools
• Best practices and principles for effective data visualization
• Getting started with data visualization

So: What is data visualization? Let’s start with a definition.

1. What is data visualization? A definition

Data visualization is the graphical or visual representation of data. It helps to highlight the most useful insights from a dataset, making it easier to spot trends, patterns, outliers, and correlations.

Imagine you’re presented with a spreadsheet containing rows and rows of data. You probably won’t be able to decipher the data without delving into it, and it’s unlikely that you’ll be able to spot trends and patterns at first glance.

Now imagine seeing the same data presented as a bar chart, or on a color-coded map. It’s much easier to see what the data is telling you, right?

That’s the whole point of data visualization. It makes insights visible to the naked eye, so that virtually anyone can see and understand what’s going on. When done well, data visualization tells a story.

This storytelling aspect is crucial as it makes your data actionable. There’s a huge difference between simply having lots of data versus actually understanding how to use it to drive actions and decisions—and data visualization bridges that gap.

There are two broad categories of data visualization: exploration and explanation. Let’s take a look at those now.

What are the two main types of data visualization? Exploration vs. explanation

We’ll look at specific types of data visualization later on, but for now, it’s important to distinguish between exploratory and explanatory data visualization.

In a nutshell, exploratory data visualization helps you figure out what’s in your data, while explanatory visualization helps you to communicate what you’ve found. Exploration takes place while you’re still analyzing the data, while explanation comes towards the end of the process when you’re ready to share your findings.

Exploration

When faced with a new dataset, one of the first things you’ll do is carry out an exploratory data analysis . This is where you investigate the dataset and identify some of its main features, laying the foundation for more thorough analysis.

At this stage, visualizations can make it easier to get a sense of what’s in your dataset and to spot any noteworthy trends or anomalies. Ultimately, you’re getting an initial lay of the land and finding clues as to what the data might be trying to tell you.

Explanation

Once you’ve conducted your analysis and have figured out what the data is telling you, you’ll want to share these insights with others.

These could be key business stakeholders who can take action based on the data, for example, or public audiences who have an interest in your topic area.

Explanatory data visualizations help you tell this story, and it’s up to you to determine which visualizations will help you to do so most effectively. We’ll introduce some of the most common types of data visualization (and when to use them) in section four.

Want to learn more about data visualization, and try your hand at creating visualizations of your own?  Give this free introductory tutorial a go. We’ll show you, step by step, how to create bar charts, line graphs, and more for a real dataset in Google Sheets.

2. Why is data visualization important?

The importance of effective data visualization is rooted in the importance of data analytics in general.

We’re living in an increasingly data-rich world; at the start of 2020, the digital universe comprised approximately 44 zettabytes of data . For perspective, one zettabyte is roughly equal to a trillion gigabytes. By 2025, it’s estimated that around 463 exabytes of data will be created every 24 hours across the globe. An exabyte is equivalent to one billion gigabytes. Basically, we’re producing tons and tons of data all the time.

Data analytics allows us to make sense of (at least some of) that data. From a business perspective, it enables companies to learn from the past and plan ahead for the future. In fields like healthcare, it can help to improve patient care and treatment. In finance and insurance, it can help to assess risk and combat fraudulent activity. Essentially, we need data analytics in order to make smart decisions—and data visualization is a crucial part of that.

Data visualization helps us to understand what certain data is telling us, presenting it in a way that’s accessible to a range of audiences—not just data experts. It’s how you bridge the gap between your expertise as a data analyst or data scientist, and those people who can use or act upon the insights you discover.

A line graph and a bar chart taken from the Fitbit app.

The advantages and benefits of effective data visualization at a glance

Data visualization allows you to:

• Get an initial understanding of your data by making trends, patterns, and outliers easily visible to the naked eye
• Comprehend large volumes of data quickly and efficiently
• Communicate insights and findings to non-data experts, making your data accessible and actionable
• Tell a meaningful and impactful story, highlighting only the most relevant information for a given context

Now we know what data visualization is and why it matters, let’s take a look at when and why you might need to visualize your data.

3. When should you visualize your data?

Aside from exploratory data visualization which takes place in the early stages, data visualization usually comprises the final step in the data analysis process . To recap, the data analysis process can be set out as follows:

• Define the question: What problem are you trying to solve?
• Collect the data: Determine what kind of data you need and where you’ll find it.
• Clean the data: Remove errors, duplicates, outliers, and unwanted data points—anything that might skew how your data is interpreted. You can learn more about data cleaning (and how to do it) in this guide .
• Analyze the data: Determine the type of data analysis you need to carry out in order to find the insights you’re looking for.
• Visualize the data and share your findings: Translate your key insights into visual format (e.g. graphs, charts, or heatmaps) and present them to the relevant audience(s).

Essentially, you visualize your data any time you want to summarize and highlight key findings and share them with others. With that in mind, let’s consider what kinds of insights you can convey with data visualizations.

What is data visualization used for?

Within the broader goal of conveying key insights, different visualizations can be used to tell different stories. Data visualizations can be used to:

• Convey changes over time: For example, a line graph could be used to present how the value of Bitcoin changed over a certain time period.
• Determine the frequency of events: You could use a histogram to visualize the frequency distribution of a single event over a certain time period (e.g. number of internet users per year from 2007 to 2021). Learn how to create a histogram in this guide .
• Highlight interesting relationships or correlations between variables: If you wanted to highlight the relationship between two variables (e.g. marketing spend and revenue, or hours of weekly exercise vs. cardiovascular fitness), you could use a scatter plot to see, at a glance, if one increases as the other decreases (or vice versa).
• Examine a network: If you want to understand what’s going on within a certain network (for example, your entire customer base), network visualizations can help you to identify (and depict) meaningful connections and clusters within your network of interest.
• Analyze value and risk: If you want to weigh up value versus risk in order to figure out which opportunities or strategies are worth pursuing, data visualizations—such as a color-coded system—could help you to categorize and identify, at a glance, which items are feasible.

So far, we’ve taken a rather broad, high-level look at data visualization. Now let’s drill down to some specific types of data visualization and when to use them.

An example of data visualization, as seen in the Fitbit app.

4. How to visualize your data: Different types of data visualization (and when to use them)

There are many different options when it comes to visualizing your data. The visualization you choose depends on the type of data you’re working with and what you want to convey or highlight. It’s also important to consider the complexity of your data and how many different variables are involved. Not all types of data visualization lend themselves to elaborate or complex depictions, so it’s important to choose a suitable technique.

Before we explore some of the most common types of data visualization, let’s first introduce five main data visualization categories.

Five data visualization categories

When considering the different types of data viz, it helps to be aware of the different categories that these visualizations may fall into:

• Temporal data visualizations are linear and one-dimensional. Examples include scatterplots, timelines, and line graphs.
• Hierarchical visualizations organize groups within larger groups, and are often used to display clusters of information. Examples include tree diagrams, ring charts, and sunburst diagrams.
• Network visualizations show the relationships and connections between multiple datasets. Examples include matrix charts, word clouds, and node-link diagrams.
• Multidimensional or 3D visualizations are used to depict two or more variables. Examples include pie charts, Venn diagrams, stacked bar graphs, and histograms.
• Geospatial visualizations convey various data points in relation to physical, real-world locations (for example, voting patterns across a certain country). Examples include heat maps, cartograms, and density maps.

With those categories in mind, let’s explore some of the most common types of data visualization.

Five common types of data visualization (and when to use them)

In this section, we’ll introduce some useful types of data visualization. We’ll also point you to our more comprehensive guide where you can learn about additional data visualization methods and how to use them.

1. Scatterplots

Scatterplots (or scatter graphs) visualize the relationship between two variables. One variable is shown on the x-axis, and the other on the y-axis, with each data point depicted as a single “dot” or item on the graph. This creates a “scatter” effect, hence the name.

Source: displayr.com

Scatterplots are best used for large datasets when there’s no temporal element. For example, if you wanted to visualize the relationship between a person’s height and weight, or between how many carats a diamond measures and its monetary value, you could easily visualize this using a scatterplot.

It’s important to bear in mind that scatterplots simply describe the correlation between two variables; they don’t infer any kind of cause-and-effect relationship.

2. Bar charts

Bar charts are used to plot categorical data against discrete values.

Categorical data refers to data that is not numeric, and it’s often used to describe certain traits or characteristics. Some examples of categorical data include things like education level (e.g. high school, undergrad, or post-grad) and age group (e.g. under 30, under 40, under 50, or 50 and over).

Discrete values are those which can only take on certain values—there are no “half measures” or “gray areas.” For example, the number of people attending an event would be a discrete variable, as would the number of sales made in a certain time period (think about it: you can’t make “half a sale” or have “half an event attendee.”)

Source: chartio.com

So, with a bar chart, you have your categorical data on the x-axis plotted against your discrete values on the y-axis.

The height of the bars is directly proportional to the values they represent, making it easy to compare your data at a glance.

3. Pie charts

Just like bar charts, pie charts are used to visualize categorical data.

However, while bar charts represent multiple categories of data, pie charts are used to visualize just one single variable broken down into percentages or proportions. A pie chart is essentially a circle divided into different “slices,” with each slice representing the percentage it contributes to the whole.

Thus, the size of each pie slice is proportional to how much it contributes to the whole “pie.”

Imagine you have a class of thirty students and you want to divide them up based on what color t-shirt they’re wearing on a given day.

The possible “slices” are red, green, blue, and yellow, with each color representing 40%, 30%, 25%, and 5% of the class total respectively. You could easily visualize this using a pie chart—and the yellow slice (5%) would be considerably thinner than the red slice (40%)! Pie charts are best suited for data that can be split into a maximum of five or six categories.

4. Network graphs

Not all data is simple enough to be summarized in a bar or pie chart. For those more complex datasets, there are a range of more elaborate data visualizations at your disposal—network graphs being one of them.

Network graphs show how different elements or entities within a network relate to one another, with each element represented by an individual node. These nodes are connected to other, related nodes via lines.

Source: networkofthrones.wordpress.com

Network graphs are great for spotting and representing clusters within a large network of data.

Let’s imagine you have a huge database filled with customers, and you want to segment them into meaningful clusters for marketing purposes. You could use a network graph to draw connections and parallels between all your customers or customer groups.

With any luck, certain clusters and patterns would emerge, giving you a logical means by which to group your audience.

5. Geographical maps

Geo maps are used to visualize the distribution of data in relation to a physical, geographical area.

For example, you could use a color-coded map to see how natural oil reserves are distributed across the world, or to visualize how different states voted in a political election. Maps are an extremely versatile form of data visualization, and are an excellent way of communicating all kinds of location-related data.

Some other types of maps used in data visualization include dot distribution maps (think scatterplots combined with a map), and cartograms which distort the size of geographical areas to proportionally represent a given variable (population density, for example).

Source: pmfias.com

Here, we’ve introduced just a handful of data visualization types. If you want to learn more, check out our complete guide to different types of data visualization and when to use them .

5. Top data visualization tools

When it comes to creating informative, eye-catching visualizations, there are plenty of tools at your disposal.

When choosing a tool, it’s important to consider your needs in terms of the kinds of visualizations you want to create, as well as your own technical expertise; some tools will require coding knowledge, while others are more suited to non-technical users.

In this section, we’ll briefly introduce some of the most popular data visualization tools. If you’re on the market for a data viz tool and want a more thorough comparison, this guide to the seven best data visualization tools will help you. For now, here are our top three data viz tools to get familiar with:

• Plotly: Open-source software built on Python. Plotly is ideal if you’ve got some coding knowledge and want to create highly customizable visualizations.
• D3.js: A free, open-source data viz library built using JavaScript. As with Plotly, you’ll need some programming knowledge in order to use this data viz tool.
• Tableau: Perhaps one of the most popular data analytics tools , Tableau is known for its user-friendliness—you don’t need any coding knowledge to create beautiful visualizations in Tableau. And, unlike some other BI tools, it’s good at handling large volumes of data.

Before deciding on a tool, it’s worth trying out a few options. The good news is that there are plenty of data viz tools on the market— as well as a number of free tools —allowing you to create beautiful and informative visualizations—even if you’re a newcomer to the field.

What are data dashboards?

Dashboards are another useful tool for data tracking and visualization. A data dashboard essentially allows you to keep track of multiple data sources, visualizing them in one single location for easy viewing.

A common example is the Google Analytics dashboard , which displays a whole host of visualizations on one page—a geo map showing where your website visitors are located, for example, or a pie chart showing what percentage of your users access your website using specific devices.

If you want multiple stakeholders to be able to access and view certain data insights, a dashboard can help you to create a single hub with easy-to-understand visualizations.

A snapshot of a data dashboard, taken from Google Analytics.

6. What are some data visualization best practices?

Data visualization truly is an art form—but the goal is always, first and foremost, to provide valuable information and insights.

If you can do this by way of beautiful visualizations, you’re onto a winner. So, when creating data visualizations, it’s important to adhere to certain best practices.

These will help you strike the right balance, keeping your audience engaged and informed. Here’s how to excel at data visualization.

1. Define a clear purpose

Like any data analytics project, it’s important to define a clear purpose for your data visualizations.

What are the priorities in terms of what you want to convey and communicate? What should your audience take away from your visualization? It’s essential to have this defined from the outset; that way, you can ensure that you’re only presenting the most valuable information—and giving your audience something they can use and act upon.

2. Know your audience

The purpose of data visualization is to communicate insights to a specific audience, so you’ll want to give some thought to who your audience is and how familiar they are with the information you’re presenting.

What kind of context can you provide around your visualizations in order to help your audience understand them? What types of visualization are likely to be most accessible to this particular group of people? Keep your audience in mind at all times.

3. Keep it simple

When creating visualizations, it’s often the case that less is more.

Ultimately, you want your visualizations to be as digestible as possible, and that means trimming away any unnecessary information while presenting key insights clearly and succinctly. The goal is to keep cognitive load to a minimum—that is, the amount of “brainpower” or mental effort it takes to process information.

Even if the data is complex, your visualizations don’t have to be, so strive for simplicity at all times.

4. Avoid distorting the data

You should strive to present your findings as accurately as possible, so avoid any kind of visual “tricks” that could bias how your data is perceived and interpreted.

Think about the labels you use, as well as how you scale your visualizations. For example, things like “blowing up” certain data segments to make them appear more significant, or starting your graph axis on a number other than zero are both bad practices which could mislead your audience. Prioritize integrity and accuracy!

5. Ensure your visualizations are inclusive

Last but by no means least, make sure that your visualizations are accessible and inclusive.

Think about how colors, contrasts, font sizes, and the use of white space affect the readability of your visualization. Is it easy for your users to distinguish between the data and see what’s going on, regardless of whether they have twenty-twenty vision or a visual impairment?

Inclusivity and accessibility are central to good data visualization, so don’t overlook this step.

7. Getting started with data visualization

By now, you hopefully have a good understanding of what data visualization is and why it matters.

Of course, the best way to get to grips with it is to see it in action. Check out our round-up of some of the most beautiful and informative data visualization examples from around the web.

Keen to give it a go yourself? Why not download a free dataset and see what you can do! If you’d like to learn it more, then check out this list of data visualization courses out there to try.

Data visualization is an excellent skill to have, whether you’re forging a career in the data industry or just want to share valuable insights with your colleagues. If you are pursuing a career as a data analyst or data scientist, be sure to include data visualizations in your data portfolio —it’s something that employers will be looking out for.

CareerFoundry’s  Data Visualizations with Python course is designed to ease you into this vital area of data analytics. You can take it as a standalone course as well as a specialization within our full Data Analytics Program, you’ll learn and apply the principles of data viz in a real-world project, as well as getting to grips with various data visualization libraries.

Want to learn more? Try your hand at this free, introductory data analytics short course , and check out the following guides:

• What is data quality and why is it important?
• What is web scraping? A beginner’s guide
• An introduction to multivariate analysis

• What is Data Analytics: Empowering Decision-Making with the Power of Data
• Top Text Analytics Companies In Singapore
• Optimizing Test Automation ROI: How Selenium, Cypress, JavaScript Frameworks Deliver Value
• F1 Betting 2024: Odds, Predictions & Hot Picks for Silverstone
• Selenium Automation Testing: In-House Team vs. Outsourcing – A Strategic Decision
• 2024 Copa América Heats Up: Argentina vs Ecuador – Who Will Reach the Semis?
• Navigating the UI Test Automation Flaky Cycle: Regaining Control and Confidence in Your Test Suite
• 15 Key Observations on NBA Free Agency Reshaping the League
• How to Analyze Test Results and Improve Test Coverage with Automation Testing
• A Beginner’s Guide to Designing a Test Automation Framework for Software Testing

The Ultimate Guide to Data Visualization with FAQs

Data visualization is the graphical representation of information and data through charts, graphs, maps, or other visual elements. The primary objective of data visualization is to present complex datasets in a clear, concise, and easily understandable format.

This helps business and marketing decision-makers to thoroughly understand the data through visual aids and make informed decisions. Colors play a huge part in enhancing the visually represented data for improving understanding.

Moreover, thankfully, the integration of geospatial data has aided localized target marketing. Marketers can use this data to target a niche audience and map out their customer journeys to further refine campaigns.

Let’s discuss each of these factors in more detail for better understanding.

Utilizing Geospatial Data for Localized Marketing

Geospatial data is the information related to a specific location. When this data is presented in the form of a map, it is known at geospatial data visualization . This method of data representation can be used by marketers to target customers in local markets.

There are several different ways to depict geographic data :

●      Impact of Geospatial Data Visualization

Geospatial data visualization provides a lot of benefits in the field of marketing as it provides businesses with invaluable insights into consumer behavior, preferences, and local market dynamics.

Location intelligence can provide companies with a competitive edge in targeting specific geographic areas. The impact of geospatial data visualization lies in its ability to transform raw data into easily understandable maps and graphics that makes complex information accessible and actionable.

In marketing, one of the key decision making aspect lies in effectively understanding the spatial distribution of customers, competitors, and market trends.

This form of data visualization allows marketers to identify hotspots of activity, pinpoint areas with high potential for growth, and uncover patterns that may be overlooked in traditional data analysis.

Viisually representing data on maps can allow businesses to make strategic choices tailored to the unique characteristics of each locality.

Using Spatial Insights to Form Strategies to Target Local Markets

One key strategy businesses can employ to effectively target local markets though geospatial data insights involves location-based advertising, where marketing efforts are concentrated in areas with a high concentration of target customers.

Geotargeting enables businesses to deliver personalized messages, promotions, and advertisements to individuals based on their physical location which increases the relevance and impact of marketing campaigns.

Furthermore, understanding the location-based relationship between competitors and potential customers is essential.

Geospatial data visualization allows businesses to assess market saturation, identify underserved areas, and allows them to strategically position themselves for maximum visibility and customer engagement.

This strategic spatial analysis empowers businesses to optimize their marketing budgets by focusing efforts where they are likely to yield the highest returns.

Practical Applications of Geospatial Visualization in Marketing

There are many practical applications of geospatial data visualization in marketing. One notable application is site selection for physical stores or service centers.

Analyzing geospatial data can allow businesses to identify optimal locations based on factors such as foot traffic, proximity to competitors, and local demographics. This ensures that the physical establishments are strategically placed to attract the target audience.

Moreover, geospatial visualization aids in demographic targeting and segmentation. Marketers can overlay demographic data onto geographic maps which allows them to understand the population in specific areas in a better way.

This information can then be used to tailor marketing messages, product offerings, and promotional strategies to align with the preferences and characteristics of local demographics.

Geospatial data visualization is a powerful tool for marketers seeking to optimize their strategies for localized marketing.

Analyzing the impact of geospatial data visualization, implementing targeted strategies, and applying practical applications in marketing can helo businesses unlock new opportunities and stay ahead in today’s dynamic and competitive market landscape.

Enhancing Insights with Customer Journey Visualization

Customer journey mapping is an intricate process consisting of multiple stages that, when coupled with data visualization, can allow marketers to make highly accurate campaign decisions personalized to each customer.

The stages of mapping out a customer journey includes:

Mapping the Customer Journey

Understanding the customer journey is at the heart of successful marketing, and customer journey visualization has become an indispensable tool for businesses aiming to enhance their insights into consumer behavior.

Customer journey mapping involves charting the various touchpoints a customer encounters from initial awareness to post-purchase engagement.

Visualizing this journey helps businesses gain a bird’s-eye view of the entire customer experience, identifying key interactions, pain points, and opportunities for improvement.

A comprehensive customer journey map typically includes stages such as awareness, consideration, purchase, retention, and advocacy.

Visual representations of the customer journey enable marketers to empathize with the customer’s perspective, leading to more customer-centric strategies.

These maps offer a visual narrative that helps teams align their efforts to create a seamless and engaging customer experience across all touchpoints.

●      Extracting Valuable Insights from Visualized Customer Journeys

The true power of customer journey visualization lies in its ability to extract valuable insights that might otherwise remain hidden in raw data.

Through visual representations, businesses can identify patterns, trends, and correlations that provide a deeper understanding of customer behavior and can uncover the critical moments that influence purchase decisions and customer satisfaction.

Visualized customer journeys also facilitate the identification of pain points or areas where customers may face challenges. Pinpointing these pain points allows businesses to implement targeted improvements, ensuring a smoother and more satisfying customer experience.

Additionally, by aligning marketing efforts with specific stages of the customer journey, businesses can tailor their messaging and strategies to address customer needs at each phase.

●      Practical Tips for Effective Customer Journey Data Visualization

Creating effective visualizations of the customer journey requires careful consideration of design, data accuracy, and interpretation.

Practical tips for enhancing customer journey data visualization include:

• Clear Representation: The visual representation of the customer journey should be clear and easy to understand. You can use intuitive icons, colors, and labels to convey different touchpoints and stages.
• Data Accuracy: Base visualizations on accurate and reliable data as inaccurate information can lead to misguided insights and ineffective strategies. Make sure to regularly update data sources to maintain relevance.
• User Feedback Integration: Incorporate customer feedback into the visualization process. This qualitative data provides valuable context and enriches the understanding of customer emotions and perceptions at each touchpoint.
• Collaboration Across Teams: Foster collaboration among cross-functional teams involved in the customer journey. Collaborative efforts ensure that insights from visualizations translate into actionable strategies across marketing, sales, and customer service departments.

Color Techniques in Data Visualization

The exploration of color psychology involves examining how colors impact human behavior, emotions, and cognitive functions. Various colors trigger distinct responses and may carry varied interpretations across different cultures.

In the realm of data visualization, a profound grasp of color psychology becomes essential, given its potential to influence the comprehension, engagement, and retention of the conveyed information.

The Role of Color Psychology in Effective Data Representation

Color plays an important role in data visualization as it greatly impacts how information is perceived and understood. Understanding the role of color psychology is crucial for creating effective and impactful data representations.

Colors stir emotions, convey significance, and guide attention im viewers which make them powerful tools for enhancing the interpretability of visualized information.

In the context of data visualization, different colors can be associated with specific meanings. For instance, warm colors like red and orange may signify urgency or importance, while cool colors like blue and green may represent tranquility or neutrality.

Color psychology can be used by data visualizers to guide viewers to focus on key elements, emphasize trends, and establish a hierarchy of information within a visualization.

Moreover, color can be employed to highlight contrasts and differences, aiding in the comparison of data points. By using distinct colors for various categories or data sets, visualizations become more accessible which allows viewers to quickly spot patterns and outliers.

However, it is essential to strike a balance and avoid overwhelming the audience with an excessive use of colors, which could lead to confusion rather than clarity.

●      Communication and Understanding in Color Displays

The effective use of color in data displays serves as a powerful communication tool that enhances understanding and engagement. Incorporating color strategically can make data visualizations more accessible, memorable, and impactful.

Color can be employed to draw attention to critical data points, trends, or anomalies. Whether through highlighting specific elements with bold colors or using subtle variations to indicate small changes.

Additionally, color can aid in storytelling within visualizations. Data storytellers can guide the audience through a memorable journey by associating colors with different segments of a narrative or specific themes.

This not only improves comprehension but also fosters a deeper connection with the information presented.

Best Practices for Using Color in Visualization

While color offers a powerful means of enhancing visualizations, adhering to best practices is essential to maximize its effectiveness.

Several key principles guide the optimal use of color in information visualization:

• Maintain Simplicity: Avoid using too many colors within a single visualization. A limited color palette enhances clarity and prevents visual overload. Select a restrained set of colors that are both aesthetically pleasing and distinct enough to convey the intended information.
• Consider Accessibility: Ensure that color choices are accessible to all viewers, including those with color vision deficiencies. Using patterns, labels, or varying line styles in addition to color can make visualizations more inclusive.
• Use Consistent Schemes: Maintain consistency in color schemes across related visualizations or charts. Consistency aids in creating a cohesive narrative and facilitates easy comparison between different data sets or time periods.
• Align Colors with Data Significance: Assign colors based on the inherent meaning of the data. For example, use a gradient scale to represent numerical values, with lighter shades indicating lower values and darker shades indicating higher values.

In conclusion, the transformative potential of geospatial data visualization and customer journey mapping offers marketers invaluable insights into consumer behavior, aiding strategic decision-making, and optimizing localized marketing efforts.

The journey through customer mapping reveals its significance in fostering a customer-centric approach, identifying pain points, and extracting actionable insights for enhanced customer experiences.

Additionally, the exploration of color techniques in data visualization emphasizes the pivotal role of color in conveying information effectively. Businesses can elevate their visualizations, ensuring clarity and impactful communication through the use of color in data displays.

Collectively, these insights underscore the importance of leveraging data visualization tools strategically to navigate the complexities of modern marketing and stay ahead in the dynamic business landscape.

FQAs on Data Visualization

Geospatial data, also known as geodata, encompasses information tied to specific locations on the Earth’s surface. This type of data allows the mapping of objects, events, and real-world phenomena to precise geographical areas, identified by latitude and longitude coordinates.

A visual depiction of the interactions between a customer and a company across the entire relationship, a customer journey map illustrates various touchpoints. This comprehensive overview of the stages a customer experiences enables marketers to foresee behavior, anticipate needs, and guide the company’s responsive strategies.

In data visualization, choosing colors goes beyond aesthetics; it’s a critical tool for accurately conveying quantitative information. Well-chosen colors ensure that the underlying data is represented faithfully, distinguishing it from common color schemes that may distort relationships between data values.

I'm an experienced SEO expert with a wealth of knowledge in digital and analytics marketing. I help organizations make sense of their marketing strategies in coherence with their business goals. Outside of my role at Analytico, I have a major in Marketing with a Bachelor's in Business Administration.

SAP Roles and Responsibilities

2024 Telecom Transformation Through the Top 10 Data Analytics Strategies

Related posts.

Seaborn Bootcamp: Install, Explore, and Master Data Visualization

Data Visualization: Transforming Raw Data into Rich Understanding

A Deep Dive into QuickSort Algorithm: Sorting Your Data with Ease

Write a comment cancel reply.

Save my name, email, and website in this browser for the next time I comment.

• Documentation

Data Visualization Essentials: Tips, Techniques, and Tools

Written by Tom Czaban   |  May 17, 2023

Table of Contents

What Is Data Visualization?

Why is data visualization important, types of data visualization, out-of-the-box visualizations, fully-customized visualizations, what is real-time data visualization, what is interactive data visualization, advantages and disadvantages of data visualization, what tools do i need for data visualization, examples of data visualization, data visualization best practices, ready to get started with data visualization, discover more about data visualization.

Data visualization is the use of visual representations to display information. This definition might sound modern, but data visualization has been around for centuries. One of the earliest and most obvious examples is maps, which developed out of the need to graphically display geographic data. Since then data visualization has continued to develop to meet the needs of today’s users.

There are multiple ways to visualize data (including charts, graphs, and infographics), and technology is constantly evolving to present information in more eye-catching and useful ways. Examples of this include making visualizations interactive and allowing the end user to filter and display different metrics. Regardless of these updates, the aim remains the same: to present key insights and make it easier to engage with and understand data.

Any discussion of the meaning of data visualization would be incomplete without a mention of creating dashboards . It is important to note that although data visualization and dashboards are closely related they are not in fact the same thing. The aim of a dashboard is to offer an overview of the key performance indicators (KPIs) of the presented area. Typically a dashboard will contain multiple visualizations, which together provide an overview of the key insights — as shown in the image below.

Data visualization helps to ensure data insights aren’t lost in delivery; most of us can’t process big blocks of statistics, our brains aren’t built like that. Anyone who has looked at a long list of numbers will understand the disconnect this can cause. Graphical representation solves this pain point by making statistics and data easier to absorb.

Data visualization is not only about creating simple and attractive visuals. It can be used to create insights by identifying patterns and trends that would otherwise be difficult to spot. Displaying a set of data on a scatter plot , for example, might reveal connections between outliers that previously went unnoticed when the statistics were in a table.

Data visualization is also an important business intelligence (BI) tool, allowing companies to effectively communicate their data and improve decision-making. High-quality visualizations can help an organization promote a data culture because the insights needn’t be explained to non-technical end users. Decisions can be made with increased accuracy and speed because everyone is on the same page. As well as improving internal processes, visualizations help to increase external engagement by making the data more accessible to partners and customers.

There are different types of visualizations to choose from, and each is better suited to showcasing certain attributes and metrics. It is important to use the visualization that makes the most sense for the insights you aim to convey. Below is a brief overview of the visualizations you might choose for some typical scenarios. For more on this, check out our post on how to choose the best chart type to visualize your data .

There are instances when you may need to display one key figure , for example, the number of customers, or the number of returned items. A KPI visualization is best suited to this purpose because it shows one big number. However, this number will mean nothing on its own; you have to, at the very least, provide a date range and compare it with another metric to give it some context.

To show comparisons between categories , for example, the number of sales each staff member has made in the last month, it is best to use a bar chart or column chart . A stacked bar chart gives you the option to add another category, so as well as showing how many sales each staff member has made, you might also include the product type they sold by adding color and a key.

When comparing parts to the whole it is best to use a pie chart , donut chart , or treemap . An example of part-to-whole comparison is the number of people who answered ‘yes’ or ‘no’ to a specific question. Generally speaking, it is a bad idea to use a pie chart or donut chart for more than three categories because it becomes difficult for users to accurately absorb the data. With more categories, it is better to use a treemap.

To show changes over time the most effective options are line charts , area charts, or column charts. You might, for instance, choose one of these to display month-by-month revenue. If you want to add an additional category (such as product type) you can use a line chart with multiple lines or a stacked area chart. But it's best to tread carefully with these because they can become confusing if not properly executed.

To show the details of many items it is best to use a table. Some people avoid using tables because they seem too basic, but when you have many items (such as a lot of customer details) a table can be the right choice. Amid the myriad of visualization options available, tables can be quite striking when combined with other types of charts and graphs on a dashboard.

Most analytics platforms offer out-of-the-box visualizations that you can use to display your data. Out-of-the-box refers to the standard visualizations available to all customers who have purchased the BI tool. These options typically include tables, column charts, bar charts, line charts, donut charts, etc. Not all out-of-the-box offerings are created equal, so when choosing a BI tool, it is helpful if it has at least some of the following capabilities:

• An interface that makes it easy to change metrics, dimensions, or data visualization types from the out-of-the-box options, such as drag-and-drop.
• The ability to add filters and drill into visualizations.
• The option to adjust certain aspects to match your brand, for example, fonts, colors, and logos.
• The potential to populate visualizations with live data.

While you can do plenty with out-of-the-box visualizations, your options are limited to the standard charts and graphs that come with the tool. For this reason, you’re often better off with a solution that allows you more creative options. With custom visualization, you can tailor your visualizations to your exact needs. You are no longer limited to standard chart types, and the only barrier to creating highly original visuals is your imagination.

A good analytics platform achieves custom visualization by enabling access to third-party charting libraries, such as D3.js, Chart.js, Fusion Charts, Google Charts, and more. These libraries are the best route to creating advanced visuals that are graphically amazing.

Want to see what GoodData can do for you?

Get a guided tour and ask us about GoodData’s features, implementation, and pricing.

With real-time data visualization, users can see the data changing as it is being updated or generated. For example, they might see the height of bar charts changing, or colors adjusting themselves on a heatmap.

To create this kind of visualization, a company needs the ability to perform real-time data reporting. In other words, their data architecture must operate in real-time to build up-to-the-minute visuals. This architecture includes components such as data processing, data streaming, and all the logic of the defined analysis that leads to the displayed insights.

Real-time analytics and their visualization can be crucial under certain circumstances. On other occasions, this can be unnecessary and even confusing. Before deciding whether to employ automated real-time visualization, a company might ask themselves the following questions:

• Do we really need to see in-the-moment data to make decisions?
• Do we have enough new data at frequent intervals to necessitate real-time updates?
• Will these instant updates make our decision-making processes more confusing and create delays?
• Will real-time visualization help our users, i.e., what value will they get from the live updates?

Based on the above questions, scenarios where real-time visualizations may be beneficial include:

• Security and fraud prevention, e.g., when monitoring for major security breaches.
• A situation where a company needs to act promptly to a crisis.
• Goal monitoring (but only if these goals are affected by rapidly changing information)
• In financial teams – where it is crucial for team members to receive up-to-the-minute financial information (for example, stock markets).

Below is an example of real-time visualization in action:

On the surface, interactive data visualization tends to look similar to regular (or static) visualization. The difference is there is an option to click a button or move a slider, so the user can interact with the data, rather than just look at it. This ability to manipulate charts, graphs, and maps can positively influence user experience (UX).

Sometimes interactive features are custom-designed for a specific purpose, but generally speaking, the most common interactive features are:

• Filtering: Allows you to filter for the exact information required, highlighting the relevant data and reducing the data that is currently unimportant.
• Drilling: Enables you to move between different visualizations and send an action from the dashboard.
• Zooming and panning: Creates the possibility to hone in on a particular detail; you can zoom in on a specific part of the visualization and pan across it.

Interactive features can be extremely useful, for example, when users quickly need to answer specific queries, which is why they’re often used in BI reports. However, there are also times when a static visualization might be the best choice; for instance, when visualizations need to be printed and shared as reports, or when it is unnecessary for users to manipulate the information and all they need to do is to look at it. For more on this, check out our article that compares interactive and non-interactive visualizations using the example of world happiness levels.

While there are huge benefits to visualizing data, if not done properly the technique can spread misinformation. Below we look at some of the main advantages of data visualization and how to mitigate its downsides.

Benefits of data visualization

• Easy to spot trends. Visualization allows users to see patterns in the data they might otherwise have missed.
• Simple sharing of information. It is far easier to share data with charts, graphs, and infographics.
• Makes data accessible to non-technical users. With visualization, you no longer need to be a mathematician to understand the data insights.
• Easy to remember. Charts and graphs are not only easier to digest; they also tend to stay in the memory more easily than lists of numbers and statistics.
• Increase revenue. When all the decision-makers have the information at their fingertips, it empowers management to make quick and accurate decisions.

Problems with data visualization

• Information still needs to be accurate. Great visualizations [/blog/8-ways-turn-good-data-great-visualizations/] don’t make up for bad data. If best practices are not followed then visualizations can fall into the trap of becoming style over substance.
• Data visualization is an investment. Companies that want to effectively organize and visualize their data, or provide this ability to their customers, will either need a lot of involvement from analytics engineers (if they have the resources)), or an integrated analytics solution. Neither of these options comes without its costs, and pricing can vary depending on requirements. This then raises the question of whether to build the analytics solution in-house or buy off the shelf .
• Correlation does not equal causation. Visualizations often show the correlation between two or more metrics, so users often assume causation. But just because there is a correlation it doesn’t necessarily mean that one is caused by the other. There may be several other factors at play that aren’t included in the visualization.
• Users can still misinterpret the information. While visualization makes it easier for users to absorb data, it is still open to misinterpretation. For example, users might focus on the wrong thing when viewing it. This once again highlights the importance of using the right visualization type for the data displayed and the desired outcome.
• Confusing visualizations. Visualizations are supposed to simplify data, but if done badly they can make matters even more complicated. Perhaps the wrong chart type has been chosen, or there is too much information — as in the picture below.

The tools required to visualize data will depend on the project and the complexities of what you need to achieve. If you’d like to visualize some basic data for a presentation, you can use Excel to create some simple charts and graphs. If your data is more complex, you’ll need to create the insights first, which may involve data analysis or data mining. To achieve this, you will likely need to learn a programming language like Python. Alternatively, you can invest in a BI solution(s) such as those listed below.

• Chart generators or plugins: These tend to be used by developers and data engineers because the software requires a more advanced level of expertise. The plugins have many visualization types to choose from and there may even be a data-processing API that allows you to create actionable insights from your data. These tools usually have the capability to categorize and analyze basic data, and so can be used as the foundation of a company’s BI platform.
• Visualization reporting software: This is most often used by report developers and BI engineers. The software creates business and data analysis reports, which can then be turned into visualizations using a selection of built-in charts.
• A fully integrated BI and analytics solution: As the name suggests, this is the most complete solution. A good BI platform will allow you to easily explore data on your own, and create interactive dashboards and charts via a user-friendly no-code UI. The top solutions offer plug-and-play integrations, no-code tools, and flexible embedding options (such as React, Iframes, and Web Components) that allow you to seamlessly embed visualizations and dashboards into your product in a way that matches the brand.

To find out more about some of the best BI solutions with visualization tools, check out Gartner’s Magic Quadrant , which provides an objective analysis of the market leaders.

Data visualization has become so omnipresent that we use it every day without even noticing. Weather maps, bus schedules, computer audio levels, and fitness trackers – all employ visualizations to provide information in a more palatable way.

The above are just some of the ways companies use customer-facing visualizations to improve UX. But they also call upon data visualization to make internal decisions, for example, when looking at how to improve the supply chain, checking product sales across different countries, or how a specific marketing campaign or project is performing. Data visualization is also useful to get a better sense of external factors that might impact the company, such as the economic climate.

Data visualization use cases across different industries

Data visualization is used across all industries, from banking to healthcare . Below we look at how it can be used to improve processes in four different sectors.

• Software as a service (SAAS): Data is the beating heart of most software companies, and many of these embed analytics into their applications. They can then share the data with their customers through visualizations in a user-friendly way. A good example of this is Zendesk, whose software is all about improving the relationship between a business and its customers. Using advanced analytics and visualization tools, Zendesk gives their customers immediate access to the insights they need, which in turn speeds up response times and increases satisfaction among the clients of their customers.
• E-commerce: Data visualization helps brands and suppliers streamline their logistics and supply chains and optimize their operations. An e-commerce platform might use visualization to both improve the consumer experience and optimize brand performance. They could, for example, use their data to get a better understanding of how shoppers are using their site, helping them to better target customers and grow their business. Visualization is necessary to share these customer insights both internally and externally.
• Financial services: Financial firms leverage data and analytics to meet compliance requirements, manage risk, improve efficiency, and grow their business. Time is money, so the quicker they can make insights available to their employees and partners, the better. Visualization can also be used to help mitigate risk in real-time, and create highly personalized experiences for customers within financial service products.
• Insurance: Insurance firms rely on analysts to create impactful insights. These insights are then shared across the enterprise with people such as adjusters, underwriters, and marketers. Data visualization is crucial for insurance companies as it allows them to present clear insights, increase speed, and reduce inaccuracy.

With the right analytics platform, it is easy to create insights from your data and beautifully visualize them on dashboards. So why do some visualizations fail to achieve what they set out to do? The simple answer is that you need to be intentional about the information you want to convey. Good data visualization requires thoughtful human design. To that end, it can be useful to ask yourself the following questions to ensure best practice.

• What story am I trying to tell?

Visualization is a form of storytelling , only you’re using visuals instead of words. You need to find a clear way to tell the story to ensure it has the greatest impact. Data metrics and attributes must be relevant to the story that you are trying to tell. If you’re allowing users to filter, consider carefully how this will allow them to create their own stories from one source of truth.

• Have I designed the visualization for the viewer’s eye?

Once you’re clear on who the visualization is designed for and what you want them to take from it, you can focus on UX. Big-picture clarity is crucial, but it’s important not to neglect the details either. For example, what do the colors you’re using mean? Red usually implies danger, green is considered more positive. Can you customize the visualization to fit the brand? What about shortening numbers to make them easier to read (e.g., using K, M, B for thousands, millions, billions)?

• Am I trying to display too much data?

When it comes to data visualization, less is more. Although it’s tempting to present as many insights as possible, there is a fine line to be walked between insightful and overwhelming.

• Have I provided context?

Without a goal or a benchmark, users may miss the key insights. It is important to standardize benchmarks so that comparisons are not being made between two different things – which can lead to misleading takeaways.

The above are just some of the questions you might consider when creating visualizations. For a deeper dive, check out our article on visualization best practices .

Our analytics platform allows you to fully integrate real-time customized dashboards into your app or product. To see how this works and get your questions instantly answered, book a demo to help you decide whether this is the right tool for your needs. Alternatively, for some first-hand experience with data visualization, sign up for a free GoodData trial .

To learn more about data visualization and how it can help your business, you might like to check out some of our other resources:

Interactive vs. non-interactive data visualizations

7 ways to create great data visualizations

Related content

Subscribe to our newsletter

Get your dose of interesting facts on analytics in your inbox every month.

Subscribe to the GoodData Newsletter

Receive regular updates from our blog, news, and more.

Thank you for subscribing to the GoodData newsletter.

• Reviews / Why join our community?
• For companies
• Frequently asked questions

Information Visualization

What is information visualization.

Information visualization is the process of representing data in a visual and meaningful way so that a user can better understand it. Dashboards and scatter plots are common examples of information visualization. Via its depicting an overview and showing relevant connections, information visualization allows users to draw insights from abstract data in an efficient and effective manner.

Information visualization plays an important role in making data digestible and turning raw information into actionable insights. It draws from the fields of human-computer interaction, visual design, computer science, and cognitive science, among others. Examples include world map-style representations, line graphs, and 3-D virtual building or town plan designs.

The process of creating information visualization typically starts with understanding the information needs of the target user group. Qualitative research (e.g., user interviews) can reveal how, when, and where the visualization will be used. Taking these insights, a designer can determine which form of data organization is needed for achieving the users’ goals. Once information is organized in a way that helps users understand it better—and helps them apply it so as to reach their goals—visualization techniques are the next tools a designer brings out to use. Visual elements (e.g., maps and graphs) are created, along with appropriate labels, and visual parameters such as color, contrast, distance, and size are used to create an appropriate visual hierarchy and a visual path through the information.

Information visualization is becoming increasingly interactive, especially when used in a website or application. Being interactive allows for manipulation of the visualization by users, making it highly effective in catering to their needs. With interactive information visualization, users are able to view topics from different perspectives, and manipulate their visualizations of these until they reach the desired insights. This is especially useful if users require an explorative experience.

Questions related to Information Visualization

There are many types of information visualization . And different types cater to diverse needs. The most common forms include charts, graphs, diagrams, and maps. Charts, like bar graphs, succinctly display data trends. Diagrams, such as flowcharts, convey processes. Maps visually represent spatial information, enhancing geographical insights. 

Each type serves a unique purpose, offering a comprehensive toolkit for effective information representation.

Information visualization and data visualization share a connection but diverge in scope. Data visualization centers on graphically representing raw data using charts or graphs. Information visualization extends beyond raw data, embracing a comprehensive array of contextual details and intricate datasets. It strives for a complete presentation, often employing interactivity to convey insights. 

Data visualization concentrates on visually representing data points. Conversely, information visualization adopts a holistic approach. It considers the context for deeper comprehension and decision-making. 

This video illustrates this concept using a routine example. It highlights the creative process and the importance of capturing and structuring ideas for effective communication.

• Transcript loading…

Information visualization and infographics play unique roles. Human memory is visual, often remembering images and patterns more than raw data. Information visualization capitalizes on this aspect. It simplifies complex data through graphics for better understanding. 

This article gives valuable insights into the properties of human memory and their significance for information visualization .

Infographics portray information in engaging formats, often for storytelling or marketing. Both use visuals, but information visualization prioritizes clarity for users and turning data into usable insights. However, the latter focuses on effective communication and engagement.

No, Information Design and data visualization are distinctive in their objectives and applications. Information Design is a broader concept. It helps organize and present information to improve communication in the bigger picture. It considers the text, images, and layout to convey information effectively. 

On the other hand, data visualization translates raw data into graphical representations. It extracts meaningful insights and patterns. The approach focuses on visual elements to simplify the analysis of complex datasets.

Information visualization is a process that transforms complex data into easy-to-understand visuals. The seven stages include: 

Data collection: Gathering relevant data from diverse sources to form the basis for visualization.

Data analysis: Examining and processing the collected data to identify patterns, trends, and insights.

Data pre-processing: Cleaning and organizing the data to make it suitable for visualization.

Visual representation: Choosing appropriate visualization techniques to represent data accurately and effectively.

Interaction design: Developing user-friendly interfaces that allow meaningful interaction with the visualized data.

Interpretation: Enabling users to interpret and derive insights from the visualized information.

Evaluation: Assessing the effectiveness of the visualization in conveying information and meeting objectives.

This article provides a comprehensive overview of the data analysis process and explores key techniques for analysis. 

Information visualization helps people understand data and make decisions. It turns complicated data into easy-to-understand visuals. This makes it easier to see patterns and get a good overall picture. It also helps people communicate by showing information in a visually exciting way. Visualizations empower individuals to interact with data, enhancing engagement and enabling deeper exploration. Additionally, visual representations facilitate easier retention and recall of information.

Data visualization has advantages and disadvantages. One big challenge is misinterpretation. The visualization of data can be misleading if presented inappropriately. It can also lead to false conclusions, especially for those who do not understand the information.

Another major problem is too much information, as this article explains: Information Overload, Why it Matters, and How to Combat It . A crowded or complex visualization can overwhelm users and make communicating difficult.

Also, making good visualizations takes time and skill. This can sometimes be challenging for newbies.

Data visualization is a powerful tool. Creating valuable and impactful visualizations requires a combination of skills. You must understand the data, choose suitable visualization methods, and tell a compelling story . All this requires a good understanding of data and design, as explained in this video.

Interpreting complex data and choosing compelling visualizations can be challenging for beginners. However, leveraging available resources and enhancing skills can simplify data visualization despite the occasional difficulty.

Check out this course to learn more about Information Visualization . The course also explains the connection between the eye and the brain in creating images. It looks at the history of information visualization, how it has evolved, and common mistakes that you must avoid in visual perception.

It will teach you how to design compelling information visualizations and use various techniques for your projects.

Literature on Information Visualization

Here’s the entire UX literature on Information Visualization by the Interaction Design Foundation, collated in one place:

Learn more about Information Visualization

Take a deep dive into Information Visualization with our course Information Visualization .

Information visualization skills are in high demand, partly thanks to the rise in big data. Tech research giant Gartner Inc. observed that digital transformation has put data at the center of every organization. With the ever-increasing amount of information being gathered and analyzed, there’s an increasing need to present data in meaningful and understandable ways.

In fact, even if you are not involved in big data, information visualization will be able to help in your work processes as a designer. This is because many design processes—including conducting user interviews and analyzing user flows and sales funnels—involve the collation and presentation of information. Information visualization turns raw data into meaningful patterns, which will help you find actionable insights. From designing meaningful interfaces, to processing your own UX research, information visualization is an indispensable tool in your UX design kit.

This course is presented by Alan Dix, a former professor at Lancaster University in the UK. A world-renowned authority in the field of human-computer interaction, Alan is the author of the university-level textbook Human-Computer Interaction . “Information Visualization” is full of simple but practical lessons to guide your development in information visualization. We start with the basics of what information visualization is, including its history and necessity, and then walk you through the initial steps in creating your own information visualizations. While there’s plenty of theory here, we’ve got plenty of practice for you, too.

All open-source articles on Information Visualization

Information overload, why it matters and how to combat it.

• 1.1k shares
• 4 years ago

Visual Representation

How to Design an Information Visualization

How to Visualize Your Qualitative User Research Results for Maximum Impact

• 3 years ago

Preattentive Visual Properties and How to Use Them in Information Visualization

• 5 years ago

How to Conduct Focus Groups

The Properties of Human Memory and Their Importance for Information Visualization

• 7 years ago

Information Visualization – A Brief Introduction

Visual Mapping – The Elements of Information Visualization

Guidelines for Good Visual Information Representations

How to Show Hierarchical Data with Information Visualization

Information Visualization – An Introduction to Multivariate Analysis

• 8 years ago

Information Visualization – Who Needs It?

How to Display Complex Network Data with Information Visualization

Vision and Visual Perception Challenges

Information Visualization an Introduction to Transformable Information Representations

The Principles of Information Visualization for Basic Network Data

The Continuum of Understanding and Information Visualization

• 6 years ago

Information Visualization – A Brief Pre-20th Century History

Information Visualization an Introduction to Manipulable Information Representations

Open Access—Link to us!

We believe in Open Access and the  democratization of knowledge . Unfortunately, world-class educational materials such as this page are normally hidden behind paywalls or in expensive textbooks.

If you want this to change , cite this page , link to us, or join us to help us democratize design knowledge !

Privacy Settings

Our digital services use necessary tracking technologies, including third-party cookies, for security, functionality, and to uphold user rights. Optional cookies offer enhanced features, and analytics.

Experience the full potential of our site that remembers your preferences and supports secure sign-in.

Governs the storage of data necessary for maintaining website security, user authentication, and fraud prevention mechanisms.

Enhanced Functionality

Saves your settings and preferences, like your location, for a more personalized experience.

Referral Program

We use cookies to enable our referral program, giving you and your friends discounts.

Error Reporting

We share user ID with Bugsnag and NewRelic to help us track errors and fix issues.

Optimize your experience by allowing us to monitor site usage. You’ll enjoy a smoother, more personalized journey without compromising your privacy.

Analytics Storage

Collects anonymous data on how you navigate and interact, helping us make informed improvements.

Differentiates real visitors from automated bots, ensuring accurate usage data and improving your website experience.

Lets us tailor your digital ads to match your interests, making them more relevant and useful to you.

Advertising Storage

Stores information for better-targeted advertising, enhancing your online ad experience.

Personalization Storage

Permits storing data to personalize content and ads across Google services based on user behavior, enhancing overall user experience.

Advertising Personalization

Allows for content and ad personalization across Google services based on user behavior. This consent enhances user experiences.

Enables personalizing ads based on user data and interactions, allowing for more relevant advertising experiences across Google services.

Receive more relevant advertisements by sharing your interests and behavior with our trusted advertising partners.

Enables better ad targeting and measurement on Meta platforms, making ads you see more relevant.

Allows for improved ad effectiveness and measurement through Meta’s Conversions API, ensuring privacy-compliant data sharing.

LinkedIn Insights

Tracks conversions, retargeting, and web analytics for LinkedIn ad campaigns, enhancing ad relevance and performance.

LinkedIn CAPI

Enhances LinkedIn advertising through server-side event tracking, offering more accurate measurement and personalization.

Google Ads Tag

Tracks ad performance and user engagement, helping deliver ads that are most useful to you.

Share Knowledge, Get Respect!

or copy link

Cite according to academic standards

Simply copy and paste the text below into your bibliographic reference list, onto your blog, or anywhere else. You can also just hyperlink to this page.

New to UX Design? We’re Giving You a Free ebook!

Download our free ebook The Basics of User Experience Design to learn about core concepts of UX design.

In 9 chapters, we’ll cover: conducting user interviews, design thinking, interaction design, mobile UX design, usability, UX research, and many more!

What Is Data Visualization and Why Is It Important?

The sheer amount of data generated today means we need new ways to understand what’s happening in order to take action faster. Every click, transaction, subscription, loyalty card swipe, and social media interaction contributes to a digital footprint that continues to grow exponentially. The result? A massive explosion of data that is revolutionizing the way we live and work. Data visualization, in particular, plays a critical role in presenting data in a meaningful and understandable format. By using a visual representation of data , it’s much easier to identify patterns, trends, and relationships that may not be immediately apparent when sifting through large data sets.

Here’s what we’ll cover in this guide to data visualization: 

• Data Visualization Definition 

Benefits of Data Visualization

Why data visualization is important .

• Types of Data Visualization and Examples
• Evaluating Data Visualization Tools
• Take the Next Step and Start Analyzing With Data Visualization 

‍ Data Visualization Definition

Data visualization is the process of transforming raw data into visual formats, such as charts, graphs, or maps, to help identify patterns, trends, and insights that might not be apparent from numerical data alone. 

Additionally, it enables data to be more accessible, understandable, and impactful, especially when communicating with stakeholders, investors, or team members who may not be familiar with the data.

For example, data visualization could help:

• In retail, gaining insights into customer behavior, purchase patterns, and product performance.
• In finance, monitoring market trends, tracking portfolio performance, and conducting risk analysis. 
• In public health, showing the geographical distribution of outbreaks and helping track the spread of infectious diseases.
• In supply chain industries, tracking inventory levels, monitoring logistics operations, and optimizing resource allocation. 
• In sports, evaluating player performance, game strategies, and match statistics.
• In education, tracking student performance, analyzing learning outcomes, and identifying areas for improvement.

Data visualization has several benefits for businesses including: the ability to process information faster, identify trends at scale, and make data more digestible. Companies regularly use data to make decisions, and through data visualization, can find insights quickly and move to action. Data visualization specifically helps with the following:

• Visualizing patterns and relationships
• Storytelling, including specifically data storytelling
• Accessibility to information 

Exploration

Let’s take a look at each of these benefits in detail. 

‍ Visualize patterns and relationships

Data visualization constitutes an excellent method for the discernment of interconnections and patterns amidst vast collections of information. For example, a scatter plot can be used to display the relationship between two variables, such as the correlation between temperature and sales. This enables users to understand the relationship and identify trends and outliers more quickly and easily.

Read a guide of Sigma’s visual library.

Storytelling

Your audience, whether it's coworkers or clients, want to hear a coherent story from your data. Storytelling with data cannot be done successfully without visualizations. Colorful charting and dynamic pivots are just as important as characters and plots are in a traditional story, so using them to communicate information makes data that much more engaging and memorable for audiences. Data can be complex and convoluted for some audiences, so data storytelling is an approach to convey important information effectively through a captivating narrative. Good visualizations are a vital part of that narrative.  

For example, if an analyst is investigating the performance of e-commerce sales for their retail company over time, they may leverage several data sources such as spreadsheets, calculations, code, etc. to do so. However, when they report these new insights to their stakeholders, the analyst will need to summarize and communicate their findings in a digestible way. 

An easy way the analyst could do this is by using the data to create a map of the U.S. with a color gradient overlaying every state that is lighter or darker based on its total sales volume. This visual story tells the least and most successful retail locations at a glance.

Accessibility / Easily Share Information

Data visualization serves as an invaluable mechanism for the facilitation of accessibility, allowing for the communication of information amongst individuals, even for those who may not usually engage with data , which broadens the audience.

Visualizations help simplify complex information by leveraging people’s ability to naturally recognize patterns. A viewer typically does not have to be taught that bigger means more and that smaller means less. In a case where an analyst wants to highlight the difference in scale between one product’s profitability vs. another, a bar chart can clearly show the user which product is more profitable and by how much, making it easy for even non-technical team members to understand and compare the performance of different products.

Exploration is a key component of successful data visualization. The more flexible charting and dashboarding is, the more follow-up questions end users can ask directly of their data. For example, an interactive dashboard can be used to explore retail sales data over time, enabling users to filter and drill down into the data to identify trends and patterns.

Data visualization exploration is often associated with the concept of “drill downs.” Drill downs in data visualization refer to the process of starting with an overview of data and then narrowing the focus to more specific aspects of it. As an example, one might start with a visualization of global climate data and drill down to data about a specific country, a specific state, a specific city, or even a specific neighborhood within that city. Each drill down reveals more precise, detailed, and nuanced information. 

The main goal of data visualization is that it helps unify and bring teams onto the same page. The human mind is wired to grasp visual information more effortlessly than raw data in spreadsheets or detailed reports. Thus, graphical representation of voluminous and intricate data is more user-friendly. Data visualization offers a swift and straightforward method to communicate ideas in a universally understood format, with the added benefit of enabling scenario testing through minor modifications.

By translating information into visual form, it ensures everyone, irrespective of the complexity of the data or the depth of the analysis, can share a unified understanding. Any industry can benefit from using data visualization, because pretty much every industry relies on data to power it. That includes finance, marketing, consumer goods, education, government, sports, history, and many more. ‍ Another thing to keep in mind is that data visualization can be a double-edged sword. For example, charts can be manipulated and skewed to force a desired outcome. Ungoverned, static, desktop tools can become the wild west in suggesting an inaccurate outcome “proven by data.” Even in the cases where the visualization builder is acting in good faith, there are still pitfalls to watch out for. Always be considerate of:

• Individual outliers having an outsized impact, skewing the visual direction of a chart
• The need for for business users to see the underlying data
• Allowing for transparency down to row-level detail in data sets

Types of Data Visualizations & Examples

There is a long list of types of data visualization techniques and methods that can be used to represent data. While no type of data visualization is perfect, we’ll walk through different examples and when to apply each one. 

We’ll be looking at:

• Line charts and area charts
• Scatter plots 
• Pivot tables
• Box-and-whisker plots
• Sankey charts 

Tables, although more commonly thought of as a data source, can also be considered a type of data visualization. Especially when conditional formatting is applied to the table’s rows and columns, the data within the table becomes more visually engaging and informative. With conditional formatting, important insights and patterns can be highlighted, making it easier for viewers to identify trends and outliers at a glance. Additionally, tables offer a structured and organized way to present information, allowing for a comprehensive comparison of data points, which further enhances data understanding and analysis. ‍ For example, Sigma’s UI is based on a spreadsheet-like interface, which means almost everything in Sigma begins in a table format. That said, you can also create visual tables that display a smaller amount of data in order to tell a clearer story. In data visualization, tables are a simplified way of representing this interface. 

When to use tables:

• For detailed numeric comparisons, or when precision of data is key
• For displaying multidimensional data; tables can handle this complexity quite well

When to avoid tables: 

• When patterns, trends, or relationships need to be highlighted at a glance
• When dealing with large amounts of data

Pie charts —similar to stacked bar charts—are useful for displaying categorical data, such as market share or customer demographics. Pie charts are often used to display data that can be divided into categories or subgroups, and to show how each category or subgroup contributes to the whole. For example, a pie chart could be used to show the proportion of sales for different product categories in a given period of time, or the percent of a company's revenue broken down by various regions.

When to use pie charts:

• You want to display a proportion or percentage of a whole
• You’re visualizing only seven categories or less

When to avoid pie charts:

• You’re visualizing more than seven categories
• You want to compare something with more details, rather than just proportion
• You want to display and pinpoint exact values 

A bar chart, or bar graph, constitutes a variety of graphs that employ rectangular bars to depict data. These bars can be oriented either horizontally or vertically, with their extent being directly proportional to the numerical values they are intended to embody. Predominantly utilized for juxtaposing data across disparate categories or illustrating shifts in data over temporal progressions, bar charts offer a straightforward, yet potent means of conveying information visually. They frequently function as the initial tool in the exploratory process of data investigation.

When to use bar charts:

• Emphasizing and contrasting different sets of data, making the disparities or similarities between categories clear
• To display a subset of a larger dataset

When to avoid bar charts: 

• When a particular field encompasses an overwhelming variety of data types
• When the differences between fields are too subtle, or when these differences exist on different scales, as it could lead to confusion or misinterpretation

Line Charts & Area Charts

Line charts and area charts are two types of charts that are commonly used to visualize data trends over time. A line chart, also called a line graph, is a distinct type of graphical representation that exhibits information in the form of a multitude of data points, which are interconnected by unbroken lines. These line charts are typically employed to demonstrate transformations in data over a certain duration, where the horizontal axis symbolizes time, and the vertical axis signifies the values under scrutiny. Furthermore, they can serve to juxtapose several series of data within the same chart, or to graphically illustrate predicted time periods. 

For example, a line chart can be used to visualize a company's stock prices over the course of a year. Similarly, an area chart can be used to visualize the temperature changes over a day.

When to use line charts:

• When you’re displaying time-based continuous data 
• When you have multiple series or larger datasets 

When to avoid line charts:

• When you have smaller datasets, bar charts are likely a better way to present the information 
• Avoid when you need to compare multiple categories at once

When to use area charts:

• When you want to display the volume of the data you have 
• When comparing data across more than one time period 

When to avoid area charts:

• Avoid if you need to compare multiple categories, as well as when you need to examine the specific data value

Scatter Plots

A scatter plot , also called a scatter chart or scatter graph, is a specialized form of chart that demonstrates the correlation between two distinct variables by mapping them as a succession of individual data points. Each data point denotes a combined value of the two variables, with its specific placement within the chart dictated by these values.

Scatter charts prove instrumental in discerning patterns and trends within data, and they also help us understand how strong and in what direction the relationship is between two variables. They also serve as effective tools for identifying outliers, or those data points that deviate significantly from anticipated values based on the pattern displayed by other data points. These charts find widespread use across a range of fields including, but not limited to, statistics, engineering, and social sciences, for the purpose of analyzing and visualizing intricate data sets. In the realm of business, they are frequently utilized to identify correlations between different variables, for instance, examining the relationship between marketing outlays and resultant sales revenue. ‍ For example, a scatter plot might be used to visualize the relationship between the age and income of a group of people. Another example would be to plot the correlation between the amount of rainfall and the crop yield for a particular region.

When to use scatter plots:

• Highlight correlations within your data
• They are useful tools for statistical investigations
• Consider scatter plots to reveal underlying patterns or trends

When to avoid scatter plots:

• For smaller datasets, scatter plots may not be optimal
• Avoid scatter plots for excessively large datasets to prevent unintelligible data clustering
• If your data lacks correlations, scatter plots may not be the best choice

Pivot Tables

While pivot tables may not be what first comes to mind for data visualization, they can give important context with hard numbers and provide strong visual indicators through formatting. ‍ Pivot tables can also be enhanced with conditional formatting to provide color scales that make performance trends more visible. Data bars can also be added to cells to run either red or green for positive and negative values. 

When to Use Pivot Tables:

• Cohort analysis performance trends or portfolio analysis with a mix of positive and negative values

What Not to Use Pivot Tables:

• When your dataset is too large to get a good understanding of the whole
• When data can easily be summarized with a bar chart instead

An example of a pivot table, where colors are used to show positive or negative progress on a company’s portfolio. The user can pivot the table to show multiple categories in different ways.

A heat map is a type of chart that uses color to represent data values. It is often used to visualize data that is organized in a matrix or table format. The color of each cell in the matrix is determined by the value of the corresponding data point. Heat maps are best used when analyzing data that is organized in a two-dimensional grid or matrix.

For example, a heat map can be used to visualize a company's website traffic, where the rows represent different pages on the website, and the columns represent different periods of time.

When to use heat maps:

• When you need to visualize the density or intensity of variables
• When you want to display patterns or trends over time or space 

When to avoid heat maps:

• When precise values are needed; heat maps are better at showing relative differences rather than precise values
• When working with small data sets 

A tree map is a type of chart that is used to visualize hierarchical data. It consists of a series of nested rectangles, where the size and color of each rectangle represent a different variable. Tree maps are best used when analyzing data that has a hierarchical structure.

For example, a tree map can be used to visualize the market share of different companies in an industry. The largest rectangle would represent the entire industry, with smaller rectangles representing the market share of each individual company.

When to use tree maps:

• When you want to visualize hierarchical data
• When you need to illustrate the proportion of different categories within a whole 

When to avoid tree maps:

• When exact values are important
• When there are too many categories

Box-and-Whisker Plots

Box plots are useful for quickly summarizing the distribution of a dataset, particularly its central tendency and variability. For example, a box-and-whisker plot can be used to visualize the test scores of a group of students. 

Colloquially recognized as a box-and-whisker plot, a box plot is a distinct form of chart that showcases the distribution of a collection of numerical data through its quartile divisions. Box plots serve as efficient tools for rapidly encapsulating the distribution of a dataset, specifically its central propensity and variability. 

A box-and-whisker plot consists of a rectangle (the "box") and a pair of "whiskers" that extend from it. The box embodies the middle 50% of the data, with the lower boundary of the box signaling the first quartile (25th percentile) and the upper boundary of the box indicating the third quartile (75th percentile). The line situated within the box signifies the median value of the data. The whiskers project from the box to the minimum and maximum values of the data, or to a designated distance from the box referred to as the "fences." Any data points that reside outside the whiskers or fences are categorized as outliers and are plotted as individual points. When to use box plot charts:

• When you want to display data spread and skewness
• When showcasing the distribution of data, including the range, quartiles, and potential outliers
• When comparing multiple groups or categories side-by-side; they allow for easy comparison of different distributions.

When to avoid box plot charts:

• If you need to show more detail, since box plots focus on a high-level summary 
• When individual data points are important to the story you’re telling
• When your audience isn’t familiar with them, since they can sometimes be less intuitive than other types of visualizations

A histogram is a type of chart that displays the distribution of a dataset. It consists of a series of vertical bars, where the height of each bar represents the number of observations in a particular range. Histograms are best used when analyzing continuous data. It’s used the most when you want to understand the frequency distribution of a numerical variable, like height, weight, or age. For example, a histogram can be used to visualize the distribution of heights in a population. Read more about building histograms in Sigma here.

When to Use a Histogram:

• When understanding the shape of a distribution; for example, whether it’s symmetric, skewed to the left or right, or bimodal
• When identifying outliers, like which data points are significantly different from the rest of the data
• When comparing distribution of a variable across different groups, such as males and females, or different age groups.
• To set boundaries for data ranges; for example, you might use a histogram to determine what constitutes a "normal" or "abnormal" value for a particular variable

When to Avoid a Histogram:

• When you need to look at multiple dimensions at the same time
• If your data isn’t all on the same scale

Sankey Charts

We end our guide with the controversial Sankey chart. A Sankey chart is a type of diagram that illustrates the movement or transfer of data, resources, or quantities through various stages of a system or process. Common applications of Sankey charts include visualizing complex sequences like energy usage, material distribution, or even a website's user journey. The structure of the chart includes nodes and links—with nodes representing the starting points, endpoints, or intermediate steps, and links depicting the transition of quantities or data between these nodes.

The thickness of the links in a Sankey chart directly corresponds to the volume of data or resources being moved, offering an intuitive comparison of the relative sizes of these transfers. They can be invaluable for recognizing inefficiencies, bottlenecks, or potential areas for enhancement in a system or process. These charts serve as a powerful tool for communicating complex information in a straightforward and comprehensible way. However, if there are too many nodes or links, Sankey charts can become cluttered and challenging to interpret, hence their use should be considerate and targeted.

‍ When to use Sankey charts:

• When you want to show the data as part of a process

When to avoid Sankey charts:

• When it starts to feel too confusing, which can quickly happen when there are too many nodes or links
• When you need to see exact values, it might not be the most intuitive option. 

Evaluating Data Visualization Tools 

Data visualization tools have become increasingly popular in recent years, with a wide variety of options available to choose from. However, determining which tool best suits your needs can be challenging with so many options. When evaluating data visualization tools, there are several key questions to consider:

• What are your goals and needs?   It's crucial to clearly understand your goals and needs before selecting a data visualization tool. Are you looking to explore your data, communicate a specific message, or both? Understanding your objectives will help you choose the right tool for your project.
• What features do you require?   Different data visualization tools come with different features. Before selecting a tool, you should consider what features you need to achieve your goals. For example, do you require interactive capabilities or the ability to create custom visualizations?
• Where will your data come from?   The source of your data is another critical factor to consider when selecting a data visualization tool. Some tools are better suited for specific types of data, such as structured or unstructured data, while others may require specific file formats or data storage solutions.
• Where will you need to see your data?   Different data visualization tools may be more suitable for specific platforms or devices. For example, some tools may be optimized for mobile devices, while others are designed for desktop computers or specific web browsers. You may also be interested in embedding visualizations elsewhere , such as internal applications or external portals.
• Where would you like to publish your visualization?   Finally, consider where you would like to publish your visualization. Some tools may provide built-in publishing capabilities, while others may require you to export your visualization to a separate platform. Selecting a tool that supports your publishing needs is important to ensure your visualization reaches your intended audience.

By considering these key questions, you can evaluate different data visualization tools and select the one that best meets your needs.

Read a side-by-side comparison of Sigma against similar BI tools.

Take the Next Step & Start Analyzing With Data Visualization

Data visualization is a powerful tool for understanding and communicating complex data. While there are many data visualization tools on the market, Sigma offers an intuitive and familiar spreadsheet interface that allows users to easily explore, analyze, and collaborate on their data. 

Explore Sigma’s capabilities and start transforming your data today via a free trial of Sigma .

Related resources

This is the default text value

Watch on-DEMAND DEMOS

ATTEND A LIVE LAB

Get a free trial

SEE WORKBOOK EXAMPLES

JOIN THE COMMUNITY

SCHEDULE A CALL

JOIN THE SIGMA COMMUNITY

• Business Essentials
• Leadership & Management
• Credential of Leadership, Impact, and Management in Business (CLIMB)
• Entrepreneurship & Innovation
• Digital Transformation
• Finance & Accounting
• Business in Society
• For Organizations
• Support Portal
• Media Coverage
• Founding Donors
• Leadership Team

• Harvard Business School →
• HBS Online →
• Business Insights →

Business Insights

Harvard Business School Online's Business Insights Blog provides the career insights you need to achieve your goals and gain confidence in your business skills.

• Career Development
• Communication
• Decision-Making
• Earning Your MBA
• Negotiation
• News & Events
• Productivity
• Staff Spotlight
• Student Profiles
• Work-Life Balance
• AI Essentials for Business
• Alternative Investments
• Business Analytics
• Business Strategy
• Business and Climate Change
• Design Thinking and Innovation
• Digital Marketing Strategy
• Disruptive Strategy
• Economics for Managers
• Entrepreneurship Essentials
• Financial Accounting
• Global Business
• Launching Tech Ventures
• Leadership Principles
• Leadership, Ethics, and Corporate Accountability
• Leading Change and Organizational Renewal
• Leading with Finance
• Management Essentials
• Negotiation Mastery
• Organizational Leadership
• Power and Influence for Positive Impact
• Strategy Execution
• Sustainable Business Strategy
• Sustainable Investing
• Winning with Digital Platforms

6 Data Visualization Examples To Inspire Your Own

• 12 Jan 2017

Data informs virtually every business decision an organization makes. Because of this, it’s become increasingly important for professionals of all backgrounds to be adept at working with data.

While data can provide immense value, it’s important that professionals are able to effectively communicate the significance of the data to stakeholders. This is where data visualization comes into play. By transforming raw data into engaging visuals using various data visualization tools , it’s much easier to communicate insights gleaned from it.

Here are six real-world examples of data visualization that you can use to inspire your own.

What Is Data Visualization?

Data visualization is the process of turning raw data into graphical representations.

Visualizations make it easy to communicate trends in data and draw conclusions. When presented with a graph or chart, stakeholders can easily visualize the story the data is telling, rather than try to glean insights from raw data.

There are countless data visualization techniques , including:

• Scatter plots

The technique you use will vary based on the type of data you’re handling and what you’re trying to communicate.

6 Real-World Data Visualization Examples

1. the most common jobs by state.

Source: NPR

National Public Radio (NPR) produced a color-coded, interactive display of the most common jobs in each state in each year from 1978 to 2014. By dragging the scroll bar at the bottom of the map, you’re able to visualize occupational changes over time.

If you’re trying to represent geographical data, a map is the best way to go.

2. COVID-19 Hospitalization Rates

Source: CDC

Throughout the COVID-19 pandemic, the Centers for Disease Control and Prevention (CDC) has been transforming raw data into easily digestible visuals. This line graph represents COVID-19 hospitalization rates from March through November 2020.

The CDC tactfully incorporated color to place further emphasis on the stark increase in hospitalization rates, using a darker shade for lower values and a lighter shade for higher values.

3. Forecasted Revenue of Amazon.com

Source: Statista

Data visualizations aren’t limited to historical data. This bar chart created by Statista visualizes the forecasted gross revenue of Amazon.com from 2018 to 2025.

This visualization uses a creative title to summarize the main message that the data is conveying, as well as a darker orange color to spike out the most important data point.

4. Web-Related Statistics

Source: Internet Live Stats

Internet Live Stats has tracked web-related statistics and pioneered methods for visualizing data to show how different digital properties have ebbed and flowed over time.

Simple infographics like this one are particularly effective when your goal is to communicate key statistics rather than visualizing trends or forecasts.

5. Most Popular Food Delivery Items

Source: Eater

Eater, Vox’s food and dining brand, has created this fun take on a “pie” chart, which shows the most common foods ordered for delivery in each of the United States.

To visualize this data, Eater used a specific type of pie chart known as a spie chart. Spie charts are essentially pie charts in which you can vary the height of each segment to further visualize differences in data.

6. Netflix Viewing Patterns

Source: Vox

Vox created this interesting visualization depicting the viewing patterns of Netflix users over time by device type. This Sankey diagram visualizes the tendency of users to switch to streaming via larger device types.

Visualizing Data to Make Business Decisions

The insights and conclusions drawn from data visualizations can guide the decision-making and strategic planning processes for your organization.

To ensure your visualizations are relevant, accurate, and ethical, familiarize yourself with basic data science concepts . With a foundational knowledge in data science, you can maintain confidence in your data and better understand its significance. An online analytics course can help you get started.

Are you interested in improving your data science and analytical skills? Download our Beginner’s Guide to Data & Analytics to learn how you can leverage the power of data for professional and organizational success.

This post was updated on February 26, 2021. It was originally published on January 12, 2017.

An official website of the United States government

The .gov means it’s official. Federal government websites often end in .gov or .mil. Before sharing sensitive information, make sure you’re on a federal government site.

The site is secure. The https:// ensures that you are connecting to the official website and that any information you provide is encrypted and transmitted securely.

• Publications
• Account settings

Preview improvements coming to the PMC website in October 2024. Learn More or Try it out now .

• Advanced Search
• Journal List
• Patterns (N Y)
• v.1(9); 2020 Dec 11

Principles of Effective Data Visualization

Stephen r. midway.

1 Department of Oceanography and Coastal Sciences, Louisiana State University, Baton Rouge, LA 70803, USA

We live in a contemporary society surrounded by visuals, which, along with software options and electronic distribution, has created an increased importance on effective scientific visuals. Unfortunately, across scientific disciplines, many figures incorrectly present information or, when not incorrect, still use suboptimal data visualization practices. Presented here are ten principles that serve as guidance for authors who seek to improve their visual message. Some principles are less technical, such as determining the message before starting the visual, while other principles are more technical, such as how different color combinations imply different information. Because figure making is often not formally taught and figure standards are not readily enforced in science, it is incumbent upon scientists to be aware of best practices in order to most effectively tell the story of their data.

The Bigger Picture

Visuals are an increasingly important form of science communication, yet many scientists are not well trained in design principles for effective messaging. Despite challenges, many visuals can be improved by taking some simple steps before, during, and after their creation. This article presents some sequential principles that are designed to improve visual messages created by scientists.

Many scientific visuals are not as effective as they could be because scientists often lack basic design principles. This article reviews the importance of effective data visualization and presents ten principles that scientists can use as guidance in developing effective visual messages.

Introduction

Visual learning is one of the primary forms of interpreting information, which has historically combined images such as charts and graphs (see Box 1 ) with reading text. 1 However, developments on learning styles have suggested splitting up the visual learning modality in order to recognize the distinction between text and images. 2 Technology has also enhanced visual presentation, in terms of the ability to quickly create complex visual information while also cheaply distributing it via digital means (compared with paper, ink, and physical distribution). Visual information has also increased in scientific literature. In addition to the fact that figures are commonplace in scientific publications, many journals now require graphical abstracts 3 or might tweet figures to advertise an article. Dating back to the 1970s when computer-generated graphics began, 4 papers represented by an image on the journal cover have been cited more frequently than papers without a cover image. 5

Regarding terminology, the terms graph , plot , chart , image , figure , and data visual(ization) are often used interchangeably, although they may have different meanings in different instances. Graph , plot , and chart often refer to the display of data, data summaries, and models, while image suggests a picture. Figure is a general term but is commonly used to refer to visual elements, such as plots, in a scientific work. A visual , or data visualization , is a newer and ostensibly more inclusive term to describe everything from figures to infographics. Here, I adopt common terminology, such as bar plot, while also attempting to use the terms figure and data visualization for general reference.

There are numerous advantages to quickly and effectively conveying scientific information; however, scientists often lack the design principles or technical skills to generate effective visuals. Going back several decades, Cleveland 6 found that 30% of graphs in the journal Science had at least one type of error. Several other studies have documented widespread errors or inefficiencies in scientific figures. 7 , 8 , 9 In fact, the increasing menu of visualization options can sometimes lead to poor fits between information and its presentation. These poor fits can even have the unintended consequence of confusing the readers and setting them back in their understanding of the material. While objective errors in graphs are hopefully in the minority of scientific works, what might be more common is suboptimal figure design, which takes place when a design element may not be objectively wrong but is ineffective to the point of limiting information transfer.

Effective figures suggest an understanding and interpretation of data; ineffective figures suggest the opposite. Although the field of data visualization has grown in recent years, the process of displaying information cannot—and perhaps should not—be fully mechanized. Much like statistical analyses often require expert opinions on top of best practices, figures also require choice despite well-documented recommendations. In other words, there may not be a singular best version of a given figure. Rather, there may be multiple effective versions of displaying a single piece of information, and it is the figure maker's job to weigh the advantages and disadvantages of each. Fortunately, there are numerous principles from which decisions can be made, and ultimately design is choice. 7

The data visualization literature includes many great resources. While several resources are targeted at developing design proficiency, such as the series of columns run by Nature Communications , 10 Wilkinson's The Grammar of Graphics 11 presents a unique technical interpretation of the structure of graphics. Wilkinson breaks down the notion of a graphic into its constituent parts—e.g., the data, scales, coordinates, geometries, aesthetics—much like conventional grammar breaks down a sentence into nouns, verbs, punctuation, and other elements of writing. The popularity and utility of this approach has been implemented in a number of software packages, including the popular ggplot2 package 12 currently available in R. 13 (Although the grammar of graphics approach is not explicitly adopted here, the term geometry is used consistently with Wilkinson to refer to different geometrical representations, whereas the term aesthetics is not used consistently with the grammar of graphics and is used simply to describe something that is visually appealing and effective.) By understanding basic visual design principles and their implementation, many figure authors may find new ways to emphasize and convey their information.

The Ten Principles

Principle #1 diagram first.

The first principle is perhaps the least technical but very important: before you make a visual, prioritize the information you want to share, envision it, and design it. Although this seems obvious, the larger point here is to focus on the information and message first, before you engage with software that in some way starts to limit or bias your visual tools. In other words, don't necessarily think of the geometries (dots, lines) you will eventually use, but think about the core information that needs to be conveyed and what about that information is going to make your point(s). Is your visual objective to show a comparison? A ranking? A composition? This step can be done mentally, or with a pen and paper for maximum freedom of thought. In parallel to this approach, it can be a good idea to save figures you come across in scientific literature that you identify as particularly effective. These are not just inspiration and evidence of what is possible, but will help you develop an eye for detail and technical skills that can be applied to your own figures.

Principle #2 Use the Right Software

Effective visuals typically require good command of one or more software. In other words, it might be unrealistic to expect complex, technical, and effective figures if you are using a simple spreadsheet program or some other software that is not designed to make complex, technical, and effective figures. Recognize that you might need to learn a new software—or expand your knowledge of a software you already know. While highly effective and aesthetically pleasing figures can be made quickly and simply, this may still represent a challenge to some. However, figure making is a method like anything else, and in order to do it, new methodologies may need to be learned. You would not expect to improve a field or lab method without changing something or learning something new. Data visualization is the same, with the added benefit that most software is readily available, inexpensive, or free, and many come with large online help resources. This article does not promote any specific software, and readers are encouraged to reference other work 14 for an overview of software resources.

Principle #3 Use an Effective Geometry and Show Data

Geometries are the shapes and features that are often synonymous with a type of figure; for example, the bar geometry creates a bar plot. While geometries might be the defining visual element of a figure, it can be tempting to jump directly from a dataset to pairing it with one of a small number of well-known geometries. Some of this thinking is likely to naturally happen. However, geometries are representations of the data in different forms, and often there may be more than one geometry to consider. Underlying all your decisions about geometries should be the data-ink ratio, 7 which is the ratio of ink used on data compared with overall ink used in a figure. High data-ink ratios are the best, and you might be surprised to find how much non-data-ink you use and how much of that can be removed.

Most geometries fall into categories: amounts (or comparisons), compositions (or proportions), distributions , or relationships . Although seemingly straightforward, one geometry may work in more than one category, in addition to the fact that one dataset may be visualized with more than one geometry (sometimes even in the same figure). Excellent resources exist on detailed approaches to selecting your geometry, 15 and this article only highlights some of the more common geometries and their applications.

Amounts or comparisons are often displayed with a bar plot ( Figure 1 A), although numerous other options exist, including Cleveland dot plots and even heatmaps ( Figure 1 F). Bar plots are among the most common geometry, along with lines, 9 although bar plots are noted for their very low data density 16 (i.e., low data-ink ratio). Geometries for amounts should only be used when the data do not have distributional information or uncertainty associated with them. A good use of a bar plot might be to show counts of something, while poor use of a bar plot might be to show group means. Numerous studies have discussed inappropriate uses of bar plots, 9 , 17 noting that “because the bars always start at zero, they can be misleading: for example, part of the range covered by the bar might have never been observed in the sample.” 17 Despite the numerous reports on incorrect usage, bar plots remain one of the most common problems in data visualization.

Examples of Visual Designs

(A) Clustered bar plots are effective at showing units within a group (A–C) when the data are amounts.

(B) Histograms are effective at showing the distribution of data, which in this case is a random draw of values from a Poisson distribution and which use a sequential color scheme that emphasizes the mean as red and values farther from the mean as yellow.

(C) Scatterplot where the black circles represent the data.

(D) Logistic regression where the blue line represents the fitted model, the gray shaded region represents the confidence interval for the fitted model, and the dark-gray dots represent the jittered data.

(E) Box plot showing (simulated) ages of respondents grouped by their answer to a question, with gray dots representing the raw data used in the box plot. The divergent colors emphasize the differences in values. For each box plot, the box represents the interquartile range (IQR), the thick black line represents the median value, and the whiskers extend to 1.5 times the IQR. Outliers are represented by the data.

(F) Heatmap of simulated visibility readings in four lakes over 5 months. The green colors represent lower visibility and the blue colors represent greater visibility. The white numbers in the cells are the average visibility measures (in meters).

(G) Density plot of simulated temperatures by season, where each season is presented as a small multiple within the larger figure.

For all figures the data were simulated, and any examples are fictitious.

Compositions or proportions may take a wide range of geometries. Although the traditional pie chart is one option, the pie geometry has fallen out of favor among some 18 due to the inherent difficulties in making visual comparisons. Although there may be some applications for a pie chart, stacked or clustered bar plots ( Figure 1 A), stacked density plots, mosaic plots, and treemaps offer alternatives.

Geometries for distributions are an often underused class of visuals that demonstrate high data density. The most common geometry for distributional information is the box plot 19 ( Figure 1 E), which shows five types of information in one object. Although more common in exploratory analyses than in final reports, the histogram ( Figure 1 B) is another robust geometry that can reveal information about data. Violin plots and density plots ( Figure 1 G) are other common distributional geometries, although many less-common options exist.

Relationships are the final category of visuals covered here, and they are often the workhorse of geometries because they include the popular scatterplot ( Figures 1 C and 1D) and other presentations of x - and y -coordinate data. The basic scatterplot remains very effective, and layering information by modifying point symbols, size, and color are good ways to highlight additional messages without taking away from the scatterplot. It is worth mentioning here that scatterplots often develop into line geometries ( Figure 1 D), and while this can be a good thing, presenting raw data and inferential statistical models are two different messages that need to be distinguished (see Data and Models Are Different Things ).

Finally, it is almost always recommended to show the data. 7 Even if a geometry might be the focus of the figure, data can usually be added and displayed in a way that does not detract from the geometry but instead provides the context for the geometry (e.g., Figures 1 D and 1E). The data are often at the core of the message, yet in figures the data are often ignored on account of their simplicity.

Principle #4 Colors Always Mean Something

The use of color in visualization can be incredibly powerful, and there is rarely a reason not to use color. Even if authors do not wish to pay for color figures in print, most journals still permit free color figures in digital formats. In a large study 20 of what makes visualizations memorable, colorful visualizations were reported as having a higher memorability score, and that seven or more colors are best. Although some of the visuals in this study were photographs, other studies 21 also document the effectiveness of colors.

In today's digital environment, color is cheap. This is overwhelmingly a good thing, but also comes with the risk of colors being applied without intention. Black-and-white visuals were more accepted decades ago when hard copies of papers were more common and color printing represented a large cost. Now, however, the vast majority of readers view scientific papers on an electronic screen where color is free. For those who still print documents, color printing can be done relatively cheaply in comparison with some years ago.

Color represents information, whether in a direct and obvious way, or in an indirect and subtle way. A direct example of using color may be in maps where water is blue and land is green or brown. However, the vast majority of (non-mapping) visualizations use color in one of three schemes: sequential , diverging , or qualitative . Sequential color schemes are those that range from light to dark typically in one or two (related) hues and are often applied to convey increasing values for increasing darkness ( Figures 1 B and 1F). Diverging color schemes are those that have two sequential schemes that represent two extremes, often with a white or neutral color in the middle ( Figure 1 E). A classic example of a diverging color scheme is the red to blue hues applied to jurisdictions in order to show voting preference in a two-party political system. Finally, qualitative color schemes are found when the intensity of the color is not of primary importance, but rather the objective is to use different and otherwise unrelated colors to convey qualitative group differences ( Figures 1 A and 1G).

While it is recommended to use color and capture the power that colors convey, there exist some technical recommendations. First, it is always recommended to design color figures that work effectively in both color and black-and-white formats ( Figures 1 B and 1F). In other words, whenever possible, use color that can be converted to an effective grayscale such that no information is lost in the conversion. Along with this approach, colors can be combined with symbols, line types, and other design elements to share the same information that the color was sharing. It is also good practice to use color schemes that are effective for colorblind readers ( Figures 1 A and 1E). Excellent resources, such as ColorBrewer, 22 exist to help in selecting color schemes based on colorblind criteria. Finally, color transparency is another powerful tool, much like a volume knob for color ( Figures 1 D and 1E). Not all colors have to be used at full value, and when not part of a sequential or diverging color scheme—and especially when a figure has more than one colored geometry—it can be very effective to increase the transparency such that the information of the color is retained but it is not visually overwhelming or outcompeting other design elements. Color will often be the first visual information a reader gets, and with this knowledge color should be strategically used to amplify your visual message.

Principle #5 Include Uncertainty

Not only is uncertainty an inherent part of understanding most systems, failure to include uncertainty in a visual can be misleading. There exist two primary challenges with including uncertainty in visuals: failure to include uncertainty and misrepresentation (or misinterpretation) of uncertainty.

Uncertainty is often not included in figures and, therefore, part of the statistical message is left out—possibly calling into question other parts of the statistical message, such as inference on the mean. Including uncertainty is typically easy in most software programs, and can take the form of common geometries such as error bars and shaded intervals (polygons), among other features. 15 Another way to approach visualizing uncertainty is whether it is included implicitly into the existing geometries, such as in a box plot ( Figure 1 E) or distribution ( Figures 1 B and 1G), or whether it is included explicitly as an additional geometry, such as an error bar or shaded region ( Figure 1 D).

Representing uncertainty is often a challenge. 23 Standard deviation, standard error, confidence intervals, and credible intervals are all common metrics of uncertainty, but each represents a different measure. Expressing uncertainty requires that readers be familiar with metrics of uncertainty and their interpretation; however, it is also the responsibility of the figure author to adopt the most appropriate measure of uncertainty. For instance, standard deviation is based on the spread of the data and therefore shares information about the entire population, including the range in which we might expect new values. On the other hand, standard error is a measure of the uncertainty in the mean (or some other estimate) and is strongly influenced by sample size—namely, standard error decreases with increasing sample size. Confidence intervals are primarily for displaying the reliability of a measurement. Credible intervals, almost exclusively associated with Bayesian methods, are typically built off distributions and have probabilistic interpretations.

Expressing uncertainty is important, but it is also important to interpret the correct message. Krzywinski and Altman 23 directly address a common misconception: “a gap between (error) bars does not ensure significance, nor does overlap rule it out—it depends on the type of bar.” This is a good reminder to be very clear not only in stating what type of uncertainty you are sharing, but what the interpretation is. Others 16 even go so far as to recommend that standard error not be used because it does not provide clear information about standard errors of differences among means. One recommendation to go along with expressing uncertainty is, if possible, to show the data (see Use an Effective Geometry and Show Data ). Particularly when the sample size is low, showing a reader where the data occur can help avoid misinterpretations of uncertainty.

Principle #6 Panel, when Possible (Small Multiples)

A particularly effective visual approach is to repeat a figure to highlight differences. This approach is often called small multiples , 7 and the technique may be referred to as paneling or faceting ( Figure 1 G). The strategy behind small multiples is that because many of the design elements are the same—for example, the axes, axes scales, and geometry are often the same—the differences in the data are easier to show. In other words, each panel represents a change in one variable, which is commonly a time step, a group, or some other factor. The objective of small multiples is to make the data inevitably comparable, 7 and effective small multiples always accomplish these comparisons.

Principle #7 Data and Models Are Different Things

Plotted information typically takes the form of raw data (e.g., scatterplot), summarized data (e.g., box plot), or an inferential statistic (e.g., fitted regression line; Figure 1 D). Raw data and summarized data are often relatively straightforward; however, a plotted model may require more explanation for a reader to be able to fully reproduce the work. Certainly any model in a study should be reported in a complete way that ensures reproducibility. However, any visual of a model should be explained in the figure caption or referenced elsewhere in the document so that a reader can find the complete details on what the model visual is representing. Although it happens, it is not acceptable practice to show a fitted model or other model results in a figure if the reader cannot backtrack the model details. Simply because a model geometry can be added to a figure does not mean that it should be.

Principle #8 Simple Visuals, Detailed Captions

As important as it is to use high data-ink ratios, it is equally important to have detailed captions that fully explain everything in the figure. A study of figures in the Journal of American Medicine 8 found that more than one-third of graphs were not self-explanatory. Captions should be standalone, which means that if the figure and caption were looked at independent from the rest of the study, the major point(s) could still be understood. Obviously not all figures can be completely standalone, as some statistical models and other procedures require more than a caption as explanation. However, the principle remains that captions should do all they can to explain the visualization and representations used. Captions should explain any geometries used; for instance, even in a simple scatterplot it should be stated that the black dots represent the data ( Figures 1 C–1E). Box plots also require descriptions of their geometry—it might be assumed what the features of a box plot are, yet not all box plot symbols are universal.

Principle #9 Consider an Infographic

It is unclear where a figure ends and an infographic begins; however, it is fair to say that figures tend to be focused on representing data and models, whereas infographics typically incorporate text, images, and other diagrammatic elements. Although it is not recommended to convert all figures to infographics, infographics were found 20 to have the highest memorability score and that diagrams outperformed points, bars, lines, and tables in terms of memorability. Scientists might improve their overall information transfer if they consider an infographic where blending different pieces of information could be effective. Also, an infographic of a study might be more effective outside of a peer-reviewed publication and in an oral or poster presentation where a visual needs to include more elements of the study but with less technical information.

Even if infographics are not adopted in most cases, technical visuals often still benefit from some text or other annotations. 16 Tufte's works 7 , 24 provide great examples of bringing together textual, visual, and quantitative information into effective visualizations. However, as figures move in the direction of infographics, it remains important to keep chart junk and other non-essential visual elements out of the design.

Principle #10 Get an Opinion

Although there may be principles and theories about effective data visualization, the reality is that the most effective visuals are the ones with which readers connect. Therefore, figure authors are encouraged to seek external reviews of their figures. So often when writing a study, the figures are quickly made, and even if thoughtfully made they are not subject to objective, outside review. Having one or more colleagues or people external to the study review figures will often provide useful feedback on what readers perceive, and therefore what is effective or ineffective in a visual. It is also recommended to have outside colleagues review only the figures. Not only might this please your colleague reviewers (because figure reviews require substantially less time than full document reviews), but it also allows them to provide feedback purely on the figures as they will not have the document text to fill in any uncertainties left by the visuals.

What About Tables?

Although often not included as data visualization, tables can be a powerful and effective way to show data. Like other visuals, tables are a type of hybrid visual—they typically only include alphanumeric information and no geometries (or other visual elements), so they are not classically a visual. However, tables are also not text in the same way a paragraph or description is text. Rather, tables are often summarized values or information, and are effective if the goal is to reference exact numbers. However, the interest in numerical results in the form of a study typically lies in comparisons and not absolute numbers. Gelman et al. 25 suggested that well-designed graphs were superior to tables. Similarly, Spence and Lewandowsky 26 compared pie charts, bar graphs, and tables and found a clear advantage for graphical displays over tabulations. Because tables are best suited for looking up specific information while graphs are better for perceiving trends and making comparisons and predictions, it is recommended that visuals are used before tables. Despite the reluctance to recommend tables, tables may benefit from digital formats. In other words, while tables may be less effective than figures in many cases, this does not mean tables are ineffective or do not share specific information that cannot always be displayed in a visual. Therefore, it is recommended to consider creating tables as supplementary or appendix information that does not go into the main document (alongside the figures), but which is still very easily accessed electronically for those interested in numerical specifics.

Conclusions

While many of the elements of peer-reviewed literature have remained constant over time, some elements are changing. For example, most articles now have more authors than in previous decades, and a much larger menu of journals creates a diversity of article lengths and other requirements. Despite these changes, the demand for visual representations of data and results remains high, as exemplified by graphical abstracts, overview figures, and infographics. Similarly, we now operate with more software than ever before, creating many choices and opportunities to customize scientific visualizations. However, as the demand for, and software to create, visualizations have both increased, there is not always adequate training among scientists and authors in terms of optimizing the visual for the message.

Figures are not just a scientific side dish but can be a critical point along the scientific process—a point at which the figure maker demonstrates their knowledge and communication of the data and results, and often one of the first stopping points for new readers of the information. The reality for the vast majority of figures is that you need to make your point in a few seconds. The longer someone looks at a figure and doesn't understand the message, the more likely they are to gain nothing from the figure and possibly even lose some understanding of your larger work. Following a set of guidelines and recommendations—summarized here and building on others—can help to build robust visuals that avoid many common pitfalls of ineffective figures ( Figure 2 ).

Overview of the Principles Presented in This Article

The two principles in yellow (bottom) are those that occur first, during the figure design phase. The six principles in green (middle) are generally considerations and decisions while making a figure. The two principles in blue (top) are final steps often considered after a figure has been drafted. While the general flow of the principles follows from bottom to top, there is no specific or required order, and the development of individual figures may require more or less consideration of different principles in a unique order.

All scientists seek to share their message as effectively as possible, and a better understanding of figure design and representation is undoubtedly a step toward better information dissemination and fewer errors in interpretation. Right now, much of the responsibility for effective figures lies with the authors, and learning best practices from literature, workshops, and other resources should be undertaken. Along with authors, journals play a gatekeeper role in figure quality. Journal editorial teams are in a position to adopt recommendations for more effective figures (and reject ineffective figures) and then translate those recommendations into submission requirements. However, due to the qualitative nature of design elements, it is difficult to imagine strict visual guidelines being enforced across scientific sectors. In the absence of such guidelines and with seemingly endless design choices available to figure authors, it remains important that a set of aesthetic criteria emerge to guide the efficient conveyance of visual information.

Acknowledgments

Thanks go to the numerous students with whom I have had fun, creative, and productive conversations about displaying information. Danielle DiIullo was extremely helpful in technical advice on software. Finally, Ron McKernan provided guidance on several principles.

Author Contributions

S.R.M. conceived the review topic, conducted the review, developed the principles, and wrote the manuscript.

Steve Midway is an assistant professor in the Department of Oceanography and Coastal Sciences at Louisiana State University. His work broadly lies in fisheries ecology and how sound science can be applied to management and conservation issues. He teaches a number of quantitative courses in ecology, all of which include data visualization.

All resources

• Blog Articles
• Use Cases & Examples
• Guides & Resources
• Dashboard Templates
• Case Studies
• Testimonials

What is data visualization

Why use data visualization, the advantages and benefits of good data visualization, principles of successful data visualization, how to choose a chart type, how to tell the visual story, designing the dashboard, why data visualization is important, reporting and visualization software tools comparison, looker studio by google, google sheets, key takeaways, what is data visualization: definition, principles, examples, tools.

Vlada Malysheva ,  Creative Writer @ OWOX

Top 30 Handpicked Google Looker Studio Dashboards for Marketers

Share with friends

65% of people are visual learners , making data visualization an effective way to communicate information.

When Excel spreadsheets aren’t enough to connect the dots between your data and there’s no possibility to involve data or digital analyst to get the report quickly, data visualization software tools and tools is what you need to become data-savvy.

In this article, which was last updated in January 2024, we’ll show you what data visualization techniques are available, how to visualize data correctly, which tools can be used for engaging and interactive visualizations without any help from developers or data professionals, and how to choose a tool that suits your specific needs.

The definition of  data visualization is the visual representation of your data. With the help of charts, maps, and other graphical elements these tools provide a simple and comprehensible way to clearly see and easily discover insights and patterns in your data.

Data visualization is the graphical representation of data using visual elements such as charts, graphs, and maps.

It is a way to communicate complex information in a visual and intuitive manner, making it easier for people to understand and analyze the data. By transforming raw data into visual representations, data visualization allows patterns, trends, and insights to be easily identified and interpreted.

Data visualization is also a powerful storytelling tool. Visual storytelling helps to uncover hidden patterns, relationships, and correlations that may not be apparent, or not visible in raw data. Through visualizations, data can be presented in a way that is engaging, impactful, and memorable, enabling effective communication and data-driven decision-making .

Data visualization is not limited to a specific field or industry. It's not only about marketing data and is used in various domains such as business, finance, healthcare, education, or journalism. In business, data visualization is used to analyze sales trends, key performance indicators, and present business metrics. In healthcare, it is used to visualize patient data, monitor disease outbreaks, and analyze medical research. In journalism, it is used to create better stories and increase reach and consumption.

If you want your Facebook post to be read by as many people as possible, what will you do? You’ll add an interesting visual. This trick works perfectly with reports too. Data-driven visuals attract more attention, are easier to understand, and assist in getting your message across to the audience quickly.

With the help of descriptive graphics and dashboards, even difficult information can be clear and comprehensible.

Why is that?

Most people are visual learners. So if you want the majority of your partners, colleagues, and clients to be able to interact with your data, you should turn boring charts into beautiful graphics. Here are some noteworthy numbers, based on research, that confirm the importance of visualization:

• People get 90% of information about their environment from the eyes.
• 50% of brain neurons take part in visual data processing.
• Pictures increase the wish to read a text up to 80%.
• People remember 10% of what they hear, 20% of what they read, and 80% of what they see.
• If a package insert doesn’t contain any data illustrations, people will remember 70% of the information. With pictures added, they’ll remember up to 95%.

With  OWOX BI , your data is collected, normalized, attributed & prepared for reporting. 

Use our templates to get reports built in minutes, or use your data to prepare the data for any report you need and visualize it in Looker Studio (formerly Google Data Studio), Google Sheets or the BI tool of your choice. Save 70+ hours on data preparation every month and automate your entire digital marketing reporting.

Relevant visualization brings lots of advantages for your business:

• Fast decision-making.  Summing up data is easy and fast with graphics, which let you quickly see that a column or touchpoint is higher than others without looking through several pages of statistics in Google Sheets or Excel... or even a database or a CRM or CMS system.
• More stakeholders are involved.  Most people are better at perceiving and remembering information presented visually and delivered on time in a visual-appealing format. 
• Higher level of involvement.  Beautiful and bright graphics with clear messages attract readers’ attention.
• Better understanding.  Perfect reports are transparent not only for technical specialists, analysts, and data scientists but also for CMOs, CEOs and other C-levels or managers, and help each and every worker make decisions in their area of responsibility.

The first thing to do before creating any chat is to check all information for accuracy and consistency.

For example, if the scaling factor is 800%, whereas the average is 120–130%, you should check where this number is coming from. Maybe it’s some kind of an outlier that you need to delete from the graph so it doesn’t skew the overall picture: 800% downplays the difference between 120% and 130%. This kind of outlying data in a report can lead to incorrect decisions made.

To increase the chances of success in marketing, the right message should be delivered to the right person at the right time. 

The same three rules are applied for data visualization:

• Choose the right chart  to visualize the answer to specific question based on your goal.​
• Confirm that the message to deliver the result of your report suits your audience (the stakeholder).
• Use an appropriate design for the chart to deliver that message.

If your message is timely but the chat or graphic isn’t dynamic, or it provides incorrect insights. or the design is not attractive, then you won’t achieve the results you were dreaming of.

If you choose the wrong chart or graph, your readers will be confused or interpret or read the results incorrectly. That’s why before creating a report with charts, it’s important to decide what data you want to visualize and for what purpose, for example: 

• To  compare different data points
• ​To show data distribution : for instance, which data points are frequent and which are not
• To show the structure of something with the help of data
• To follow the connections, references or correlation between data points

Let’s have a look at the most popular types of charts and the goals they can help you achieve.

1. Line chart

A line chart is a type of data visualization that uses a series of data points connected by straight lines. It is commonly used to show the relationship between two variables over a continuous period of time. Foe example, the x-axis represents the time or the independent variable, while the y-axis represents the value or the dependent variable.

By plotting the data points and connecting them with lines, the line chart provides a visual representation of how the values change over time.

Pros of Line Charts

One of the main advantages of line charts is their ability to display trends and patterns in data . They make it easy to identify the overall direction of change, whether it is increasing, decreasing, or remaining stable.

Line charts also allow for the comparison of multiple data series on the same chart, making it simple to  analyze the correlation between different variables .

Additionally, line charts are visually appealing and easy to understand, making them accessible to a wide range of audiences.

Cons of Line Charts

However, line charts also have some limitations. They are most effective when used with continuous data, such as time series data, and may not be suitable for categorical or discrete data .

Line charts can become cluttered and confusing if there are too many data points or series plotted on the chart. They may also not be the best choice for displaying data with irregular or inconsistent intervals. It is important to consider these factors when deciding whether to use a line chart for data visualization.

Use cases of Line Charts

The best use cases for line charts include analyzing sales or revenue data over time , tracking website traffic or user engagement metrics , visualizing stock market trends, or monitoring changes in weather patterns.

Line charts are particularly useful when there is a need to understand the overall trend or pattern in the data and identify any significant changes or anomalies. They are also effective for presenting data to a non-technical stekholders, as they provide a clear and really easy and intuitive representation of the data.

2. Bar chart

Type of diagram that represents data using rectangular bars is called bar chart. Each bar corresponds to a specific metric or variable, while its length or height represents the value associated with that metric.

Bar charts are typically used to compare different metrics or track changes over time  providing simplicity and versatility.

Horizontal bar charts are often used when you need to compare lots of data sets or to visually emphasize the distinct advantage of one of the data sets.

Vertical bar charts display how data points change over time — for example, how the annual company profit has changed over the past few years.

Pros of Bar Charts

Bar charts are ease of read and consume, no background in data analysis is required. 

The clear and straightforward presentation of data in bar charts allows for quick insights and understanding.

Additionally, bar charts can accommodate large datasets and display multiple variables simultaneously (and stay usable).

Cons of Bar Charts

Continuous data, such as temperature measurements over time, may not be as suitable for bar charts.

Bar charts may also not be the best choice for displaying complex relationships or correlations between variables, as they primarily focus on comparing values within categories.

Use cases of Bar Charts

Some common use cases include sales analysis, market research, financial reporting, and survey results. For example, a bar chart can be used to compare the market share of different companies in a specific industry , or to visualize the responses to a survey question with multiple answer options.

A bar chart can also represent the sales figures of different products in a given month , with each bar representing a product or a category, and its height indicating the sales quantity. This visual representation allows for easy comparison and identification of trends or patterns in the data.

The  of bar charts make them a valuable tool for data visualization in various domains.

3. Histogram

A histogram is often mistaken for a bar chart due to their visual similarities, but the goals of these charts are different.

 A histogram shows the distribution of a dataset across a continuous interval or a definite time period. It is a graphical representation of the frequency of data values in different intervals or bins. The x-axis of a histogram represents the range of values in the dataset, divided into equal intervals, while the y-axis represents the frequency or count of data values falling within each interval. The height of each bar in the histogram corresponds to the frequency of data values in that interval. This chart provides a visual summary of the underlying data distribution.

Unlike a histogram, a bar chart doesn’t show any continuous interval; each column displays a category of its own. It’s easier to demonstrate the number of purchases in different years with the help of a bar chart. 

If you want to know the number of order beween $10 and 100, $101 and 200, $201 and 300, etc. of purchases, it’s better to choose a histogram. The histogram will show you the frequency of orders falling within each price range, allowing us to identify patterns such as a normal distribution, skewed distribution, or outliers.

Histogram allows you to quickly identify the central tendency, spread, and shape of the dataset. Histograms are particularly useful when dealing with large datasets or continuous data, as they provide a visual summary without overwhelming the viewer with individual data points.

What are the limitation of the histogram?

First, the choice of bin size or interval width can impact the interpretation of the data. A smaller bin size can provide more detailed information but may also result in a cluttered or noisy chart . At the same time, a larger bin size can oversimplify the data distribution .

Second, histograms may not be suitable for datasets with categorical or ordinal variables , as they require numerical data to create meaningful intervals.

4. Pie chart

A pie chart is a type of data visualization that displays shares of each value in a data set.

It is divided into slices, where each slice represents a proportion or percentage of the whole. The size of each slice is determined by the value it represents in relation to the total value of the data set. Pie charts are commonly used to show the distribution or composition of a categorical variable.

Pie chart visually displays the relative proportions of different categories within a data set. It allows viewers to quickly grasp the overall distribution of the data and easily compare the sizes of different categories.

The angles of the slices in the pie chart represent the proportions of the categories, making it easy to understand the relationship between the parts and the whole. For instance, what percentage of general sales is attributed to each product category?

Pie charts are particularly useful when dealing with data that has a small number of categories or when the emphasis is on comparing the parts to the whole. They can also be useful for highlighting a specific category or identifying outliers.

The biggest pie chart limitation is that they can become difficult to interpret when there are too many categories or when the differences between the categories are small. It can be challenging to accurately compare the sizes of the slices, especially if they are similar in magnitude.

Additionally, pie charts do not easily don't represent the trends over time. Pie charts are commonly used in business and marketing to represent market share, customer demographics, or product sales by category. Pie charts are also used in survey data to display the distribution of responses for multiple-choice questions. Overall, pie charts are most effective when the data is simple, the categories are distinct, and the emphasis is on comparing the parts to the whole.

5. Scatter plot

A scatter plot chart displays the relationship between two numerical variables . It uses a Cartesian coordinate system, where each data point is represented by a dot or marker on the chart.

The x-axis represents one variable, while the y-axis represents the other variable. By plotting the data points on the chart, you can visually analyze the correlation or pattern between the variables.

The scatter plot chart allows you to identify trends, clusters, or outliers in the data.

Additionally, scatter plots can be used to detect any patterns or irregularities in the data distribution. 

The main limitation is that it can only represent two variables at a time. If there are more than two variables to analyze, additional charts are required. Also, scatter plots may not be suitable for large datasets, as the overlapping data points can make it difficult to make decisions based on the chart accurately.

For example, with the help of a scatter plot, you can find out how the conversion rate changes depending on the size of the product discount.

6. Bubble chart

This is an interesting chart that allows you to compare two parameters by means of a third.

It is a variation of a scatter plot, where the size of the bubbles is used to convey additional information. The bubble chart is particularly useful when visualizing three variables, as it allows for the representation of two continuous variables on the x and y axes, while the size of the bubbles represents the third variable. This makes it easy to identify patterns and relationships between the variables in a single chart.

Bubble chat allows you to display large amounts of data in a visually appealing and intuitive way. By using different colors or shades, you can also incorporate a fourth variable into the chart, further enhancing the information conveyed. The size of the bubbles provides a quick visual cue, allowing for easy comparisons between data points.

Additionally, the bubble chart can be interactive, allowing users to hover over or click on the bubbles to reveal more detailed information.

Basically, the main drawback of the bubble charts is that the size of the bubbles can sometimes be misleading, as it may not accurately represent the magnitude of the data point. This can be mitigated by scaling the size of the bubbles appropriately or by providing a clear legend or scale.

Also, it is important to strike a balance between the number of data points and the readability of the chart.

The best use cases for bubble charts are situations where you want to visualize relationships between three variables.

For example, you can use a bubble chart to show the relationship between the price, size, and number of orders of different products. It can also be used to compare data across different categories or groups, such as comparing the revenue, market share, and growth rate of different companies in an industry.

Bubble charts are particularly effective when the size of the bubbles is meaningful and provides valuable insights into the data.

7. Geo chart

The geo chart is a simple one. It’s used when you need to demonstrate a certain data distribution across regions, countries, and continents.

By visualizing data on a map, a geo chart provides a clear and intuitive way to understand spatial patterns and user behavior. For example, a geo chart can show shopping frequency across countries, GDP per capita by country, or election results by region. It allows viewers to quickly grasp the variations and disparities between different locations. Basically, geo chart works best if metric dimension is geographical.

By mapping data onto a familiar geographic locations, it becomes easier for viewers to interpret and remember the information.

Since a geo chart relies on colors or patterns to represent data, it is important to choose appropriate color schemes and legends to avoid confusion or bias. Furthermore, a geo chart may not be suitable for displaying complex or detailed data, as the level of granularity is often limited to the size and boundaries of the regions on the map. It is important to carefully select the level of detail and aggregation that best suits the purpose of the report.

For example, when analyzing sales data, a geo chart can show the distribution of sales across different regions , helping businesses identify potential markets or areas of improvement.

Overall, geo charts are particularly effective when the spatial dimension of the data is crucial for decision-making or storytelling.

The second important thing that you have to take into account while working with visualization is  choosing the right message for the audience . The information you talk about, the story you tell in the report should be clear and informative for your readers.

Here’s a chart that was awarded the prestigious Data Journalism Award. 

For people who aren’t familiar with the background to the story, this chart looks like a picture made by a three-year-old. However, when you find out a little bit more about it, you can see the huge amount of work done by its authors.

Charles Seife and Peter Aldhous, Buzzfeed News editors, used the R language to  visualize flight data obtained by FBI and DHS agents as part of air surveillance. Specifically, this chart shows flights above the house and mosque of those responsible for the mass shooting in December 2015 in San Bernardino, California.

While choosing the parameters you want to visualize on one chart, you have to confirm that they can be combined. Some combinations just aren’t logical, though at first sight the information correlates perfectly. Here’s an example of such a chart with a faulty correlation. It shows that the number of people who drowned by falling into a pool correlates with the number of Nicolas Cage films.

The next things you should take into account when creating a chart are the scale and scope. People are used to the fact that measurements on axes start from the bottom and from the left. If you change the direction of measurement, it will confuse an inattentive audience. Although we should mention that reversing the measurement is possible when used as a tactical maneuver, as in this example:

At first sight, it may seem that the number of murders committed using firearms has been decreasing over the years. In fact, it’s the opposite, as the scale starts from the top. Perhaps the author of the chart did this on purpose to decrease the negative response to the results shown.

A suitable scale also makes your chart clearer. If a report shows data points that are too close and you can’t see any movement, try to change the scale. Start the measurements not from zero or divide the scale into smaller parts and the picture will clear up.

Before giving a report to the stakeholder, make sure that the chart loads fast. Slow loading kills all your efforts.

For example, if you’re visualizing data in Google Sheets, most likely your data is stored on the same page or on the next page and doesn’t come from a third-party source.

But when you create a report in Looker Studio (ex. Data Studio) or Power BI, data will be imported from somewhere else. In this case, you have to pay close attention to the source accessibility and the data flow rate. Otherwise, you’ll see a sad looking picture when there’s a chart template but data hasn’t been loaded.

Remember, the golden rule when you're crafting your chart design is to keep it simple. 

When you're tasked with putting together a standard report, don’t fret about making it look fancy. You don't need to dress it up.

Avoid any extra elements that only clutter the chart: too many colors and structures, 3D volume, shadows, gradients, etc.

The simpler a chart is - the easier it is for the readers to understand the information you want to share.

Don’t make your visualizations too small, and don’t put all charts on the same dashboard page. It’s considered bad style to use more than three types of charts on one slide or the same dashboard page. If you really need so many chart types, put them on different pages, or make a clear separation, so it’s easy to understand them.

Don’t be afraid to experiment. If you have a task that's not typical, perhaps your solution should also be non-standard. In the infographic below, we can see the wing movement patterns of different animals. The dynamic visualization is totally relevant.

Let’s have a look at some data visualization tools examples and discuss how to choose the right one for your goals.

Visualizing data is an undeniable benefit in any niche, and it doesn’t matter if you’re building a career in marketing, design, retail or anything else.

Making information easy to consume and quickly make smart decisions is one of the keys to finding growth zones and developing your business.

When your colleagues would see the visual charts outlining the current state of the main metrics for you, it’s easier to make sure that all of the team members are  on the same page and everyone understands the strong and weak points of the current strategy.

While visualizing reports itself cannot fix the issues, it gives you the wheel to drive the car, to make the necessary changes and improve the KPIs.

Nowadays, there are lots of data visualization and reporting tools on the market. Some of them are paid, others are available for free. Some of them work fully on the web, others can be installed on a desktop but work online, and others are offline only. 

Best Reporting Tools

We’ve crafted a list of 10 most popular reporting and data visualization software:

1. Google Spreadsheets

Explore BigQuery Data in Google Sheets

Bridge the gap between corporate BigQuery data and business decisions. Simplify reporting in Google Sheets without manual data blending and relying on digital analyst resources availability

2.  Looker Studio (ex. Google Data Studio)

3.  Tableau

4.  Power BI

6. QlikView

7. R Studio

8. Visual.ly

First six tools and services are created by companies specializing in visualization. 

Numbers seven through ten are quite interesting tools, mostly free and online. They offer non-standard types of data visualization and may offer new ways of approaching your business information.

How to select a reporting tool

What to look for when choosing a reporting tool:

• Start from the goals and tasks you want to accomplish.  For example, a major trend on the market nowadays is dynamic reports. If a tool cannot work with dynamic reports, that’s a strike against it.
• Consider the amount of money you’re ready to pay.  If your team is big enough and every employee has to work with the visualization tool, then the cost per user may be a stop sign.
• Decide who will use the tool and how:
• Is there a possibility for group editing? 
• How simple is it to start working with the tool? 
• Is the interface user-friendly? 
• Is there a possibility to create a report without any knowledge of programming? 

For example, R Studio is a great service, especially for searching for trends and building attribution and correlation models. But if you are not familiar with coding, you won't be able to connect any specific libraries, and it would be difficult for you to start working with R Studio.

We'll dive deeper into a few services and guide you through their pros & cons, as well as the main features and advantages. But before we start, let us explain how  dynamic data visualization  and  dynamic reports  differs.

Dynamic reports   refer to the possibility to import data from different sources in real time. 

For example, Looker Studio (formerly Google Data Studio) doesn’t have dynamic reports in place. Let’s say we’ve connected a Looker Studio request from Google BigQuery and then changed something in this request. To record these changes in the report, we need to at least refresh the page. 

However, if we add or delete some fields in Google BigQuery (not just change the logic of the calculation but change the table structure), then Looker Studio would show an error. You’ll have to rebuild the dashboard to get the visualizations in place.

Dynamic visualization   concept refers to the possibility to look at summary statistics over different dates during one session. 

For example, in Google Analytics 4 you can change the time period and get statistics for the date range you need.

OWOX BI is a comprehensive analytics platform that covers everything from data collection and streaming to attribution modeling and reporting. With OWOX BI, companies get a complete view of their marketing activities across various channels, empowering advertising specialists to optimize their ad spending and achieve better ROI.

3 whales of data management and analysis

OWOX BI Pipelines   facilitates seamless  data collection  from various advertising platforms, CRMs, and website builders, enabling organizations to consolidate all their data in one place in order to have a  data source of truth  and gain better insights.

OWOX BI Streaming  is a  cookieless real-time user behavior tracking  system, ensuring privacy compliance with regulation and extending the lifespan of cookies. Marketers can accurately track the entire conversion journey, find the  true sources of conversions , and gain a deeper understanding of customer behavior.

OWOX BI Transformation   saves time on data preparation  (avg. of 70 hours per month). With pre-built  low- or no-code transformation templates  (based on 100’s delivered projects across multiple industries), businesses can quickly produce  trusted datasets for reporting , modeling, and operational workflows:

• Sessionization: Group on-site events into sessions to  find conversion sources
• Cost data blending: Merge ad cost data across channels to  compare campaign KPIs in a single report
• Attribute ad costs to sessions to  measure  cohorts   and pages' ROI;
• Create cross-device user profiles across  different devices
• Identify new and returning user types for  accurate analysis
• Apply a set of  attribution models : Choose from standard attribution models like First-Click, LNDC, Linear, U-shape, and Time Decay, or create a custom  Machine Learning Funnel-based attribution model
• Use modeled conversion for cookieless measurements and  conversion predictions
• Prepare data for  marketing reports  in minutes

Lastly, OWOX BI integrates with visualization tools like Looker Studio, Tableau, or Power BI, enhancing data-driven decision-making by building customizable reports & keeping the data always up-to-date.

OWOX BI Advantages & Benefits

• No technical background, coding experience or knowledge of SQL is required.
• Simple and user-friendly interface: you can collect all of the data and generate reports using our dashboard templates and customize what matters for you the most.
• If you want to working with your data in Google Sheets, you can easily export an aggregated dataset from BigQuery to Google Sheets with our reports add-on . 
• You can copy SQL queries generated by OWOX BI. 
• You can then modify those queries or use them, for example, to automate a data-based report in Google Sheets or BigQuery.
• You retain complete control over access to that data .
• You can merge digital marketing data with CRM/CMS data.
• Full transperancy.

Note: For enterprise customers, OWOX BI expert team will set up a data model tailored to your business. You’ll be able to evaluate the impact of all marketing efforts — both online and offline.

Types of data you can use

User actions on your site:

• You can set up the collection of raw data from the site in Google BigQuery using OWOX BI Streaming.
• Or you can use the native standard export from Google Analytics 360 or Google Analytics 4 to Google BigQuery.

Transactions Data:

• Google Analytics → Google BigQuery
• Google Sheets → Google BigQuery
• CRM → Google BigQuery

Advertising campaign costs:

• Advertising services → Google Analytics 4
• Advertising services → Google BigQuery
• Other marketing tools → Google BigQuery

Looker Studio, also  known as data studio  allows you to connect data sources, easily build charts, reports and add elements to visualize and share reports with colleagues in a way that’s similar to other Google products.

Advantages:

• Free​ (with paid version announced in 2023)
• More than 860 connectors to the data sources that are easy to integrate
• Allows to use data from several sources via one dashboard
• Convenient to share reports

Looker Studio is a free tool with  21 native connectors  provided by Google:

• Connect to your Looker semantic models. 
• Connect to Google Analytics 4 reporting views. 
• Connect to Google Ads performance report data.
• Connect Google Sheets .
• Connect to BigQuery tables and custom queries.
• Connect to AppSheet app data.
• File Upload - Use CSV ( comma-separated values ) files. 
• Connect to Amazon Redshift .
• Connect to Campaign Manager 360 data.
• Connect to MySQL databases.
• Connect to Display & Video 360 report data.
• Connect to Microsoft SQL Server databases.
• Connect to PostgreSQL databases.
• Connect to Search Console data.
• Connect to YouTube Analytics data. 

and more...

They’re checked, approbated, work well, and perfectly suit to the most common reporting tasks. 

There are also connectors provided by Google partners, though you have to understand that connectors can be presented by developers with different skill levels and there’s no guarantee they’ll perform correctly.

By the way, if you want to see any Facebook or Yahoo Gemini statistics in reports built in Looker Studio, you can  import ad cost data into Google BigQuery  with OWOX BI. While you may lose some of the important data with other data connectors, with our Facebook Ads to Google BigQuery pipeline you receive complete data ready for analysis and reporting from your Facebook account.

You can also merge your Facebook Ads data with the advertising cost data from Google Ads, Twitter Ads, and LinkedIn ads and get a helicopter view of your marketing performance and optimize your cross-channel budget easily.

Automate your digital marketing reporting

Manage and analyze all your data in one place! Access fresh & reliable data with OWOX BI — an all-in-one reporting and analytics tool

We also have a ready-to-use dashboard templates of our own that we want to share.

We've prepared a comprehensive Looker Studio dashboard template gallery with  ready-to-use templates  so that you can quickly create a guide to your business results, KPIs and performance.

The first is a  All-in-one Performance Dashboard . With this dashboard, you can find all of the basic metrics and metrics to stay on top of your advertising and marketing performance and achieve the desired ROI.

All-in-one Digital marketing Dashboard

Another dashboard template we'd like to share is the  Digital Marketing Paid Channels KPI dashboard , which is segmented by data sources (shown in detail). In other words, it shows filtered data on Facebook marketing campaigns, etc.

Those are the dashboard templates. Make a copy, change the data sources to your own, and use them to build beautiful reports based on your data. 

One of the recent Looker Studio updates adds the possibility to filter information by view. For example, you can compare data points over the current period and the previous year.

One more interesting update allows you to change the type of an already created chart, graph or element. Earlier, when changing a chart, you had to delete it and create a new one. 

Useful links:

• Webinar: Mastering Marketing KPIs
• Google Looker Studio dashboard template gallery

Mastering Marketing KPIs: How to Evaluate Your Marketing Performance

You'll learn how you can effectively evaluate your marketing performance to fuel your business growth.

Marketing Ninja @ OWOX

This is one of the two most popular data reporting tools (together with Microsoft Excel) that’s used by any marketing specialist at least once. The Google Sheets interface is quite simple and easy-to-use, especially for those who just starting analytics out.

• Flexible — supports dynamic parameters, vlookups, pivot tables, formulas, app scripts etc.
• Easy to integrate with data sources (but not so easy to automate updates)
• Convenient to share reports via links

The charts and reports types in Google Sheets are the same as that is in Looker Studio.

Conditional Formatting

Conditional formatting in Google Sheets allows users to apply formatting rules to cells based on specific conditions .

These conditions can be based on the cell's value, text, or even a formula . By using conditional formatting, users can visually highlight important data, identify trends , and make their spreadsheets more visually appealing and easier to understand.

For example, let's say you have a sales report in Google Sheets and you want to highlight all the cells that have sales numbers above a certain threshold . With conditional formatting, you can set a rule that applies a different background color to those cells automatically. This makes it easier to quickly identify the high-performing sales figures without manually scanning through the entire spreadsheet.

Pivot Tables

Perhaps the main advantage of Google Sheets as your go-to reporting tool is pivot tables.

Pivots allow users to summarize and analyze significant amounts of data. They are used to transform raw flat table data or data sets into insights by organizing and summarizing it in an easy and structured relatively small table.

With Pivot tables you can quickly explore data from different perspectives, change columns and rows, sort values, identify patterns , and uncover trends or anomalies . They are particularly useful for data aggregation tasks.

For example , let's say you have a spreadsheet with sales data for a company. The data includes columns for product names, sales dates, sales quantities, and sales amounts . By creating a pivot table, you can easily summarize this data to answer questions like: ' What are the total sales amounts for each product? ' or ' What are the average sales quantities by month? '

Pivot tables allow you to group and aggregate data based on different criteria , such as product , sales date , or any other relevant attribute.

VLOOKUP formula

VLOOKUP is a function in Google Sheets that stands for vertical lookup . It is used to search for a specific value in the leftmost column of a range of cells, and then return a corresponding value from a different column in the same row . Vlookup is commonly used to find and retrieve data from large datasets or tables.

Imagine having a list of products and their corresponding prices. You can use VLOOKUP to search for a specific product name in the leftmost column, and then retrieve the price of that product from a different column in the same row.

This can be useful for tasks such as pricing analysis.

The syntax of the VLOOKUP function in Google Sheets is as follows: =VLOOKUP(search_key, range, index, is_sorted).

The search_key is the value you want to search for, the range is the range of cells where the search will be performed, the index is the column number from which the corresponding value should be returned, and the is_sorted is an optional parameter that specifies whether the range is sorted in ascending order or not.

Everything about VLOOKUP in Google Sheets

BigQuery <> Google Sheets

If you want to get data from BigQuery to visualized in Google Sheets, there is a Google sheets extension that allows you to query data directly from Google BigQuery and build reports based on the imported data.

Basically, you can request any data stored in your Google BigQuery project directly from the Sheets interface. 

Simplify BigQuery Reporting in Sheets

Easily analyze corporate data directly into Google Sheets. Query, run, and automatically update reports aligned with your business needs

Cohort Analysys

Last but not least, we wanted to share with you one of our favorite reports for Google Sheets — the cohort analysis report .

Cohort analysis is a powerful analytical technique used to understand the behavior and common details of a specific group of individuals over time. 

It involves dividing a larger audience into smaller groups, or cohorts , based on a common characteristic or event. These cohorts are then analyzed to identify patterns and trends  that can help businesses make informed decisions and improve their strategies.

The most common case of cohort analysis in marketing is to track the behavior of customers who made their first purchase in a particular month . By analyzing this cohort, businesses can determine the retention rate, average purchase value, and lifetime value of these customers. This information can be valuable in identifying the most effective marketing channels, optimizing customer acquisition strategies, and improving customer loyalty.

Additionally, you can read our  detailed guide to cohort analysis in Google Analytics 4  and Google Sheets, where we provide very detailed instructions. We’ve also hosted a webinar on cohort analysis .

Finally, we want to share some useful links and books on data visualization:

• Edward Tufte,  The Visual Display of Quantitative Information
• Stephen Few,  Big Data, Big Dupe
• «The Joy of Stats»  (documentary film)

Visualization services can help you make your reports visually appealing and comprehensive, you can highlight valuable insights in your data easily. 

If you want to keep up with the pace of modern business, adding visual storytelling and data exploration to your reports will allow you to accelerate the process of decision-making.

If you still don’t know which of all data visualization tools would fit your business needs, book a free demo to discuss your specific situation with our data experts and discover the ideal solution designed for you. 

Gain clarity for better decisions without chaos

No switching between platforms. Get the reports you need to focus on campaign optimization

How can I create effective data visualizations?

What are some popular data visualization tools, what are the benefits of data visualization.

Don't forget to share post!

You might also like

The Role of Business Dashboards: Real-life Examples for Practical Use

Google Data Studio Templates for Marketers — The Ultimate Guide

How to Create Meaningful Dashboards for Your Facebook Data

BigQuery to Looker Studio: A New Approach to Dynamic Visualization and Report Automation

30 Best Marketing Dashboard Examples and Templates [2023]

Now You See Me! What is CPA Fraud and How to Fight it

Top 5 Tools for Advertising Reporting

Dashboards for businesses: Why you need one and top services for building it

2,000 companies rely on us

• SUGGESTED TOPICS
• The Magazine
• Newsletters
• Managing Yourself
• Managing Teams
• Work-life Balance
• The Big Idea
• Data & Visuals
• Reading Lists
• Case Selections
• HBR Learning
• Topic Feeds
• Account Settings
• Email Preferences

Visualizations That Really Work

• Scott Berinato

Not long ago, the ability to create smart data visualizations (or dataviz) was a nice-to-have skill for design- and data-minded managers. But now it’s a must-have skill for all managers, because it’s often the only way to make sense of the work they do. Decision making increasingly relies on data, which arrives with such overwhelming velocity, and in such volume, that some level of abstraction is crucial. Thanks to the internet and a growing number of affordable tools, visualization is accessible for everyone—but that convenience can lead to charts that are merely adequate or even ineffective.

By answering just two questions, Berinato writes, you can set yourself up to succeed: Is the information conceptual or data-driven? and Am I declaring something or exploring something? He leads readers through a simple process of identifying which of the four types of visualization they might use to achieve their goals most effectively: idea illustration, idea generation, visual discovery, or everyday dataviz.

This article is adapted from the author’s just-published book, Good Charts: The HBR Guide to Making Smarter, More Persuasive Data Visualizations.

Know what message you’re trying to communicate before you get down in the weeds.

Idea in Brief

Knowledge workers need greater visual literacy than they used to, because so much data—and so many ideas—are now presented graphically. But few of us have been taught data-visualization skills.

Tools Are Fine…

Inexpensive tools allow anyone to perform simple tasks such as importing spreadsheet data into a bar chart. But that means it’s easy to create terrible charts. Visualization can be so much more: It’s an agile, powerful way to explore ideas and communicate information.

…But Strategy Is Key

Don’t jump straight to execution. Instead, first think about what you’re representing—ideas or data? Then consider your purpose: Do you want to inform, persuade, or explore? The answers will suggest what tools and resources you need.

Not long ago, the ability to create smart data visualizations, or dataviz, was a nice-to-have skill. For the most part, it benefited design- and data-minded managers who made a deliberate decision to invest in acquiring it. That’s changed. Now visual communication is a must-have skill for all managers, because more and more often, it’s the only way to make sense of the work they do.

• Scott Berinato is a senior editor at Harvard Business Review and the author of Good Charts Workbook: Tips Tools, and Exercises for Making Better Data Visualizations and Good Charts: The HBR Guide to Making Smarter, More Persuasive Data Visualizations .

Partner Center

• Data Science
• Data Analysis
• Data Visualization
• Machine Learning
• Deep Learning
• Computer Vision
• Artificial Intelligence
• AI ML DS Interview Series
• AI ML DS Projects series
• Data Engineering
• Web Scrapping

Visualizing Text Data: Techniques and Applications

Text data visualization refers to the graphical representation of textual information to facilitate understanding, insight, and decision-making. It transforms unstructured text data into visual formats, making it easier to discern patterns, trends, and relationships within the text. Common techniques include word clouds, bar charts, network diagrams, and heatmaps, among others.

Visualizing Text Data

This article delves into the concept of text data visualization, its importance, various techniques, tools, and when to use it.

Table of Content

Importance of Text Data Visualization

• When to Use Text Data Visualization

Techniques for Text Data Visualization

1. word clouds, 2. bar charts, 3. bigram network, 4. word frequency distribution plot:, 5. network graphs, examples and use cases for text data visualization.

The importance of text data visualization lies in its ability to simplify complex data. Key benefits include:

• Enhanced Comprehension: Visualizations make it easier to grasp large volumes of text data quickly.
• Pattern Recognition: Helps identify trends, frequent terms, and associations that might not be apparent from raw text.
• Improved Communication: Visual representations can convey insights more effectively to stakeholders who may not be familiar with textual analysis techniques.
• Data Exploration: Facilitates exploratory data analysis, allowing users to interactively explore and understand the text data.
• Facilitates Decision-Making : By providing clear insights, text data visualization aids in informed decision-making.

When to Use Text Data Visualization?

Text data visualization is particularly useful in the following scenarios:

• Exploratory Data Analysis: When you need to explore large text datasets to identify key themes and patterns.
• Summarizing Large Text Corpora: To condense and present the essence of lengthy documents or collections of text.
• Comparative Analysis: When comparing text data across different sources, time periods, or categories.
• Communication and Reporting: To present findings from text analysis to a non-technical audience.
• Detecting Anomalies or Outliers: In contexts like social media monitoring or customer feedback analysis, where identifying unusual patterns is crucial.

Visualizing text data can be done using several techniques, each of which can highlight different aspects of the data. There are several types of text data visualizations, each serving different purposes:

Word clouds are one of the most popular and straightforward text visualization techniques. Display the most frequent words in a text dataset, with the size of each word reflecting its frequency.

Use Cases :

• Summarizing large text datasets.
• Identifying key themes in customer feedback or social media posts.

Here’s a simple example of text data visualization using a word cloud.

• This code uses the wordcloud library to generate a word cloud from a sample text.
• If you don’t have the wordcloud and matplotlib libraries installed, you can install them using pip install wordcloud matplotlib.

Bar charts can be used to visualize the frequency of specific words or phrases in a text dataset. They provide a clear and precise comparison of word frequencies.

• Comparing the frequency of keywords in different documents.
• Analyzing the distribution of topics in a dataset.
• Used to show the frequency of specific terms or categories within the text.

Code Implementation:

A Bigram Network is a visualization technique used to illustrate the relationships between pairs of words (bigrams) in a text dataset. This network graphically represents the most frequent pairs of words that appear consecutively in the text, with nodes representing words and edges representing the connections between them.

• Understanding the contextual relationship between words in large text datasets.
• Analyzing patterns in customer feedback or social media posts to identify common themes or issues.
• Exploring text data from research articles, books, or any large corpus to discover hidden connections.

Bigram Network

A Word Frequency Distribution Plot is a graphical representation that shows how frequently different words appear in a text dataset. It typically displays words on the x-axis and their corresponding frequencies on the y-axis. This plot helps in understanding the distribution of words in the text, identifying the most common words, and observing the overall frequency pattern.

• Analyzing the vocabulary usage in a text dataset.
• Identifying the most important words in customer feedback or social media posts.
• Comparing word frequencies across different texts or corpora.

Word Frequency Distribution Plot

Network graphs visualize the relationships between words or entities in a text dataset. Nodes represent words or entities, and edges represent the relationships between them.

• Analyzing co-occurrence of words in a text.
• Exploring relationships between entities in a document.

Install necessary Libraries:

Network Graphs

• Social Media Analysis: Visualizing the frequency of hashtags or keywords in tweets to understand trending topics.
• Customer Feedback: Using sentiment analysis visualizations to gauge customer satisfaction from reviews or survey responses.
• Academic Research: Topic modeling visualizations to summarize the main themes in a large set of academic papers.
• Market Research: Word clouds to highlight key terms in consumer opinions or competitor analysis reports.
• News Analysis: Network diagrams to show relationships between entities mentioned in news articles.

Text data visualization is a powerful tool for unlocking the hidden potential of textual information. By applying the right visualization techniques, you can extract valuable insights, gain deeper understanding, and effectively communicate your findings to a broader audience. As text data continues to grow in volume and importance, text data visualization will play an increasingly crucial role in extracting knowledge and making informed decisions.

Please Login to comment...

Similar reads.

• AI-ML-DS With Python
• Data Science Blogathon 2024

Improve your Coding Skills with Practice

What kind of Experience do you want to share?

• Research Article
• Neuroscience

Factorized visual representations in the primate visual system and deep neural networks

• Jack W Lindsey

• Zuckerman Mind Brain Behavior Institute, Columbia University, United States ;
• Department of Neuroscience, Columbia University, United States ;
• Open access
• Copyright information

Share this article

Cite this article.

• Elias B Issa
• Copy to clipboard
• Download BibTeX
• Download .RIS

eLife assessment

Elife digest, introduction, data availability, article and author information.

Object classification has been proposed as a principal objective of the primate ventral visual stream and has been used as an optimization target for deep neural network models (DNNs) of the visual system. However, visual brain areas represent many different types of information, and optimizing for classification of object identity alone does not constrain how other information may be encoded in visual representations. Information about different scene parameters may be discarded altogether (‘invariance’), represented in non-interfering subspaces of population activity (‘factorization’) or encoded in an entangled fashion. In this work, we provide evidence that factorization is a normative principle of biological visual representations. In the monkey ventral visual hierarchy, we found that factorization of object pose and background information from object identity increased in higher-level regions and strongly contributed to improving object identity decoding performance. We then conducted a large-scale analysis of factorization of individual scene parameters – lighting, background, camera viewpoint, and object pose – in a diverse library of DNN models of the visual system. Models which best matched neural, fMRI, and behavioral data from both monkeys and humans across 12 datasets tended to be those which factorized scene parameters most strongly. Notably, invariance to these parameters was not as consistently associated with matches to neural and behavioral data, suggesting that maintaining non-class information in factorized activity subspaces is often preferred to dropping it altogether. Thus, we propose that factorization of visual scene information is a widely used strategy in brains and DNN models thereof.

The study makes a valuable empirical contribution to our understanding of visual processing in primates and deep neural networks, with a specific focus on the concept of factorization. The analyses provide convincing evidence that high factorization scores are correlated with neural predictivity. This work will be of interest to systems neuroscientists studying vision and could inspire further research that ultimately may lead to better models of or a better understanding of the brain.

• Read the peer reviews
• About eLife assessments

When looking at a picture, we can quickly identify a recognizable object, such as an apple, applying a single word label to it. Although extensive neuroscience research has focused on how human and monkey brains achieve this recognition, our understanding of how the brain and brain-like computer models interpret other complex aspects of a visual scene – such as object position and environmental context – remains incomplete.

In particular, it was not clear to what extent object recognition comes at the expense of other important scene details. For example, various aspects of the scene might be processed simultaneously. On the other hand, general object recognition may interfere with processing of such details.

To investigate this, Lindsey and Issa analyzed 12 monkey and human brain datasets, as well as numerous computer models, to explore how different aspects of a scene are encoded in neurons and how these aspects are represented by computational models. The analysis revealed that preventing effective separation and retention of information about object pose and environmental context worsened object identification in monkey cortex neurons. In addition, the computer models that were the most brain-like could independently preserve the other scene details without interfering with object identification.

The findings suggest that human and monkey high level ventral visual processing systems are capable of representing the environment in a more complex way than previously appreciated. In the future, studying more brain activity data could help to identify how rich the encoded information is and how it might support other functions like spatial navigation. This knowledge could help to build computational models that process the information in the same way, potentially improving their understanding of real-world scenes.

Artificial deep neural networks (DNNs) are the most predictive models of neural responses to images in the primate high-level visual cortex ( Cadieu et al., 2014 ; Schrimpf et al., 2020 ). Many studies have reported that DNNs trained to perform image classification produce internal feature representations broadly similar to those in areas V4 and IT of the primate cortex, and that this similarity tends to be greater in models with better classification performance ( Yamins et al., 2014 ). However, it remains opaque what aspects of the representations of these more performant models drive them to better match neural data. Moreover, beyond a certain threshold level of object classification performance, further improvement fails to produce a concomitant improvement in predicting primate neural responses ( Schrimpf et al., 2020 ; Nonaka et al., 2021 ; Linsley, 2023 ). This weakening trend motivates finding new normative principles, besides object classification ability, that push models to better match primate visual representations.

One strategy for achieving high object classification performance is to form neural representations that discard some (are tolerant to) or all (are invariant to) information besides object class. Invariance in neural representations is in some sense a zero-sum strategy: building invariance to some parameters improves the ability to decode others. We also note that our use of ‘invariance’ in this context refers to invariance in neural representations, rather than behavioral or perceptual invariance ( DiCarlo and Cox, 2007 ). However, high-level cortical neurons in the primate ventral visual stream are known to simultaneously encode many forms of information about visual input besides object identity, such as object pose ( Freiwald and Tsao, 2010 ; Hong et al., 2016 ; Kravitz et al., 2013 ; Peters and Kriegeskorte, 2021 ). In this work, we seek to characterize how the brain simultaneously represents different forms of information.

In particular, we introduce methods to quantify the relationships between different types of visual information in a population code (e.g., object pose vs. camera viewpoint), and specifically the degree to which different forms of information are ‘factorized’. Intuitively, if the variance driven by one parameter is encoded along orthogonal dimensions of population activity space compared to the variance driven by other scene parameters, we say that this representation is factorized. We note that our definition of factorization is closely related to the existing concept of manifold disentanglement ( DiCarlo and Cox, 2007 ; Chung et al., 2018 ) and can be seen as a generalization of disentanglement to high-dimensional visual scene parameters like object pose. Factorization can enable simultaneous decoding of many parameters at once, supporting diverse visually guided behaviors (e.g., spatial navigation, object manipulation, or object classification) ( Johnston and Fusi, 2023 ).

Using existing neural datasets, we found that both factorization of and invariance to object category and position information increase across the macaque ventral visual cortical hierarchy. Next, we leveraged the flexibility afforded by in silico models of visual representations to probe different forms of factorization and invariance in more detail, focusing on several scene parameters of interest: background content, lighting conditions, object pose, and camera viewpoint. Across a broad library of DNN models that varied in their architecture and training objectives, we found that factorization of all of the above scene parameters in DNN feature representations was positively correlated with models’ matches to neural and behavioral data. Interestingly, while neural invariance to some scene parameters (background scene and lighting conditions) predicted neural fits, invariance to others (object pose and camera viewpoint) did not. Our results generalized across both monkey and human datasets using different measures (neural spiking, fMRI, and behavior; 12 datasets total) and could not be accounted for by models’ classification performance. Thus, we suggest that factorized encoding of multiple behaviorally relevant scene variables is an important consideration, alongside other desiderata such as classification performance, in building more brain-like models of vision.

Disentangling object identity manifolds in neural population responses can be achieved by qualitatively different strategies. These include building invariance of responses to non-identity scene parameters (or, more realistically, partial invariance; DiCarlo and Cox, 2007 ) and/or factorizing non-identity-driven response variance into isolated (factorized) subspaces ( Figure 1A , left vs. center panels, cylindrical/spherical-shaded regions represent object manifolds). Both strategies maintain an ‘identity subspace’ in which object manifolds are linearly separable. In a non-invariant, non-factorized representation, other variables like camera viewpoint also drive variance within the identity subspace, ‘entangling’ the representations of the two variables ( Figure 1A , right; viewpoint-driven variance is mainly in identity subspace, orange flat-shaded region).

Framework for quantifying factorization in neural and model representations.

( A ) A subspace for encoding a variable, for example, object identity, in a linearly separable manner can be achieved by becoming invariant to non-class variables (compact spheres, middle column, where the volume of the sphere corresponds to the degree of neural invariance, or tolerance, for non-class variables; colored dots represent example images within each class) and/or by encoding variance induced by non-identity variables in orthogonal neural axes to the identity subspace (extended cylinders, left column). Only the factorization strategy simultaneously represents multiple variables in a disentangled fashion. A code that is sensitive to non-identity parameters within the identity subspace corrupts the ability to decode identity (right column) (identity subspace denoted by orange plane). ( B ) Variance across images within a class can be measured in two different linear subspaces: that containing the majority of variance for all other parameters ( a, ‘other_param_subspace’ ) and that containing the majority of the variance for that parameter ( b, ‘param_subspace’ ). Factorization is defined as the fraction of parameter-induced variance that avoids the other-parameter subspace (left). By contrast, invariance to the parameter of interest is computed by comparing the overall parameter-induced variance to the variance in response to other parameters ( c, ‘var_other_param’ ) (right). ( C ) In a simulation of coding strategies for two binary variables out of 10 total dimensions that are varying (see ‘Methods ’ ), a decrease in orthogonality of the relationship between the encoding of the two variables (alignment a > 0, or going from a square to a parallelogram geometry), despite maintaining linear separability of variables, results in poor classifier performance in the few training-samples regime when i.i.d. Gaussian noise is present in the data samples (only 3 of 10 dimensions used in simulation are shown).

To formalize these different representational strategies, we introduced measures of factorization and invariance to scene parameters in neural population responses ( Figure 1B ; see Equations 2–4 in ‘Methods’). Concretely, invariance to a scene variable (e.g., object motion) is computed by measuring the degree to which varying that parameter alone changes neural responses, relative to the changes induced by varying other parameters (lower relative influence on neural activity corresponds to higher invariance, or tolerance, to that parameter). Factorization is computed by identifying the axes in neural population activity space that are influenced by varying the parameter of interest and assessing how much it overlaps the axes influenced by other parameters (‘ a’ in Figure 1B and C ; lower overlap corresponds to higher factorization). We quantified this overlap in two different ways (‘principal components analysis (PCA)-based’ and ‘covariance-based’ factorization, corresponding to Equations 2 and 4 in ‘Methods’), which produced similar results when compared in subsequent analyses (unless otherwise noted, factorization scores will generally refer to the PCA-based method, and the covariance method is shown in Figures 5–7 for comparison). Intuitively, a neural population in which one neural subpopulation encodes object identity and another separate subpopulation encodes object position exhibits a high degree of factorization of those two parameters (however, note that factorization may also be achieved by neural populations with mixed selectivity in which the ‘subpopulations’ correspond to subspaces, or independent orthogonal linear projections, of neural activity space rather than physical subpopulations). Though the example presented in Figure 1 focused on factorization of and invariance to object identity versus non-identity variables, we stress that our definitions can be applied to any scene variables of interest. Furthermore, we presented a simplified visual depiction of the geometry within each scene variable subspace in Figure 1 . We emphasize that our factorization metric does not require a particular geometry within a variable’s subspace, whether parallel linearly ordered coding of viewpoint as in the cylindrical class manifolds shown in Figure 1A and B , or a more complex geometry where there is a lack of parallelism and/or a more nonlinear layout.

While factorization and invariance are not mutually exclusive representational strategies, they are qualitatively different. Factorization, unlike invariance, has the potential to enable the simultaneous representation of multiple scene parameters in a decodable fashion. Intuitively, factorization increases with higher dimensionality as this decreases overlap, all other things being equal (in the limit, the angle between points will approach 90 o or a fully orthogonal code in high dimensions), and for a given finite, fixed dimension, factorization is mainly driven by the angle between this dimension and the other variable subspaces which measures the degree of contamination ( Figure 1C ; square vs. parallelogram). In a simulation, we found that the extent to which the variables of interest were represented in a factorized way (i.e., along orthogonal axes, rather than correlated axes) influenced the ability of a linear discriminator to successfully decode both variables in a generalizable fashion from a few training samples ( Figure 1C ).

Given the theoretically desirable properties of factorized representations, we next asked whether such representations are observed in neural data, and how much factorization contributes empirically to downstream decoding performance in real data. Specifically, we took advantage of an existing dataset in which the tested images independently varied object identity versus object pose plus background context ( Majaj et al., 2015 ; https://github.com/brain-score/vision/blob/master/examples/data_metrics_benchmarks.ipynb ). We found that both V4 and IT responses exhibited more significant factorization of object identity information from non-identity information than a shuffle control (which accounts for effects on factorization due to dimensionality of these regions) ( Figure 2—figure supplement 1 ; see ’Methods’). Furthermore, the degree of factorization increased from V4 to IT ( Figure 2A ). Consistent with prior studies, we also found that invariance to non-identity information increased from V4 to IT in our analysis ( Figure 2A , right, solid lines; Rust and DiCarlo, 2010 ). Invariance to non-identity information was even more pronounced when measured in the subspace of population activity capturing the bulk (90%) of identity-driven variance as a consequence of increased factorization of identity from non-identity information ( Figure 2A , right, dashed lines).

Benefit of factorization to neural decoding in macaque V4 and IT.

( A ) Factorization of object identity and position increased from macaque V4 to IT (PCA-based factorization, see ‘Methods’; dataset E1 – multiunit activity in macaque visual cortex) (left). Like factorization, invariance also increased from V4 to IT (note, ‘identity’ refers to invariance to all non-identity position factors, solid black line) (right). Combined with increased factorization of the remaining variance, this led to higher invariance within the variable’s subspace (orange lines), representing a neural subspace for identity information with invariance to nuisance parameters which decoders can target for read-out. ( B ) An experiment to test the importance of factorization for supporting object class decoding performance in neural responses. We applied a transformation to the neural data (linear basis rotation) that rotated the relative positions of mean responses to object classes without changing the relative proportion of within- vs. between-class variance (Equation 1 in ’Methods’). This transformation preserved invariance to non-class factors (leftmost pair of bars in each plot), while decreasing factorization of class information from non-class factors (center pair of bars in each plot). Concurrently, it had the effect of significantly reducing object class decoding performance (light vs. dark red bars in each plot, chance = 1/64; n = 128 multi-unit sites in V4 and 128 in IT).

To illustrate the beneficial effect of factorization on decoding performance, we performed a statistical lesion experiment that precisely targeted this aspect of representational geometry. Specifically, we analyzed a transformed neural representation obtained by rotating the population data so that inter-class variance more strongly overlapped with the principal components (PCs) of the intra-class variance in the data (see Equation 1 in ’Methods’). Note that this transformation, designed to decrease factorization, acts on the angle between latent variable subspaces. The applied linear basis rotation leaves all other activity statistics completely intact (such as mean neural firing rates, covariance structure of the population, and its invariance to non-class variables) yet has the effect of strongly reducing object identity decoding performance in both V4 and IT ( Figure 2B ). Our analysis shows that maintaining invariance alone in the neural population code was insufficient to account for a large fraction of decoding performance in high-level visual cortex; factorization of non-identity variables is key to the decoding performance achieved by V4 and IT representations.

We next asked whether factorization is found in DNN model representations and whether this novel, heretofore unconsidered metric, is a strong indicator of more brainlike models. When working with computational models, we have the liberty to test an arbitrary number of stimuli; therefore, we could independently vary multiple scene parameters at sufficient scale to enable computing factorization and invariance for each, and we explored factorization in DNN model representations in more depth than previously measured in existing neural experiments. To gain insight back into neural representations, we also assessed the ability of each model to predict separately collected neural and behavioral data. In this fashion, we may indirectly assess the relative significance of geometric properties like factorization and invariance to biological visual representations – if, for instance, models with more factorized representations consistently match neural data more closely, we may infer that those neural representations likely exhibit factorization themselves ( Figure 3 ). To measure factorization, invariance, and decoding properties of DNN models, we generated an augmented image set, based on the images used in the previous dataset ( Figure 2 ), in which we independently varied the foreground object identity, foreground object pose, background identity, scene lighting, and 2D scene viewpoint. Specifically for each base image from the original dataset, we generated sets of images that varied exactly one of the above scene parameters while keeping the others constant, allowing us to measure the variance induced by each parameter relative to the variance across all scene parameters ( Figure 3 , top left; 100 base scenes and 10 transformed images for each source of variation). We presented this large image dataset to models (4000 images total) to assess the relative degree of representational factorization of and invariance to each scene parameter. We conducted this analysis across a broad range of DNNs varying in architecture and objective as well as other implementational choices to obtain the widest possible range of DNN representations for testing our hypothesis. These included models using supervised training for object classification ( Krizhevsky et al., 2012 ; He et al., 2016 ), contrastive self-supervised training ( He et al., 2020 ; Chen et al., 2020 ), and self-supervised models trained using auxiliary objective functions ( Tian et al., 2019 ; Doersch et al., 2015 ; He et al., 2017 ; Donahue and Simonyan, 2019 ; see ’Methods’ and Supplementary file 1b ).

Measurement of factorization in deep neural network (DNN) models and comparison to brain data.

Schematic showing how meta-analysis on models and brain data was conducted by first computing various representational metrics on models and then measuring a model’s predictive power across a variety of datasets. For computing the representational metrics of factorization of and invariance to a scene parameter, variance in model responses was induced by individually varying each of four scene parameters (n = 10 parameter levels) for each base scene (n = 100 base scenes) (see images on the top left). The combination of model-layer metric and model-layer dataset predictivity for a choice of model, layer, metric, and dataset specifies the coordinates of a single dot on the scatter plots in Figures 4 and 7 , and the across-model correlation coefficient between a particular representational metric and neural predictivity for a dataset summarizes the potential importance of the metric in producing more brainlike models (see Figures 5 and 6 ).

First, we asked whether, in the course of training, DNN models develop factorized representations at all. We found that the final layers of trained networks exhibited consistent increases in factorization of all tested scene parameters relative to a randomly initialized (untrained) baseline with the same architecture ( Figure 4A , top row, rightward shift relative to black cross, a randomly initialized ResNet-50). By contrast, training DNNs produced mixed effects on invariance, typically increasing it for background and lighting but reducing it for object pose and camera viewpoint ( Figure 4A , bottom row, leftward shift relative to black cross for left two panels). Moreover, we found that the degree of factorization in models correlated with the degree to which they predicted neural activity for single-unit IT data ( Figure 4A , top row), which can be seen as correlative evidence that neural representations in IT exhibit factorization of all scene variables tested. Interestingly, we saw a different pattern for representational invariance to a scene parameter. Invariance showed mixed correlations with neural predictivity ( Figure 4A , bottom row), suggesting that IT neural representations build invariance to some scene information (background and lighting) but not to others (object pose and observer viewpoint). Similar effects were observed when we assessed correlations between these metrics and fits to human behavioral data rather than macaque neural data ( Figure 4B ).

Neural and behavioral predictivity of models versus their factorization and invariance properties.

( A ) Scatter plots, for example, neural dataset (IT single units, macaque E2 dataset) showing the correlation between a model’s predictive power as an encoding model for IT neural data versus a model’s ability to factorize or become invariant to different scene parameters (each dot is a different model, using each model’s penultimate layer). Note that factorization (PCA-based, see ‘Methods’) in trained models is consistently higher than that for an untrained, randomly initialized Resnet-50 DNN architecture (rightward shift relative to black cross). Invariance to background and lighting but not to object pose and viewpoint increased in trained models relative to the untrained control (rightward versus leftward shift relative to black cross). ( B ) Same as ( A ) except for human behavior performance patterns across images (human I2 dataset). Increasing scene parameter factorization in models generally correlated with better neural predictivity (top row). A noticeable drop in neural predictivity was seen for high levels of invariance to object pose (bottom row, second panel).

To assess the robustness of these findings to choice of images and brain regions used in an experiment, we conducted the same analyses across a large and diverse set of previously collected neural and behavioral datasets, from different primate species and visual regions (six macaque datasets [ Majaj et al., 2015 ; Rust and DiCarlo, 2012 ; Rajalingham et al., 2018 ]: two V4, two ITC (inferior temporal cortex), and two behavior; six human datasets [ Rajalingham et al., 2018 ; Kay et al., 2008 ; Shen et al., 2019 ]: two V4, two HVC (higher visual cortex), and two behavior; Supplementary file 1a ). Consistently, increased factorization of scene parameters in model representations correlated with models being more predictive of neural spiking responses, voxel BOLD signal, and behavioral responses to images ( Figure 5A , black bars; see Figure 4—figure supplements 1 – 3 for scatter plots across all datasets). Although invariance to appearance factors (background identity and scene lighting) correlated with more brainlike models, invariance for spatial transforms (object pose and camera viewpoint) consistently did not (zero or negative correlation values; Figure 5C , red and green open circles). Our results were preserved when we re-ran the analyses using only the subset of models with the identical ResNet-50 architecture ( Figure 5—figure supplement 1 ) or when we evaluated model predictivity using representational dissimilarity matrices of the population (RDMs) instead of linear regression (encoding) fits of individual neurons or voxels ( Figure 5—figure supplement 2 ). Furthermore, the main finding of a positive correlation between factorization and neural predictivity was robust to the particular choice of PCA threshold we used to quantify factorization ( Figure 5—figure supplement 3 ). We found similar results using a covariance-based method for computing factorization that does not have any free parameters ( Figure 5C , faded filled circles; see Equations 4 in ‘Methods’).

Scene parameter factorization correlates with more brainlike deep neural network (DNN) models.

( A ) Factorization of scene parameters in model representations computed using the PCA-based method consistently correlated with a model being more brainlike across multiple independent datasets measuring monkey neurons, human fMRI voxels, or behavioral performance in both macaques and humans (left vs. right column) (black bars). By contrast, increased invariance to camera viewpoint or object pose was not indicative of brainlike models (gray bars). In all cases, model representational metric and neural predictivity score were computed by averaging scores across the last 5 model layers. ( B ) Instead of computing factorization scores using our synthetic images ( Figure 3 , top left), recomputing camera viewpoint or object pose factorization from natural movie datasets that primarily contained camera or object motion, respectively, gave similar results for predicting which model representations would be more brainlike (right: example movie frames; also see ’Methods’). Error bars in ( A and B ) are standard deviations over bootstrapped resampling of the models. ( C ) Summary of the results from ( A ) across datasets (x-axis) for invariance (open symbols) versus factorization (closed symbols) (for reference, ‘ x ’ symbols indicate predictive power when using model classification performance). Results using a comparable, alternative method for computing factorization (covariance-based, Equation 4 in ’Methods’; light closed symbols) are shown adjacent to the original factorization metric (PCA-based, Equation 2 in ‘Methods’; dark closed symbols).

Finally, we tested whether our results generalized across the particular image set used for computing the model factorization scores in the first place. Here, instead of relying on our synthetically generated images, where each scene parameter was directly controlled, we re-computed factorization from two types of relatively unconstrained natural movies, one where the observer moves in an urban environment (approximates camera viewpoint changes) ( Lee et al., 2012 ) and another where objects move in front of a fairly stationary observer (approximates object pose changes) ( Monfort, 2019 ). Similar to the result found for factorization measured using augmentations of synthetic images, factorization of frame-by-frame variance (local in time, presumably dominated by either observer or camera motion; see ‘Methods’) from other sources of variance across natural movies (non-local in time) was correlated with improved neural predictivity in both macaque and human data while invariance to local frame-by-frame differences was not ( Figure 5B ; black versus gray bars). Thus, we have shown that a main finding – the importance of object pose and camera viewpoint factorization for achieving brainlike representations – holds across types of brain signal (spiking vs. BOLD), species (monkey vs. human), cortical brain areas (V4 vs. IT), images for testing in experiments (synthetic, grayscale vs. natural, color), and image sets for computing the metric (synthetic images vs. natural movies).

Our analysis of DNN models provides strong evidence that greater factorization of a variety of scene variables is consistently associated with a stronger match to neural and behavioral data. Prior work had identified a similar correlation between object classification performance (measured fitting a decoder for object class using model representations) and fidelity to neural data ( Yamins et al., 2014 ). A priori, it is possible that the correlations we have demonstrated between scene parameter factorization and neural fit can be entirely captured by the known correlation between classification performance and neural fits ( Schrimpf et al., 2020 ; Yamins et al., 2014 ) as factorization and classification may themselves be correlated. However, we found that factorization scores significantly boosted cross-validated predictive power of neural/behavioral fit performance compared to simply using object classification alone, and factorization boosted predictive power as much if not slightly more when using RDMs instead of linear regression fits to quantify the match to the brain/behavior ( Figure 6 ). Thus, considering factorization in addition to object classification performance improves upon our prior understanding of the properties of more brainlike models ( Figure 7 ).

Scene parameter factorization combined with object identity classification improves correlations with neural predictivity.

Average across datasets of brain predictivity of classification (faded black bar), dimensionality (faded pink bar), and factorization (remaining faded colored bars) in a model representation. Linearly combining factorization with classification in a regression model (unfaded bars at right) produced significant improvements in predicting the most brainlike models (performance cross-validated across models and averaged across datasets, n = 4 datasets for each of V4, IT/HVC and behavior). The boost from factorization in predicting the most brainlike models was not observed for neural and fMRI data when combining classification with a model’s overall dimensionality (solid pink bars; compared to black dashed line for brain predictivity when using classification alone). Results are shown for both the PCA-based and covariance-based factorization metric (top versus bottom row). Error bars are standard deviations over bootstrapped resampling of the models.

Combining classification performance with object pose factorization improves predictions of the most brainlike models on IT/HVC data.

Example scatter plots for neural and fMRI datasets (macaque E1 and E2, IT multi units and single units; human F1 and F2, fMRI voxels) showing a saturating and sometimes reversing trend in neural (voxel) predictivity for models that are increasingly good at classification (top row). This saturating/reversing trend is no longer present when adding object pose factorization to classification as a combined, predictive metric for brainlikeness of a model (middle and bottom rows). The x-axis of each plot indicates the predicted encoding fit or representational dissimilarity matrix (RDM) correlation after fitting a linear regression model with the indicated metrics as input (either classification or classification + factorization).

Object classification, which has been proposed as a normative principle for the function of the ventral visual stream, can be supported by qualitatively different representational geometries ( Yamins et al., 2014 ; Nayebi, 2021 ). These include representations that are completely invariant to non-class information ( Caron et al., 2019b ; Caron, 2019a ) and representations that retain a high-dimensional but factorized encoding of non-class information, which disentangles the representation of multiple variables ( Figure 1A ). Here, we presented evidence that factorization of non-class information is an important strategy used, alongside invariance, by the high-level visual cortex ( Figure 2 ) and by DNNs that are predictive of primate neural and behavioral data ( Figures 4 and 5 ).

Prior work has indicated that building representations that support object classification performance and representations that preserve high-dimensional information about natural images are both important principles of the primate visual system ( Cadieu et al., 2014 ; Elmoznino and Bonner, 2022 ; though see Conwell et al., 2022 ). Critically, our results cannot be accounted for by classification performance or dimensionality alone ( Figure 6 , gray and pink bars); that is, the relationship between factorization and matches to neural data was not entirely mediated by classification or dimensionality. That said, we do not regard factorization and dimensionality, or factorization and object classification performance, as mutually exclusive hypotheses for useful principles of visual representations. Indeed, high-dimensional representations could be regarded as a means to facilitate factorization, and likewise factorized representations can better support classification ( Figure 1C ).

Our notion of factorization is related to, but distinct from, several other concepts in the literature. Many prior studies in machine learning have considered the notion of disentanglement, often defined as the problem of inferring independent factors responsible for generating the observed data ( Kim and Mnih, 2018 ; Eastwood and Williams, 2018 ; Higgins, 2018 ). One prior study notably found that machine learning models designed to infer disentangled representations of visual data displayed single-unit responses that resembled those of individual neurons in macaque IT ( Higgins et al., 2021 ). Our definition of factorization is more flexible, requiring only that independent factors be encoded in orthogonal subspaces, rather than by distinct individual neurons. Moreover, our definition applies to generative factors, such as camera viewpoint or object pose, that are multidimensional and context dependent. Factorization is also related to a measure of ‘abstraction’ in representational geometry introduced in a recent line of work ( Bernardi et al., 2020 ; Boyle et al., 2024 ), which is observed to emerge in trained neural networks ( Johnston and Fusi, 2023 ; Alleman et al., 2024 ). In these studies, an abstract representation is defined as one in which variables are encoded and can be decoded in a consistent fashion regardless of the values of other variables. A fully factorized representation should be highly abstract according to this definition, though factorization emphasizes the geometric properties of the population representation while these studies emphasize the consequences for decoding performance in training downstream linear read-outs. Relatedly, another recent study found that orthogonal encoding of class and non-class information is one of several factors that determines few-shot classification performance ( Sorscher et al., 2022 ). Our work can be seen as complementary to work on representational straightening of natural movie trajectories in the population space ( Hénaff et al., 2021 ). This work suggested that visual representations maintain a locally linear code of latent variables like camera viewpoint, while our work focused on the global arrangement of the linear subspaces affected by different variables (e.g., overall coding of camera viewpoint-driven variance versus sources of variance from other scene variables in a movie). Local straightening of natural movies was found to be important for early visual cortex neural responses but not necessarily for high-level visual cortex ( Toosi and Issa, 2022 ), where the present work suggests factorization may play a role.

Our work has several limitations. First, our analysis is primarily correlative. Going forward, we suggest that factorization could prove to be a useful objective function for optimizing neural network models that better resemble primate visual systems, or that factorization of latent variables should at least be a by-product of other objectives that lead to more brain-like models. An important direction for future work is finding ways to directly incentivize factorization in model objective functions so as to test its causal impact on the fidelity of learned representations to neural data. Second, our choice of scene variables to analyze in this study was heuristic and somewhat arbitrary. Future work could consider unsupervised methods (in the vein of independent components analysis) for uncovering the latent sources of variance that generate visual data, and assessing to what extent these latent factors are encoded in factorized form. Third, in our work we do not specify the details of how a particular scene parameter is encoded within its factorized subspace, including whether the code is linear (‘straightened’) or nonlinear ( Hénaff et al., 2021 ; Hénaff et al., 2019 ). Neural codes could adopt different strategies, resulting in similar factorization scores at the population level, each with some support in visual cortex literature: (1) each neuron encodes a single latent variable ( Field, 1994 ; Chang and Tsao, 2017 ), (2) separate brain subregions encode qualitatively different latent variables but using distributed representations within each region ( Tsao et al., 2006 ; Lafer-Sousa and Conway, 2013 ; Vaziri et al., 2014 ), and (3) each neuron encodes multiple variables in a distributed population code, such that the factorization of different variables is only apparent as independent directions when assessed in high-dimensional population activity space ( Field, 1994 ; Rigotti et al., 2013 ). Future work can disambiguate among these possibilities by systematically examining ventral visual stream subregions ( Kravitz et al., 2013 ; Vaziri et al., 2014 ; Kravitz et al., 2011 ) and the single neuron tuning curves within them ( Leopold et al., 2006 ; Freiwald et al., 2009 ).

Monkey datasets

Macaque monkey datasets were of single-unit neural recordings ( Rust and DiCarlo, 2012 ), multi-unit neural recordings ( Majaj et al., 2015 ), and object recognition behavior ( Rajalingham et al., 2018 ). Single-unit spiking responses to natural images were measured in V4 and anterior ventral IT ( Rust and DiCarlo, 2012 ). The advantages of this dataset are that it contains well-isolated single neurons, the gold standard for electrophysiology. Furthermore, the IT recordings were obtained from penetrating electrodes targeting the anterior ventral portion of IT near the base of skull, reflecting the highest level of the IT hierarchy. On the other hand, the multi-unit dataset was obtained from across IT with a bias toward where multi-unit arrays are more easily placed such as CIT and PIT ( Majaj et al., 2015 ), complementing the recording locations of the single-unit dataset. An advantage of the multi-unit dataset using chronic recording arrays is that an order of magnitude more images were tested per recording site (see dataset comparisons in Supplementary file 1a ). Finally, the monkey behavioral dataset came from a third study examining the image-by-image object classification performance of macaques and humans ( Rajalingham et al., 2018 ).

Human datasets

Three datasets from humans were used, two fMRI datasets and one object recognition behavior dataset ( Nonaka et al., 2021 ; Rajalingham et al., 2018 ; Kay et al., 2008 ). The fMRI datasets used different images (color versus grayscale) but otherwise used a fairly similar number of images and voxel resolution in MR imaging. Human fMRI studies have found that different DNN layers tend to map to V4 and HVC human fMRI voxels ( Nonaka et al., 2021 ). The human behavioral dataset measured image-by-image classification performance and was collected in the same study as the monkey behavioral signatures ( Rajalingham et al., 2018 ).

Computational models

In recent years, a variety of approaches to training DNN vision models have been developed that learn representations that can be used for downstream classification (and other) tasks. Models differ in a variety of implementational choices including in their architecture, objective function, and training dataset. In the models we sampled, objectives included supervised learning of object classification (AlexNet, ResNet), self-supervised contrastive learning (MoCo, SimCLR), and other unsupervised learning algorithms based on auxiliary tasks (e.g., reconstruction or colorization). A majority of the models that we considered relied on the widely used, performant ResNet-50 architecture, though some in our library utilized different architectures. The randomly initialized network control utilized ResNet-50 (see Figure 4A and B ). The set of models we used is listed in Supplementary file 1b .

Simulation of factorized versus non-factorized representational geometries

For the simulation in Figure 1C , we generated data in the following way. First, we randomly sampled the values of N = 10 binary features. Feature values corresponded to positions in an N-dimensional vector space as follows: each feature was assigned an axis in N-dimensional space, and the value of each feature (+1 or –1) was treated as a coefficient indicating the position along that axis. All but two of the feature axes were orthogonal to the rest. The last two features, which served as targets for the trained linear decoders, were assigned axes whose alignment ranged from 0 (orthogonal) to 1 (identical). In the noiseless case, factorization of these two variables with respect to one another is given by subtracting the square of the cosine of the angle between the axes from 1. We added Gaussian noise to the positions of each data point and randomly sampled K positive and negative examples for each variable of interest to use as training data for the linear classifier (a support vector machine).

Macaque neural data analyses

For the shuffle control used as a null model for factorization, we shuffled the object identity labels of the images ( Figure 2—figure supplement 1 ). For the transformation used in Figure 2B , we computed the PCs of the mean neural activity response to each object class (‘class centers,’ x c ), referred to as the inter-class PCs, v 1 inter , v 2 inter , …, v N inter . We also computed the PCs of the data with corresponding class centers subtracted (i.e., x - x c ) from each activity pattern, referred to as the intra-class PCs v 1 intra , v 2 intra , …, v N intra . We transformed the data by applying to the class centers a change of basis matrix W inter→intra that rotated each inter-class PC into the corresponding intra-class PC: W inter→intra = v 1 intra ( v 1 inter ) T + …1 v N intra ( v N inter ) T . That is, the class centers were transformed by this matrix, but the relative positions of activity patterns within a given class were fixed. For an activation vector x belonging to a class c for which the average activity vector over all images of class c is x c , the transformed vector was

This transformation has the effect of preserving intra-class variance statistics exactly from the original data and preserving everything about the statistics of inter-class variance except its orientation relative to intra-class variance. That is, the transformation is designed to affect (specifically decrease) factorization while controlling for all other statistics of the activity data that may be relevant to object classification performance (considering the simulation in Figure 1C of two binary variables, this basis change of the neural data in Figure 2B is equivalent to turning a square into the maximally flat parallelogram, the degenerate one where all the points are collinear).

Scene parameter variation

Our generated scenes consisted of foreground objects imposed upon natural backgrounds. To measure variance associated with a particular parameter like the background identity, we randomly sampled 10 different backgrounds while holding the other variables (e.g., foreground object identity and pose constant). To measure variance associated with foreground object pose, we randomly varied object angle from [–90, 90] along all three axes independently, object position on the two in-plane axes, horizontal [–30%, 30%] and vertical [–60%, 60%], and object size [×1/1.6, ×1.6]. To measure variance associated with camera position, we took crops of the image with scale uniformly varying from 20 to 100% of the image size, and position uniformly distributed across the image. To measure variance associated with lighting conditions, we applied random jitters to the brightness, contrast, saturation, and hue of an image, with jitter value bounds of [–0.4, 0.4] for brightness, contrast, and saturation and [–0.1, 0.1] for hue. These parameter choices follow standard data augmentation practices for self-supervised neural network training, as used, for example, in the SimCLR and MoCo models tested here ( He et al., 2020 ; Chen et al., 2020 ).

Factorization and invariance metrics

Factorization and invariance were measured according to the following equations:

Variance induced by a parameter ( var param ) is computed by measuring the variance (summed across all dimensions of neural activity space) of neural responses to the 10 augmented versions of a base image where the augmentations are those obtained by varying the parameter of interest. This quantity is then averaged across the 100 base images. The variance induced by all parameters is simply the sum of the variances across all images and augmentations. To define the ‘other-parameter subspace,’ we averaged neural responses for a given base image over all augmentations using the parameter of interest, and ran PCA on the resulting set of averaged responses. The subspace was defined as the space spanned by top PCA components containing 90% of the variance of these responses. Intuitively, this space captures the bulk of the variance driven by all parameters other than the parameter of interest (due to the averaging step). The variance of the parameter of interest within this ‘other-parameter subspace,’ var param|other_param_subspace , was computed the same way as var param but using the projections of neural activity responses onto the other-parameter subspace. In the main text, we refer to this method of computing factorization as PCA-based factorization.

We also considered an alternative definition of factorization referred to as covariance-based factorization. In this alternative definition, we measured the covariance matrices cov param and cov other_param induced by varying (in the same fashion as above) the parameter of interest, and all other parameters. Factorization was measured by the following equation:

This is equal to 1 minus the dot product between the normalized, flattened covariance matrices, and thus covariance-based factorization is a measure of the discrepancy of the covariance structure induced by the parameter of interest and other parameters. The main findings were unaffected by our choice of method for computing the factorization metric, whether PCA or covariance based ( Figures 5 — 7 ). An advantage of the PCA-based method is that as an intermediate one recovers the linear subspaces containing parameter variance, but in so doing requires an arbitrary choice of the explained variance threshold used to choose the number of PCs. By contrast, the covariance-based method is more straightforward to compute and has no free parameters. Thus, these two metrics are complementary and somewhat analogous in methodology to two metrics commonly used for measuring dimensionality (the number of components needed to explain a certain fraction of the variance, analogous to our original PCA-based definition, and the participation ratio, analogous to our covariance-based definition) ( Ding and Glanzman, 2010 ; Litwin-Kumar et al., 2017 ).

Natural movie factorization metrics

For natural movies, variance is not induced by explicit control of a parameter as in our synthetic scenes but implicitly, by considering contiguous frames (separated by 200 ms in real time) as reflective of changes in one of two motion parameters (object versus observer motion) depending on how stationary the observer is (MIT Moments in Time movie set: stationary observer; UT-Austin Egocentric movie set: nonstationary) ( Lee et al., 2012 ; Monfort, 2019 ). Here, the all parameters condition is simply the variance across all movie frames, which in the case of MIT Moments in Time dataset includes variance across thousands of video clips taken in many different settings and in the case of the UT-Austin Egocentric movie dataset includes variance across only four movies but over long durations of time during which an observer translates extensively in an environment (3–5 hr). Thus, movie clips in the MIT Moments in Time movie set contained new scenes with different object identities, backgrounds, and lightings and thus effectively captured variance induced by these non-spatial parameters ( Monfort, 2019 ). In the UT-Austin Egocentric movie set, new objects and backgrounds are encountered as the subject navigates around the urban landscape ( Lee et al., 2012 ).

Model neural encoding fits

Linear mappings between model features and neuron (or voxel) responses were computed using ridge regression (with regularization coefficient selected by cross-validation) on a low-dimensional linear projection of model features (top 300 PCA components computed using images in each dataset). We also tested an alternative approach to measuring representational similarity between models and experimental data based on representational similarity analysis ( Kriegeskorte and Kievit, 2013 ), computing dot product similarities of the representations of all pairs of images and measuring the Spearman correlation coefficient between these pairwise similarity matrices obtained from a given model and neural dataset, respectively.

Model behavioral signatures

We followed the approach of Rajalingham et al., 2018 . We took human and macaque behavioral data from the object classification task and used it to create signatures of image-level difficulty (the ‘I1’ vector) and image-by-distractor-object confusion rates (the ‘I2’ matrix). We did the same for the DNN models, extracting model ‘behavior’ by training logistic regression classifiers to classify object identity in the same image dataset used in the experiments of Rajalingham et al., 2018 , using model layer activations as inputs. Model behavioral accuracy rates on image by distractor object pairs were assessed using the classification probabilities output by the logistic regression model, and these were used to compute I1 and I2 metrics as was done for the true behavioral data. Behavioral similarity between models and data was assessed by measuring the correlation between the entries of the I1 vectors and I2 matrices (both I1 and I2 results are reported).

Model layer choices

The scatter plots in Figure 4A and B and Figure 4—figure supplements 1 – 3 use metrics (factorization, invariance, and goodness of neural fit) taken from the final representational layer of the network (the layer prior to the logits layer used for classification in supervised network, prior to the embedding head in contrastive learning models, or prior to any auxiliary task-specific layers in unsupervised models trained using auxiliary tasks). However, representational geometries of model activations, and their match to neural activity and behavior, vary across layers. This variability arises because different model layers correspond to different stages of processing in the model (convolutional layers in some cases, and pooling operations in others), and may even have different dimensionalities. To ensure that our results do not depend on idiosyncrasies of representations in one particular model layer and the particular network operations that precede it, summary correlation statistics in all other figures ( Figures 5 — 7 , Figure 5—figure supplements 1 – 3 ) show the results of the analysis in question averaged over the five final representational layers of the model. That is, the metrics of interest (factorization, invariance, neural encoding fits, RDM correlation, behavioral similarity scores) were computed independently for each of the five final representational layers of each model, and these five values were averaged prior to computing correlations between different metrics.

Correlation of model predictions and experimental data

A Spearman linear correlation coefficient was calculated for each model layer by biological dataset combination (six monkey datasets and six human datasets). Here, we do not correct for noise in the biological data when computing the correlation coefficient as this would require trial repeats (for computing intertrial variability) that were limited or not available in the fMRI data used. In any event, normalizing by the data noise ceiling applies a uniform scaling to all model prediction scores and does not affect model comparison, which only depends on ranking models as being relatively better or worse in predicting brain data. Finally, we estimated the effectiveness of model factorization, invariance, or dimensionality in combination with model object classification performance for predicting model neural and behavioral fit by performing a linear regression on the particular dual metric combination (e.g., classification plus object pose factorization) and reporting the Spearman correlation coefficient of the linearly weighted metric combination. The correlation was assessed on held-out models (80% used for training, 20% for testing), and the results were averaged over 100 randomly sampled train/test splits.

The current manuscript is a computational study, so no data have been generated for this manuscript. Publicly available datasets and models were used. Analysis code is available at https://github.com/issalab/Lindsey-Issa-Factorization , (copy archived at Issa, 2024 ).

• Naselaris T
• Google Scholar
• Siegelbaum SA
• Bojanowski P
• Kornblith S
• Sompolinsky H
• Williams CKI
• Elmoznino E
• Freiwald WA
• Livingstone MS
• Simoncelli EP
• Charlton JA
• Summerfield C
• Botvinick M
• Johnston WJ
• Ungerleider LG
• Kriegeskorte N
• Krizhevsky A
• Sutskever I
• Lafer-Sousa R
• Litwin-Kumar A
• Rajalingham R
• Prescott-Roy J
• Tootell RBH

Author details

• Zuckerman Mind Brain Behavior Institute, Columbia University, New York, United States
• Department of Neuroscience, Columbia University, New York, United States

Contribution

Competing interests.

For correspondence

Doe csgf (de-sc0020347), klingenstein-simons foundation (fellowship in neuroscience), sloan foundation (fellowship), grossman-kavli center at columbia (scholar award).

The funders had no role in study design, data collection and interpretation, or the decision to submit the work for publication.

Acknowledgements

This work was performed on the Columbia Zuckerman Institute Axon GPU cluster and via generous access to Cloud TPUs from Google’s TPU Research Cloud (TRC). JWL was supported by the DOE CSGF (DE-SC0020347). EBI was supported by a Klingenstein-Simons fellowship, Sloan Foundation fellowship, and Grossman-Kavli Scholar Award. We thank Erica Shook for comments on a previous version of the manuscript. The authors declare no competing interests.

Version history

• Preprint posted: August 4, 2023 (view preprint)
• Sent for peer review: August 22, 2023
• Preprint posted: February 2, 2024 (view preprint)
• Preprint posted: June 5, 2024 (view preprint)
• Version of Record published: July 5, 2024 (version 1)

Cite all versions

You can cite all versions using the DOI https://doi.org/10.7554/eLife.91685 . This DOI represents all versions, and will always resolve to the latest one.

© 2024, Lindsey and Issa

This article is distributed under the terms of the Creative Commons Attribution License , which permits unrestricted use and redistribution provided that the original author and source are credited.

• 21 downloads
• 0 citations

Views, downloads and citations are aggregated across all versions of this paper published by eLife.

Download links

Downloads (link to download the article as pdf).

• Article PDF
• Figures PDF

Open citations (links to open the citations from this article in various online reference manager services)

Cite this article (links to download the citations from this article in formats compatible with various reference manager tools), categories and tags.

• visual cortex
• deep neural networks
• neurophysiology
• visual scenes
• object recognition

Research organisms

• Rhesus macaque

Further reading

Early parafoveal semantic integration in natural reading.

Humans can read and comprehend text rapidly, implying that readers might process multiple words per fixation. However, the extent to which parafoveal words are previewed and integrated into the evolving sentence context remains disputed. We investigated parafoveal processing during natural reading by recording brain activity and eye movements using MEG and an eye tracker while participants silently read one-line sentences. The sentences contained an unpredictable target word that was either congruent or incongruent with the sentence context. To measure parafoveal processing, we flickered the target words at 60 Hz and measured the resulting brain responses (i.e. Rapid Invisible Frequency Tagging, RIFT ) during fixations on the pre-target words. Our results revealed a significantly weaker tagging response for target words that were incongruent with the previous context compared to congruent ones, even within 100ms of fixating the word immediately preceding the target. This reduction in the RIFT response was also found to be predictive of individual reading speed. We conclude that semantic information is not only extracted from the parafovea but can also be integrated with the previous context before the word is fixated. This early and extensive parafoveal processing supports the rapid word processing required for natural reading. Our study suggests that theoretical frameworks of natural reading should incorporate the concept of deep parafoveal processing.

Perceptual error based on Bayesian cue combination drives implicit motor adaptation

The sensorimotor system can recalibrate itself without our conscious awareness, a type of procedural learning whose computational mechanism remains undefined. Recent findings on implicit motor adaptation, such as over-learning from small perturbations and fast saturation for increasing perturbation size, challenge existing theories based on sensory errors. We argue that perceptual error, arising from the optimal combination of movement-related cues, is the primary driver of implicit adaptation. Central to our theory is the increasing sensory uncertainty of visual cues with increasing perturbations, which was validated through perceptual psychophysics (Experiment 1). Our theory predicts the learning dynamics of implicit adaptation across a spectrum of perturbation sizes on a trial-by-trial basis (Experiment 2). It explains proprioception changes and their relation to visual perturbation (Experiment 3). By modulating visual uncertainty in perturbation, we induced unique adaptation responses in line with our model predictions (Experiment 4). Overall, our perceptual error framework outperforms existing models based on sensory errors, suggesting that perceptual error in locating one’s effector, supported by Bayesian cue integration, underpins the sensorimotor system’s implicit adaptation.

The breath shape controls intonation of mouse vocalizations

Intonation in speech is the control of vocal pitch to layer expressive meaning to communication, like increasing pitch to indicate a question. Also, stereotyped patterns of pitch are used to create distinct sounds with different denotations, like in tonal languages and, perhaps, the 10 sounds in the murine lexicon. A basic tone is created by exhalation through a constricted laryngeal voice box, and it is thought that more complex utterances are produced solely by dynamic changes in laryngeal tension. But perhaps, the shifting pitch also results from altering the swiftness of exhalation. Consistent with the latter model, we describe that intonation in most vocalization types follows deviations in exhalation that appear to be generated by the re-activation of the cardinal breathing muscle for inspiration. We also show that the brainstem vocalization central pattern generator, the iRO, can create this breath pattern. Consequently, ectopic activation of the iRO not only induces phonation, but also the pitch patterns that compose most of the vocalizations in the murine lexicon. These results reveal a novel brainstem mechanism for intonation.

Be the first to read new articles from eLife

Thank you for visiting nature.com. You are using a browser version with limited support for CSS. To obtain the best experience, we recommend you use a more up to date browser (or turn off compatibility mode in Internet Explorer). In the meantime, to ensure continued support, we are displaying the site without styles and JavaScript.

• View all journals
• Explore content
• About the journal
• Publish with us
• Sign up for alerts
• Open access
• Published: 28 June 2024

Immersive scene representation in human visual cortex with ultra-wide-angle neuroimaging

• Jeongho Park   ORCID: orcid.org/0000-0003-4260-3435 1 ,
• Edward Soucy 2 ,
• Jennifer Segawa 2 ,
• Ross Mair   ORCID: orcid.org/0009-0004-8085-407X 2 , 3 , 4 &
• Talia Konkle   ORCID: orcid.org/0000-0003-1738-4744 1 , 2 , 5  

Nature Communications volume  15 , Article number:  5477 ( 2024 ) Cite this article

336 Accesses

1 Altmetric

Metrics details

• Object vision

While human vision spans 220°, traditional functional MRI setups display images only up to central 10-15°. Thus, it remains unknown how the brain represents a scene perceived across the full visual field. Here, we introduce a method for ultra-wide angle display and probe signatures of immersive scene representation. An unobstructed view of 175° is achieved by bouncing the projected image off angled-mirrors onto a custom-built curved screen. To avoid perceptual distortion, scenes are created with wide field-of-view from custom virtual environments. We find that immersive scene representation drives medial cortex with far-peripheral preferences, but shows minimal modulation in classic scene regions. Further, scene and face-selective regions maintain their content preferences even with extreme far-periphery stimulation, highlighting that not all far-peripheral information is automatically integrated into scene regions computations. This work provides clarifying evidence on content vs. peripheral preferences in scene representation and opens new avenues to research immersive vision.

Similar content being viewed by others

Mapping brain function during naturalistic viewing using high-density diffuse optical tomography

Attention modulates neural representation to render reconstructions according to subjective appearance

Higher-contrast images are better remembered during naturalistic encoding

Introduction.

When we look at the world, we feel immersed in a broader visual environment. For example, the experience of a view of an expansive vista from the top of a mountain is not the same as when looking at a picture of the same view. One key difference is that in the real world, we sense a >180 degrees view of the environment at each glance. Indeed, while our fovea and macula ensure high-resolution input at the center of gaze, there is an equally impressive expanse of peripheral vision: with 170 degrees sensed by a single eye, and up to 220 degrees of the extreme far-periphery sensed by the two eyes combined 1 . What are the neural processes by which this immersive visual experience of the broader environment is constructed in the human visual system?

Seminal research has identified three brain regions in the human brain that show a clear role in high-level visual scene perception 2 , 3 . There are parahippocampal place area (PPA 4 ) in the temporo-occipital cortex, retrosplenial cortex (RSC 5 ) or medial place area (MPA 6 ) in the medial side along the parietal-occipital sulcus, and occipital place area (OPA 7 , 8 ) in the parieto-occipital cortex. Extensive neuroimaging studies have characterized tuning properties of these regions and their complementary roles in scene perception, regarding recognition 9 , 10 , 11 , 12 , 13 and navigation 14 , 15 , 16 , 17 , 18 , 19 , 20 in particular.

However, the constraints of standard fMRI image projection setup have limited scene perception research to the central 10-20 degrees of the visual field, with scene properties inferred from postcard-like picture perception. Thus, it remains unknown how a scene activates the visual system when it is presented across the full visual field, providing a more immersive first-person view. Would this alter the way we define the scene regions along the cortical surface (e.g., a larger cortical extent, or new scene regions)? More generally, what are the neural processes that construct a visual scene representation when far-peripheral information is available?

Here, drawing inspiration from an infant fMRI study 21 , we introduce an innovative image projection setup, which enables the presentation of ultra-wide-angle visual stimuli in an fMRI scanner. In typical scanning setups, stimuli are presented to humans lying supine in the scanner by projecting onto a screen outside of the scanner bore, while the participants look out through a head coil at a small mirror reflecting the screen behind them. With this setup, the maximum visual angle of a projected image is ~15–20 degrees. We modified this setup, by bouncing the projected image off two angled mirrors, directly onto a large, curved screen inside the scanner bore. This allowed us to project images about 175 degrees wide, stimulating almost the entire visual field.

While there have been prior approaches to establish wide-angle presentation, they were mainly centered on studying retinotopic properties in early visual areas, presenting expanding rings and rotating wedges in black and white 22 , 23 , 24 , 25 . Thus, for testing high-level visual areas, which involves presenting more complex images (e.g., faces or scenes), different solutions imposed specific limitations. For example, one approach enabled researchers to project images up to 120 degrees, but only to one eye at a time, and onto a screen that was 3 cm from an eye, requiring participants to view stimuli with a custom contact lens 24 , 25 , 26 , 27 , 28 , 29 . More recently, a high-resolution MR-compatible head mounted display was developed, but the maximum field-of-view is ~52 degrees wide (Nordic Neuro Lab). Our solution was developed with the intention of studying high-level visual perception by providing as expansive and natural visual experience as possible. Further, our approach does not require participants to wear additional devices, and leverages a relatively low-tech solution that can be implemented in other scanning facilities.

With this full-field neuroimaging setup, we first chart the cortex with far-peripheral sensitivity. Then, we leverage this wide-angle setup to entertain questions about what it means to be a scene and the implications for the responses of classic scene-selective regions. For example, perhaps any image content presented in the far-periphery is part of a scene, and should be automatically integrated into the computations of high-level scene regions. From an embodied, ego-centric perspective, this is a reasonable account. Alternatively, perhaps the scene regions are more like high-level pattern analyzers that are sensitive to particular kinds of image statistics (e.g., open/closed spatial layout, contour junctions, etc.) rather than to the retinotopic location of the visual stimulation per se. Indeed, in the scene perception literature, there is evidence for both accounts. The neuroimaging studies with 0–20 degrees of the visual field showed that the classic scene regions are modulated both by the scene content (over other semantic category contents like faces) and by peripheral stimulation 6 , 7 , 30 , 31 , 32 . We now extend the scope of this investigation to the entire visual field and revisit this question.

Ultra-wide-angle fMRI

To accomplish ultra-wide-angle visual presentation in the scanning environment, we installed two angled mirrors near the projector such that the projected image was cast directly into the scanner bore, onto a custom-built curved screen positioned around a person’s head (Fig.  1 , Supplementary Fig.  3) . Additionally, given the visual obstruction of the top of the head coil, we simply removed it, allowing participants to have an unobstructed view of the curved screen. Through signal quality check protocols, we confirmed that the lack of top head coil did not have critical impacts on MRI signals for occipital and parietal cortices (see Supplementary Fig.  1 for more details).

An image is bounced off two angled mirrors and directly projected onto a curved screen inside the scanner bore.

To compensate for the curved screen, we developed code to computationally warp any image, to account for the screen curvature and tilted projection angle (Fig.  2 ). Given the geometrical constraints of our MRI room, only a subset of pixels could be projected onto the screen, resulting in substantially lower image resolution compared to other standard projection systems, particularly in the vertical dimension (see Methods).

A rectangular image (1024 × 768 pixels) was computationally warped to match the size and curvature of the tilted screen. Due to the geometrical constraints of the room, only a subset of pixels could be projected onto the screen (828 × 284 pixels). On the curved screen, the aspect ratio of the original image was maintained on the display surface.

Further, we found that when projecting natural scene images across the full field, using standard pictures taken from typical cameras lead to highly distorted perceptions of space—a picture with a compatible wide field-of-view was required. Thus, for the present studies, we built virtual 3D environments in Unity game engine (Unity Technologies, Version 2017.3.0), where we could control the viewport height and field-of-view when rendering scene images. Further details about the full-field fMRI setup can be found in the Methods and on our website ( https://jpark203.github.io/fullfield-neuroimaging ). Taken together, our solution enabled us to present images over 175 degrees, providing natural and immersive viewing experience.

Full-field eccentricity map

In Experiment 1, we first attempted to map the full visual field and chart an extended eccentricity map along the visual cortex. We used a classic retinotopic mapping protocol, where participants performed a fixation dot color detection task. Flashing checkerboards were presented in rings at five levels of eccentricity: (1) a center circle of 1.8 degrees radius, and (2) the inner and outer rings of 2.0–5.6 degrees, (3) 6.3–16.5 degrees, (4) 18.5–50.3 degrees, and (5) >55.3 degrees radius. The two farthest eccentricities were not possible with typical scanning setups, allowing us to stimulate cortical territory that has been inaccessible via direct visual input.

The cortical map of eccentricity preferences is shown in Fig.  3 . For each voxel, we compared responses to different eccentricity conditions, and colored the voxel based on the condition with the highest activation (hue). The resulting map revealed a systematic progression of preference from the center to far-periphery, covering an expansive cortical territory along the medial surface of the occipital lobe. In particular, we mapped strong responses to far-peripheral stimulation near the parieto-occipital sulcus (POS), extending beyond our typical eccentricity band maps (black dotted line, Fig.  3) . These results validate the technical feasibility of our ultra-wide-angle projection method, and to our knowledge, show the full-field mapping of eccentricity in the human brain that exceeds the scope of prior studies.

An example participant’s right occipital cortex is shown from a medial view. Each voxel is colored based on its preference for one of five eccentricity rings (right). In the group data, the black dotted line shows where a typical eccentricity map would end, and the black arrows show how much more the cortex can be stimulated with full-field neuroimaging. Individual brain maps from nine participants also show a consistent pattern of results. POS parieto-occipital sulcus.

Full-field scene perception

With this full-field neuroimaging set up, we next measured visual system responses to ultra-wide-angle, immersive real-world scenes and compared them to responses from visually-smaller postcard scenes and unstructured image-statistical counterparts.

Specifically, we created four different stimulus conditions that varied in presentation size (full-field vs. postcard), and content (intact vs. phase-scrambled scenes). The full-field images filled up the entire screen (175 deg wide), and the postcard images were presented at the center of screen in a much smaller size (though still 44 deg wide). The chosen size of postcard images was bigger than the maximum size in typical fMRI setups due to limited image resolution. We discuss this limitation further in the Discussion.

To match the image content across presentation sizes, the postcard images were rescaled from the entire full-field images, instead of cropping the center only. To vary the image content, the same scenes were phase-scrambled, preserving the summed spatial frequency energy across the whole image but disrupting all second-order and higher-level image statistics present in scenes 33 , 34 . Additionally, we also included a postcard-face condition where a single face was presented at the center of screen, in a similar visual size to the postcard-scenes. Each stimulus condition was presented in a standard blocked design (12 sec), and participants performed a one-back repetition detection task (see Methods for further details).

First, we asked how the visual cortex responds to full-field size images with intact scene content, compared to images with phase-scrambled scene statistics (Fig.  4 a). This contrast is matched in full-field retinotopic footprint, but different in the image content. Will the immersive full-field scenes recruit additional brain regions, e.g., with more extensive scene regions (in terms of cortical surface area), or separate brain areas away from the classic scene regions that were not found with the traditional fMRI setups due to the limited stimulus size?

The group data is shown on an example subject brain. Zoom-in views at each row are captured around the classic scene regions. a Image content contrast. A large portion of high-level visual areas, including the scene regions, shows higher activation for the intact scenes compared to the phase-scrambled scenes. b Visual size contrast. A large swath of cortex near the parieto-occipital sulcus is strongly activated when viewing a full-field scene compared to a postcard scene. PPA parahippocampal place area, RSC retrosplenial cortex, OPA occipital place area, POS parieto-occipital sulcus.

The whole-brain contrast map is shown with the group data in Fig.  4 a (Supplementary Fig.  4 for individual participants). We found qualitatively higher responses for intact scenes over the scrambled scenes along the ventral medial cortex, as well as dorsal occipito-parietal cortex. For comparison, we defined three scene ROIs by contrasting the postcard-scene vs. postcard-face condition, reflecting a more typical (non-full field) definition of these regions. Fig.  4 a shows the overlaid outlines of these classically-defined ROIs (PPA; OPA; RSC). Note that these ROIs reflect group-level ROIs for visualization, but all ROIs were defined in individual subjects in independent data. Qualitative inspection reveals that these ROIs largely encircle the strongest areas of scene-vs-scrambled response preferences. In other words, it is not the case that the full-field stimulation leads to strong scene content-preferring responses that clearly extend well beyond the postcard-defined ROI boundaries.

One important note is that our postcard-sized stimulus was still rather large (44 degrees) relative to the visual size presented in typical set ups (15-20 degrees). Thus, the present data indicate only that the extent of activated cortical surface is not much increased by a relatively dramatic stimulus size increase from 44 to 175 deg. If there is increasing cortical scene-selective territory as a function of visual angle, it is limited to visual size increases from 15-44 degrees. More detailed parametric visual size mapping is required to answer this question. For the purposes of the present work, these results reveal that the standard contrasts for defining classic scene regions reflect stable functionally defined regions, across both our postcard and full-field presentation sizes.

Next, we asked how the visual cortex responds to full-field scenes compared to postcard scenes. This contrast is matched in content (i.e., identical scene images that have been rescaled), but different in retinotopic footprint (Fig.  4 b). This allows us to examine which cortical territory is more active under an immersive visual experience of a scene view, compared to postcard scene perception.

A whole-brain contrast map is shown in Fig.  4 b (Supplementary Fig.  5 for individual participants). This map shows that cortex near the POS is activated significantly more to full-field scenes than postcard scenes. This cortex showed far-peripheral visual field preference in Experiment 1, and effectively corresponds to the far-peripheral parts of early visual areas. Thus, it is likely that this cortex is not uniquely attributed to scene content presentation per se, but to any far-peripheral visual stimulation (which we explore further in the next experiments). Anatomically, this swath of cortex is largely adjacent to and mostly non-overlapping with classic scene regions, PPA and OPA, and anterior part of RSC. Thus, while it could have been that the full-field vs. postcard contrast would strongly encompass the scene-selective regions, this was not the case.

Effects of visual size and scene content

The whole-brain contrasts did not show clear evidence for a new scene region, or more extensively activated cortical surface area from the classic scene regions. Thus, we focused our quantitative analyses on these classic scene ROIs defined at the postcard visual size, and explored the extent to which each scene region is modulated by the visual size and scene content.

In addition to the scene ROIs, we defined a “Peripheral-POS” (parietal-occipital sulcus) region, using the retinotopy protocol data from Experiment 1. Specifically, we selected voxels that survived a conjunction contrast between pairs of the far-peripheral eccentricity ring condition and all other eccentricity ring conditions. Further, we removed the small proportion of voxels in the Peripheral-POS which spatially overlapped with independently defined RSC (mean = 5.6%, std = 7.7% of Peripheral-POS voxels).

The results of the ROI analyses are shown in Fig.  5 . Broadly, this 2 × 2 design reveals a powerful transition in the invariances of the responses, from cortex with retinotopic selectivities to scene content selectivities. Specifically, the Peripheral-POS region showed clear retinotopic modulation: there was a large effect of full-field vs. postcard sizes (F(1, 36) = 518.6, p  < 0.01, etaSq = 0.91), with only weak effect of image content (F(1, 36) = 11.7, p  < 0.01, etaSq = 0.02), and no interaction between these factors (F(1, 36) = 1.8, p  = 0.2). Put succinctly, this region shows clear retinotopic modulation, with little sensitivity to higher-order scene image content.

The anatomical locations of each ROI are illustrated on a schematic brain map in the middle (top: medial side, bottom: ventral surface of the right hemisphere). Each ROI panel shows the mean beta averaged across participants ( n  = 10) for each condition. Individual data are overlaid on top of the bars as dots. The main effect of visual size (blue vs. purple) and the main effect of content (dark vs. light) were significant in all ROIs. The significant interaction was found only in the PPA and RSC. The FFA result is in Supplement Fig.  6 . Post Postcard, PostScr Postcard Scrambled, FF full-field scenes, FFscr full-field scenes scrambled, PPA parahippocampal place area, RSC retrosplenial cortex, OPA occipital place area, FFA fusiform face area, POS parieto-occipital sulcus.

In contrast, both the PPA and the OPA showed the opposite pattern. That is, there were large effects of scene content vs. scrambled content (PPA: F(1, 36) = 535.2, p  < 0.01, etaSq = 0.86; OPA: F(1, 36) = 168.9, p  < 0.01, etaSq = 0.8), with only small effects of image size (PPA: F(1, 36) = 44.7, p  < 0.01, etaSq = 0.07; OPA: F(1, 36) = 5.1, p  < 0.05, etaSq = 0.02). There was a very small interaction of these factors in PPA, but not in OPA, with slightly higher activation in PPA for scenes in full-field presentation (PPA: F(1, 36) = 6.5, p  < 0.05, etaSq = 0.01; OPA: F(1, 36) = 0.6, n.s.). Thus, intact scenes drive much higher response than the phase-scrambled scenes in PPA and OPA, generally independently of the presentation size (darker vs. lighter color bars, Fig.  5) .

The RSC immediately abuts the Peripheral-POS region. Interestingly, it has a slightly more intermediate pattern, though it is more like the other high-level scene regions. That is, RSC showed a large effect of scene content (RSC: F(1, 32) = 141.1, p  < 0.01, etaSq = 0.52) and a moderate effect of visual size (RSC: F(1, 32) = 93.1, p  < 0.01, etaSq = 0.34), with only very weak interaction between them (RSC: F(1, 32) = 4.3, p  < 0.05, etaSq = 0.02). Taken together, these data reveal a clear pattern: classic scene regions have strong overall responses for image content, which is maintained over dramatically different visual sizes and a qualitatively different immersive experience, with relatively weaker modulation by the visual size of stimulus.

As a control, we also examined responses in the face-selective FFA (Supplementary Fig.  6) . While the overall responses to all four conditions were quite low, there was a small but statistically reliable main effect of visual size, with higher overall activation in full-field over postcard views (F(1, 36) = 8.9, p  < 0.01, etqSq = 0.19). The responses of this control region suggest that full-field stimulation might partly provide a more general boost to the visual system (e.g., via arousal). On this account, the scene regions’ slight preference for full-field stimulation might reflect a more general drive, further amplifying the dissociation between tuning for content and peripheral stimulation.

Thus, from the far-peripheral retinotopic cortex to the classic scene regions, there is a relatively abrupt transition in tuning along the cortical sheet. The far-peripheral retinotopic cortex shows only weak content differences. Adjacent scene-selective cortex amplifies these scene vs. scrambled content differences, regardless of whether or not the content stimulates the far periphery.

Far-peripheral stimulation without the central visual field

The previous experiment showed that scene regions are modulated dominantly by the image content, much less so by the visual size. However, postcard and full-field scenes both stimulate the central 45 degrees of the visual field. Thus, it is possible that the scene content preferences we observed are actually primarily due to central visual field stimulation. Are these scene content preferences also evident when only stimulating the far-periphery? In Experiment 3, we asked how far in eccentricity this scene preference is maintained.

We also asked the parallel question for face-selective regions. FFA is traditionally defined by contrasting responses to face vs. object image content presented in the center of the visual field. What happens when faces are presented in the far-periphery? Do face-selective regions also maintain their face content preferences when only presenting the content in the very far-peripheral visual field? Or, will any structured image content be represented increasingly more like a “scene” and drive scene regions, as it is presented farther from the center?

To directly test these questions, we generated a new stimulus set, depicting different content across the visual field, with increasing degrees of central “scotoma” 35 , that have matched retinotopic footprint to full-field scenes but differ in their content (Fig.  6 ). As in the previous experiment, we included both wide-angle rendered 3D scenes and their phase-scrambled counterparts. As a proxy for “full-field faces”, we made face arrays, in which multiple individual faces were presented throughout the full visual field. To avoid crowding effect and make each face recognizable (at basic category level), we adjusted the size of faces as a function of eccentricity (see Methods). Object arrays were generated in the same manner with individual small objects.

To stimulate only the peripheral visual field, we removed the central portion of the image by creating an artificial scotoma that systematically varied in size. There were five levels of scotomas including the no-scotoma condition (columns). We filled in the remaining space with four different kinds of image content: intact scenes, phase-scrambled scenes, object array, and face arrays (rows). For the object and face arrays, the size of individual items was adjusted to account for cortical magnification. *For copyright reasons, human faces have been substituted with illustrations in this manuscript, and objects were substituted with example images without copyright.

Then, we parametrically masked the central portion of images at 5 sizes (0, 30, 58, 88, and 138 degrees in diameter; see Fig.  6) . We measured brain responses to these 20 conditions, using a blocked design (see Methods). Participants were asked to perform a one-back repetition detection task while fixating their eyes at the center of screen. As before, we defined the classic scene ROIs using the same method (i.e., postcard-scene vs. postcard-face) from independent localizer runs.

We first examined responses of scene and face ROIs (Fig.  7 ). As expected, when there is no scotoma, all regions showed preferences for either scenes or faces relative to other categories. As the size of the central scotoma increases, leaving only increasingly peripheral stimulation, the results showed that content preferences across all ROIs were generally maintained. Through the penultimate scotoma condition (88 deg), all scene regions showed significantly higher activation for scenes compared to face arrays, object arrays, and phase-scrambled scenes (see Supplementary Tables for statistical test results).

In each panel, the line plot (error bands = standard error of the mean) shows how the response of each ROI changed as we increasingly removed the central visual field stimulation via scotoma, leaving only the peripheral stimulation. The call-out box with a bar plot (* p  < 0.05, two-sided paired t-test; see Supplement Tables for the full report of the statistical tests) shows responses for each image content at the largest scotoma condition (138 deg diameter). a , b Overall, PPA and RSC maintained their scene preference over faces across all scotoma conditions, whereas c the OPA maintained the preference until the penultimate condition. d The FFA also maintained its content preference for faces across all scotoma conditions. PPA parahippocampal place area, RSC retrosplenial cortex, OPA occipital place area, FFA fusiform face area.

The pattern at the farthest scotoma condition (138 deg) varied by the ROI and stimulus. RSC showed strong scene preference against all other image contents (Fig.  7 b, Supplementary Table.  2) . However, OPA’s scene preference did not hold at the 138 deg scotoma condition (Fig.  7 c, Supplementary Table.  3) . The PPA showed significantly higher activation for scenes compared to face arrays, but this activation level was not different from object arrays (t(9) = 2.2, n.s.; Figure  7 a; Supplementary Table.  1) . These results are also depicted on the cortical surface in Fig.  8 (Supplementary Fig.  7 for individual participants), showing the contrast of face vs. scene content, as the presentation is restricted increasingly peripherally. Overall, our results show that scene regions can be driven by content differences through a purely peripheral route, beyond at least 88 deg, that does not require central presentation.

This figure shows the whole-brain contrast between the scenes (red) and faces (blue), at each scotoma condition (columns). a Ventral view with PPA and FFA. b Medial view with RSC. c Lateral view with OPA. PPA parahippocampal place area, RSC retrosplenial cortex, OPA occipital place area, FFA fusiform face area.

Next we turned to FFA. If the presence of faces at the central visual field is necessary to drive FFA responses, then we would have expected the face preference to exist only in the no-scotoma or small scotoma conditions. However, that is not what we found. Instead, face-selective FFA shows the same pattern as the scene-selective regions. That is, FFA responded more to face content than other image content, across all scotoma levels, even at 138 degrees (see Supplementary Table.  4 for stats). This pattern of results is also evident in the cortical maps of Fig.  8 (Supplementary Fig.  7 for individual participants). Overall, these results clearly demonstrate that face-selectivity is present even when faces are presented in the very far periphery only. Thus, this result suggests that there is also a far-peripheral route to drive face-selective responses in the FFA, which does not require direct stimulation of the central visual field.

Finally, we wondered whether participants would actually be aware of the stimulus condition when it was presented in the far 138+ degrees of the visual field. To explore this, we conducted a brief categorization test during the anatomical scan. Either an object array or face array was presented with one of four scotoma sizes, and participants did a 2-alternative-forced-choice task. We found that participants were nearly perfect through the penultimate scotoma condition (30 deg: mean = 0.98, s.e = 0.02; 58 deg: mean = 0.96, s.e. = 0.03; 88 deg: mean = 0.99, s.e. = 0.01). The accuracy at the farthest eccentricity was more variable, but still statistically above chance (mean = 0.64, s.e. = 0.04; t(11) = 4.0, p  < 0.01). We note that only a limited number of trials were conducted due to time constraints, so these results should be interpreted with caution. However, the current results suggest that participants, on average, were weakly able to do the basic-level categorization, with only extreme peripheral visual information present.

Peripheral bias in scene regions

Lastly, in the classic scene regions, we found only minimally higher activation for full-field scenes relative to postcard scenes. Is this finding at odds with previously reported “peripheral bias”? Previous studies indicating a peripheral bias have shown increased activation in the PPA when the stimulated location moves from the central visual field to the periphery, up to 20 deg in diameter 30 , 36 . Two points are worth clarifying. First, our comparison between full-field scenes vs. postcard scenes is not actually a direct test of central vs. peripheral tuning, as both of these conditions stimulate the central visual field. Second, how much a region is activated depends on its receptive field (RF) size and location. So, for example, if a region’s RF completely encompasses the 44 deg diameter center of the visual field (i.e., postcard scene presentation size), that means this brain region’s RF would be stimulated in both postcard and full-field scenes, predicting not much activation difference.

We thus ran an exploratory analysis that examined each ROI’s response to the increasing eccentricity ring checkerboards used in Experiment 1. A peripheral bias account would intuitively predict that increasing peripheral stimulation would lead to a corresponding activation increase in each of these scene regions. However, that is not what we found. Instead, each scene ROI showed a different pattern in response to these eccentricity rings (Fig.  9 ).

a PPA response increases until the penultimate condition then drops at the extreme periphery. b RSC response was rather flat then jumped after the third ring, clearly showing its preference for the far-periphery. c OPA showed a mild peak around the third ring. d FFA showed the opposite pattern to a – c , demonstrating its preference for the central visual field. PPA parahippocampal place area, RSC retrosplenial cortex, OPA occipital place area, FFA fusiform face area.

The PPA had increasing activation with increasingly peripheral eccentricity rings (up to 37–100.6 deg diameter) but dropped at the farthest, most peripheral stimulation condition (>110 degrees). The OPA was similar to PPA, but with a nominal peak activation at the 3rd level of eccentricity (12.6–33 deg). Finally, RSC’s activation to central checkerboards was not significantly different from baseline for the first three levels, and then abruptly increased for both the two most extreme peripheral rings. Thus, neither PPA nor OPA showed strong sensitivity to ultra-peripheral generic stimulation (flashing checkerboard), showing a limit on the general peripheral bias hypothesis of scene regions.

Are these ROI responses across levels of eccentricity consistent with the visual size effects between full-field and postcard conditions? The size of the postcard scene (44 deg diameter) is most similar to the size of the inner circle at the fourth eccentricity ring (37 deg). So, in a rudimentary way, the stimulated visual field by both the last two eccentricity rings (>37 deg) roughly corresponds to the additionally stimulated visual field by the full-field scenes compared to the postcard scenes (>44 deg). Both PPA and OPA have stronger responses for the first three levels of eccentricity than the final two levels, and consistently showed little additional response to full-field scenes relative to postcard scenes. Meanwhile, RSC shows weaker responses for the first three levels of eccentricity, and more for the most peripheral conditions; and consistently, RSC showed stronger responses for full-field conditions regardless of content. Thus, the activation differences over eccentricity rings are indeed consistent with the visual size modulation effect of each scene region, observed in Experiment 2.

In sum, this post-hoc analysis is consistent with the previously known notion that peripheral bias—peripheral stimulation activates the scene regions more than the foveal stimulation. However, our results also place updated constraints on this account. The peripheral bias in the scene regions is present only up to a certain eccentricity, and this differs depending on each scene region. We offer that the thinking of a general peripheral bias is thus not appropriate, and the responsiveness over the visual field might be better understood in the context of RFs. Future work employing population RF mapping can be used to further clarify and chart the far-peripheral RF structure across these cortical regions.

In this study, we established a method to present ultra-wide-angle visual stimuli in the scanning environment. With this new tool, we were able to measure neural responses to the extreme far-periphery and chart the ultra-wide eccentricity map in the human brain beyond the scope of prior studies. We then examined the neural basis of full-field scene perception. We found that classic scene regions are tuned to scene content that is robust to changes in the visual size of scenes, suggesting a sharp tuning transition from adjacent far-peripheral retinotopic cortex to scene content regions. We also found scene and face-selective regions maintained their content preferences even in conditions of extreme peripheral stimulation, highlighting the existence of a far-peripheral route that has yet to be fully investigated. Finally, only RSC showed systematically higher responses at the farthest eccentricity, where both PPA and OPA had weaker responses, clarifying new limits on the peripheral bias of scene regions. Broadly, this work brings unique empirical evidence to clarify debates about the issues of content and peripheral preferences in scene representation and introduces an innovative method for investigating more naturalistic, immersive scene perception inside a scanner.

The full-field neuroimaging method allowed us to gain some new insights into the classic scene regions. First, we gained a better understanding of what it means to be a scene. While it has been well established that PPA, RSC, and OPA are scene-selective regions, the definition of a scene has been used in a somewhat mixed way. On one hand, a scene can be a set of visual patterns with particular kinds of higher-order image statistics. On the other hand, anything (including objects or faces) that falls in the far periphery can be part of a scene. This account is motivated by intuitions that part of what it means to be a scene is to have content that extends beyond the view. Leveraging the ultra-wide-angle image projection, our study directly compared these two accounts.

Overall results are clearly in favor of the first hypothesis. That is, not just any information in the far-periphery becomes a scene and is automatically integrated into the computations of scene regions. Even when faces or objects are at the far periphery, they do not drive the scene regions more than they would normally do at the central visual field. Instead, the classic scene regions are tuned to particular higher-order image statistics that are distinctive from visual features of other semantic categories, although there are some further differences among the scene regions 2 , 3 . This makes sense: many of the core visual features important for the scene regions are not much disrupted by the visual size or location change. For example, spatial layout 34 , 37 , statistics of contour junctions 38 , surface properties like material or texture 12 , 39 , or objects present in a scene 40 can be extracted similarly in both postcard and full-field scenes. However, it is also worth emphasizing that while these features do not have to be present in specific retinotopic locations, in real visual experience, useful visual cues for those (e.g., walls, planes, or boundaries) tend to be at the periphery rather than the center, providing an ecological explanation why the scene regions are developed to have sensitivity to visual information at the periphery.

Additionally, the access to the far-periphery provided a new perspective on the anatomical locations of the scene regions. We showed that three scene regions are very closely positioned to the far-peripheral cortex along the POS. When we perceive a full-field view, this medial brain region and the classic scene regions are activated together, forming a large ring-shaped portion of the cortex along the POS. In other words, the classic scene regions might be connected together by the far-periphery preferring cortex. This observation allows us to realize that the scene regions are actually proximal to each other anatomically. This intuition is not easily captured from the typical flattened brain map, because the cut is made along the fundus of calcarine sulcus 41 , splitting the retinotopic map into upper and lower visual quadrants, which in turn places PPA and OPA on opposite sides of this flat map (e.g., see ref. 42 for how object-selective cortex is organized between these PPA and OPA regions). Our schematic map of the medial surface (Fig.  10 ), in contrast, keeps the visual field maps intact, emphasizing the proximity between the scene regions and their relationship to the retinotopic map.

The scale and shape of retinotopic map is not accurately presented as the actual data. Instead, this flattened map of the medial view emphasizes the idea that the three scene regions might be connected via the far-peripheral cortex. VF visual field, PPA parahippocampal place area, RSC retrosplenial cortex, OPA occipital place area.

This view naturally lends an explanation why the PPA has upper visual field bias, OPA has lower visual field bias, and RSC does not show clear bias to either upper or lower visual field 31 . Further, this large-scale cortical organization may be related to recently proposed place-memory areas that are positioned immediately anterior to each of the scene-perception areas 43 . In particular, the organization is suggestive of a hierarchical representational motif, with systematic transformations of representation from retinotopic far-peripheral cortex to perceptual scene structure of the current view to more abstract scene memory.

Another interesting question is the relationship between RSC and area prostriata, which is located in the fundus of the calcarine sulcus, just anterior to the far-peripheral V1 29 . The prostriata has a distinctive representation from V1 over a wide visual field (up to 60 deg), and responds more to fast motion (570 deg per sec) than the moderate-speed motion (38 deg per sec) 29 . Moving dorsally along the POS, there is also human V6 that has the sensitivity to coherent field motion (e.g., spiral motion vs. random dots) or optic flow when compatible with self-motion 44 , 45 . While it requires further investigation whether the prostriata overlaps with the functionally defined RSC, it is possible that its sensitivity to peripheral motion might be used for representing dynamic scenes, to support spatial navigation.

The scene regions and even the fusiform face area both showed their content preference at the extreme far-periphery. How do these regions process stimuli at the far periphery?

Many studies have shown that face-selective regions respond more strongly to foveal stimulation, whereas scene-selective regions respond more strongly to peripheral stimulation 30 , 36 , 46 . Further, stronger functional connectivity was found between foveal V1 and face-selective regions (and between peripheral V1 and scene-selective regions), in human adults 47 , as well as in human and macaques neonates 48 , 49 . More recent study using diffusion MRI also showed higher proportion of white matter connection between foveal early visual cortex and ventral face regions (e.g., fusiform face area; FFA) 50 . Together, these results imply eccentricity-based preferential connection between early visual cortex and higher category-selective regions, which does not easily explain our findings.

One possibility is that there are meaningful connections across all eccentricities between the early visual cortex and the higher visual areas, even though some connections to a particular eccentricity are more weighted (e.g., FFA and foveal V1). Then, FFA might still show a somewhat weaker but preferential response to faces at the far periphery, as long as the stimuli are presented with appropriate visual size and arrangement to accommodate cortical magnification and crowding.

Another possibility is that attention temporarily adjusts RF properties of high-level visual areas. A study using the population receptive field (pRF) method showed that the pRFs of FFA were located more peripherally and larger during a face task (one-back judgment) than during a digit judgment task, resulting in extended coverage of the peripheral visual field 51 . While there was no control task condition in our experiments, the one-back repetition detection task could have helped incorporate far-peripheral stimuli into computations.

Additionally, there might be other input connections to the high-level visual areas outside the ventral pathway, perhaps via a subcortical route (e.g., superior colliculus) 52 , 53 or from the lateral surface. For example, the diffusion MRI study showed that lateral face regions (e.g., posterior STS-faces) have uniformly distributed connections across overall early visual cortex eccentricities, in contrast to the foveal-biased ventral FFA 50 . This suggests that the processing of faces is not limited to the central visual field, as they can also be processed at the periphery, especially in dynamic or social situations 54 , 55 . It is possible that the peripheral face selectivity observed in FFA may reflect responses from those lateral face areas. Further investigation is necessary to better understand these peripheral routes and how they support the transition from eccentricity-based to content tuning.

Lastly, another possibility to consider is that this effect is driven by non-compliant subjects who moved their eyes to the periphery. However, if participants always shifted their gaze towards the periphery, activation levels at the largest scotoma would match those in the no-scotoma condition, and if such eye movements happened only occassionally, it would likely result in greater variance in the far-periphery condition, which we did not observe. Further, moving your eyes beyond 70 degrees requires considerable effort and some discomfort. Thus we think this account of the responses is unlikely. While the setup here precludes the use of traditional eye-tracking equipment, emerging computational eye-tracking methods that extract eye gaze from the EPI images could prove a valuable complement to this method in future studies 56 .

Achieving a wide-angle (>15–20 deg) visual stimulation in an fMRI scanner has been a goal since the early days of fMRI 57 , 58 . For example, in the early 2000s, researchers were able to stimulate wider visual field up to 100 deg, mapping retinotopic area V6 in humans 22 . To achieve this, a relatively large flat screen (260 × 185 mm) was positioned inside the scanner bore, and the closer distance between the screen and the eyes allowed it to stimulate a larger portion of the visual field. However, this screen size was too large to be easily adaptable to other MRI scanners or conventional head coils. Another research group achieved 100 deg wide stimulation with a smaller screen (140 mm wide), but they used the combination of glasses and prism to enlarge the size of projected stimuli 23 , 59 .

In the next decades, the angle of image projection was pushed up to 120 deg wide 24 , 25 , 26 , 27 , 28 , 29 . These approaches leveraged monocular viewing–presenting the image to only one eye. In these setups, a small screen was positioned very close to the eyes (3 cm), and participants had to wear a contact lens to get help with focus and fixation at such a short distance. And, most recently, stimulation of the entire visual field was achieved 60 . Using custom-built goggles with white light-emitting diodes (LEDs), they were able to functionally localize the temporal monocular crescent, which requires wide-angle projection beyond 120 deg.

While much of this early work focused on retinotopy, our aim was to develop an approach that does not require participants to wear any devices and allows them to see stimuli as naturally as possible as they do outside the scanner. And, we focus here on exploring the perception and representation of high-level visual information presented extensively across the visual field. An advantage of our approach is that apparatus can be built at relatively low cost. We used a pair of mirrors to control the image projection trajectory, and the curved screen can be assembled with 3D-printed plastic parts. We share the design files and all specifications via a public website and community mailing list to support ultra-wide-angle neuroimaging ( https://jpark203.github.io/fullfield-neuroimaging ).

One of the current challenges of our ultra-wide-angle projection setup is that we are scanning without the top head coil because it blocks the peripheral view. While the data quality was still viable, there was a clear decrease of tSNR in all of the main ROIs (Supplementary Fig.  2) . The lack of top head coil could also limit the scope of research topics, especially if they involve investigating on the frontal lobe. Another main challenge is a limited image resolution (2–4 pixels/degree). Due to physical constraints of the scanner room, only ~30% of pixels from the projected image could be on the screen. This is because as the distance between the projector and the screen (inside the scanner bore) gets farther, the size of the projected image also gets larger. However, this limitation in spatial resolution can be overcome with our new projector that supports much higher resolution (up to 4k), compared to the old one (maximum 1024 × 786 pixels), increasing the projected resolution more than threefold (8–15 pixels/degree).

Regardless of these limitations, our full-field scanning method provides promising new research avenues that can be explored in future investigations. One such avenue is to explore how the brain represents the spatial scale of a view in a more ecologically valid manner. Traditionally, we study object-focused views by cropping a picture closely to an object, eliminating all contextual peripheral visual information. However, this picture editing approach does not reflect how we actually experience the world, as we continuously receive visual information from the periphery even when focusing on an object. By simply moving the camera position (as an agent moves in an environment) and maintaining the same wide field-of-view, the spatial scale of the view is naturally determined by the distance between the focused object and the camera (agent). This positions us to investigate how we obtain a sense of object-focused view in the real visual world. Moreover, this method allows us to re-examine previous studies on various aspects of spatial representation in the brain. We can revisit how the continuous dimension of space is represented from an object-focused view to a far-scale navigable scene view 61 , how intermediate-scale scenes (e.g., a view of a chopping board) are represented in the brain 62 , and how the memory of a view is biased depending on the depicted spatial scale 63 , 64 . Importantly, this can be done while isolating field-of-view manipulation (e.g., cropping) from viewing distance manipulation.

Another promising research direction is to investigate the role of peripheral vision in computing one’s body position (relative to objects or environments) in complex, dynamically moving situations. This task is crucial for activities ranging from maneuvering a vehicle and playing sports to everyday walking and navigation. For this, extracting relevant visual cues such as optic flow, and sensitivity to the peripheral visual field in particular would be important. Notably, brain regions involved in these processes, such as the peripheral POS, human V6, prostriata, and potentially OPA 17 , 22 , 29 , are spatially adjacent along the POS. Full-field scanning offers a unique opportunity to directly stimulate these regions. This approach can enhance our understanding of how these areas interact and contribute to ego-motion computation, with wide-reaching implications for applied vision research.

The present findings reveal that classic scene regions are modulated by structured image and scene content, over dramatic changes in visual size, suggesting that they are tuned to particular higher-order image statistics rather than to any peripheral stimulation. Broadly, this study demonstrates how full-field neuroimaging allows us to investigate visual perception under more realistic, immersive experiences.

Participants

Twenty-two participants were recruited from the Harvard University Public Study Pool (10 females aged 20–54 years). All participants completed Experiment 1 (retinotopy protocol), ten participants in Experiment 2, and twelve participants in Experiment 3. All participants had normal or corrected-to-normal vision, gave informed consent, and were financially compensated. The experiments were performed in accordance with relevant guidelines and regulations and all procedures were approved by the Harvard University Human Subjects Institutional Review Board.

To enable ultra-wide-angle projection during scanning, several modifications were made to the typical scanning setup. In order to achieve an unobstructed view for the participant, we did not attach the top head coil and scanned only with the bottom head coil. Instead, we placed a custom-built curved screen right above the participant’s head. The screen was built with 3D-printed plastic parts and acrylic solvent. The curved shape was maintained by gluing a polystyrene sheet (1/16 inch thick) to a custom-fabricated acrylic hull (Supplementary Fig.  3 b). The radius of the cylindrical screen was 11 inches. The screen was made as large as possible while remaining rigidly supported and still fitting inside the MRI bore (about 12-inch radius). The one-inch difference allowed for the supporting ribs of the hull and a bit of clearance when moving in and out of the bore. Adjustable “legs” were attached at the bottom of the screen with nylon screws, and these legs were slotted into the scanner bed, allowing the screen to be securely anchored. Design files of the screen can be downloaded at https://jpark203.github.io/fullfield-neuroimaging/screen .

We also removed the standard flat projection screen at the back of the scanner bore. We bounced the projected image off of a pair of angled mirrors installed near the projector, directly into this curved screen inside the bore (Supplementary Fig.  3 a, Fig.  1) . For this, we constructed an inverted periscope. A pair of front surface mirrors were supported on a non-ferromagnetic stand. The lower mirror remains fixed, and the upper mirror is hinged. Tilting the upper mirror up removes the periscope from the projection path. With the periscope in place, the projector appears to originate from a virtual point further back and below the floor of the room.

Since this step changed how far the image on the screen is cast from the projector, we also adjusted the focus setting of the projector. Next, we used a reference image that was warped to fit the screen to check whether the image was accurately projected on the curved screen. If necessary, we carefully adjusted the projector position and/or the mirror angle. After this initial calibration stage, we refined the screen setup after a participant was put inside the scanner. First, we asked the participant to adjust their head position such that they were looking directly toward the center fixation mark on the screen. Second, we further adjusted the focus setup of the projector based on individual participants’ feedback. Overall, we allocated ~15–20 minutes of additional time for setting up the full-field scanning.

Image projection

To increase the spatial extent of stimulus, our goal was to project an image onto the inner wall of the cylinder bore. Ideally, the projector would be incident on the screen at 90 physical degrees. The physical geometry of the scanner bore makes this nearly impossible. The geometry of the room and our projection path are schematized in Supplementary Fig.  3 . The next best solution would be to place the projector (or the final mirror) closer to the cylinder bore in order to obtain the steepest angle possible. We did not pursue this route because any alterations must be minimally intrusive to alter any other ongoing study, as the MRI serves many labs.

If we projected directly onto the MRI bore, the light rays would be incident at just over 18 physical degrees. This shallow angle results in large distortion along the vertical (Y) axis of the projected image. To somewhat mitigate this, we angled the projection screen. Rather than being parallel to the magnet bore, we tilted it by 10 physical degrees. The edge of the screen at the top of the subject’s head nearly touches the bore. The screen edge near their collarbone is closer to the subject than the bore. Tilting angles larger than 10 physical degrees were ruled out for reasons of comfort–eye strain, feelings of confinement, etc. Effectively, this leads to the projector being angled slightly over 28 physical degrees relative to the screen (i.e., combining the tilted angle of the mirror and the screen).

As a result, approximately 1/3 of the Y pixels of the projector fall onto the screen, limiting our vertical resolution to 284 pixels rather than the native 768. In the case of the x pixels, about 828 pixels fall onto the screen, out of the native 1024 pixels (Fig.  2 , Supplementary Fig.  3 a). Pixels that did not intercept the display screen were set to black.

The visual angle of the display screen ranges from 168–182 degrees in width and 106–117 degrees in height (Supplementary Fig.  3 c). This variation depends on the distance between the participant’s eyes and the screen, which is affected by head size and head cushion options, for distances between 13.5–16.5 cm. For the stimulus size to be reported in the manuscript, we picked the middle viewing distance (15 cm) and calculated a stimulus angular extent. Perspective estimates did not take into account subject variability or binocularity.

The resolution of our current screen was 4.6–4.9 pixels per degree in width and 2.4–2.7 pixels/degree in height. It is noteworthy that the current limits on the low resolution can be overcome by our new projector, which has a much higher maximum resolution (4k). For example, if we keep the same aspect ratio of 4:3 (3200 × 2400), the pixels/degree will increase by the scaling factor of 3.125 (i.e., 2400/768 = 3.125).

Computational image warping

Because of the curvature and angle of the screen, all projected images were first computationally warped using a custom function to compensate for the geometry of the curved screen. Specifically, we developed a computational method that transforms a regular, rectangular image (1024 × 768 pixels; 4:3 aspect ratio) into a curved shape that matches the size and curvature of our custom-built screen. The transformed image on the cylindrical display surface preserved the same original aspect ratio (4:3) as it is measured 58.5 cm (arc length) × 44 cm (linear). Our image-warping algorithm allowed us to project the images onto the cylindrical screen without stretch or distortion; similar to the real-world action of pasting a sheet of wallpaper onto a cylindrical wall.

To link the warping algorithm parameters to the physical set up, we developed a calibration procedure, in which we use an MR-compatible mouse to obtain the × and y coordinates of the projector image that correspond with the three points along the screen outline (e.g., measuring points along both the top and bottom of screen curvature separately, as the bottom screen was slightly narrower than the top). This resulted in a 2d mapping, which takes an original image, and then resizes and warps it to be positioned directly into the part of the projected image that is being projected onto the screen (Fig.  2) .

Signal quality check

Several quality assurance tests were conducted with and without the top head coil separately, to check how much fMRI signal was impacted by removing the top head coil. First, we ran the CoilQA sequence that calculates and provides an Image SNR map. Second, we ran one of our BOLD protocols (i.e., one of the functional runs), computed tSNR maps, and examined BOLD quality check results. Third, we also ran the T1-weighted scan for a qualitative comparison between the two cases. The test results are reported in Supplementary Fig.  1 .

Additionally, we also computed tSNR within each ROI. For this, we preprocessed the data using the identical protocol as the main experiments and normalized it into Talairach space. The voxel-wise tSNR was calculated by dividing the mean by the standard deviation of time-course data. Then, we extracted voxels for each ROI, and averaged their tSNRs to get an ROI tSNR value. The comparison between with and without the top head coil is reported in Supplementary Fig.  2 .

Rendering full-field views from virtual 3D environments

Computer-generated (CGI) environments were generated using the Unity video game engine (Unity Technologies, Version 2017.3.0). We constructed twenty indoor environments, reflecting a variety of semantic categories (e.g., kitchens, bedrooms, laboratories, cafeterias, etc.). All rooms had the same physical dimensions (4 width × 3 height × 6 depth arbitrary units in Unity), with an extended horizontal surface along the back wall, containing a centrally positioned object. Each environment was additionally populated with the kinds of objects typically encountered in those locations, creating naturalistic CGI environments. These environments were also used in refs. 61 , 63 .

Next, for each environment, we rendered an image view, set to mimic the view of an adult standing in a room looking at the object on the back counter/surface. During the development of these protocols, we found that it was important to get the camera parameters related to the field of view (FOV) right to feel as if you were standing in the room with objects having their familiar sizes; otherwise, viewers were prone to experience distortions of space. Here the camera FOV was fixed at 105 degrees in height and 120.2 degrees in width. This FOV was chosen based on the chord angle of our physical screen (120 deg) and empirical testing by experimenters. Since there was no ground truth for the size of virtual reality environments (e.g., how large the space should be), experimenters compared a few different FOVs and made subjective judgments on which parameter feels most natural. Relatedly, we set the camera height to be 1.6 (arbitrary units), and tilted the camera angle down (mean rotation angle = 5.2 deg, s.d. = 0.5 deg, across 20 environments), so that the center object was always at the center of the image. For these stimuli, we positioned the camera at the back of the environment, to give a view of the entire room. Each image was rendered at 1024 × 768 pixels.

Experiment 1

In the retinotopy runs (5.8 min, 174 TRs), there were 7 conditions: horizontal bands, vertical bands, and five levels of eccentricities (e.g., from foveal stimulation to far-peripheral stimulation). A center circle was 1.8 degrees radius, and the inner and outer rings of the rest of the conditions were 2.0–5.6 degrees, 6.3–16.5 degrees, 18.5–50.3 degrees, and >55.3 degrees radius. All stimuli cycled between states of black-and-white, white-and-black, and randomly colored, at 4Hz. Each run consisted of 7 blocks per condition (6-sec block), with seven 6-sec fixation blocks interleaved throughout the experiment. An additional 6-sec fixation block was added at the beginning and the end of the run. Participants were asked to maintain fixation and press a button when the fixation dot turned blue, which happened at a random time once per block.

Experiment 2

In Experiment 2, participants completed 8 runs of the main protocol (one participant completed 6, and two participants completed 5 runs) and 3 retinotopy runs (two participants completed 2 runs).

In the main protocol, there were 7 stimulus conditions. (1) Full-field scenes: 15 full-field scene images were chosen (randomly selected from the 20 total environments). (2) Full-field Phase-scrambled image. First, the images were fast Fourier transformed (FFT) to decompose them into amplitude and phase spectrum. Then, the phase spectrum was randomized by adding random values to the original phase spectrum. The resulting phase spectrum was combined with the amplitude spectrum, then transformed back to an image using an inverse Fourier transform 65 . (3) Postcard scenes. These images were generated by rescaling the full-field scenes. Instead of cropping the central portion of the original image, an entire image was rescaled from 1024 × 786 pixels to 205 × 154 pixels (44 degrees wide). This rescaled image was positioned at the center, and the rest of area around it was filled with the background color, such that the size of whole image (i.e., small scene at the center with the padding around it) was kept the same as the original image (1024 × 768 pixels). (4) Postcard-scrambled scenes. The same rescaling procedure was followed for the phase-scrambled scenes. The final three conditions consisted of fifteen images from each category of (5) faces, (6) big animate objects, and (7) small inanimate objects. They were rescaled to fit a bounding box (171 × 129 pixels; 37 degrees wide) with white background color. This bounding box was positioned at the center with the padding, so that the size of an output image is 1024 × 768 pixels.

A single run of the main protocol was 6.5 min in duration (195 TRs) and was a classic on-off blocked design. A condition block was 12 sec, and was always followed by 6 sec fixation period. Within each block, six trials from one condition were presented. Each trial consisted of 1.5 sec stimulus presentation and 500 ms blank screen. The stimulus duration was chosen to be a little longer than the typical scanning, because flashing full-field images too fast can be uncomfortable and may cause nausea. Among those six images, five were unique images, and one of those images was randomly chosen and repeated twice in a row. Participants were instructed to press a button when they saw the repeated image (one-back repetition detection task). The presentation order of blocks was pseudo-randomized for each run as follows. Seven conditions within an epoch were randomized 3 times independently and concatenated with a constraint that the same condition cannot appear in two successive blocks. Thus, each of 7 condition blocks were presented 3 times per run. Fifteen unique images per condition were randomly split across those three blocks, for each run.

Experiment 3

In Experiment 3, participants completed 8 runs of the main protocol (one participant completed 7, and five participants completed 6 runs), 2 runs of classic category localizer (two participants completed 1 run, and two participants did not complete any localizers and were excluded from ROI analyses), and 2 retinotopy runs (two participants completed 3 runs).

In the main protocol of Experiment 3, stimuli were varied with 2 factors: image content (scenes, phase-scrambled scenes, face arrays, object arrays), and scotoma size (0, 29, 58, 88, 140 degrees in diameter). The scene images were captured from 20 virtual environments built in Unity, using the same camera parameters as in Experiment 2. For face and object conditions, 58 individual faces and objects were collected. We matched the luminance across scenes, faces, and objects, by equating the luminance histograms using Color SHINE toolbox 66 . The phase-scrambled scenes were generated from the luminance-matched scenes, using the same parameters as in Experiment 1.

Face and object arrays were generated with those luminance-matched images. For each face array, 13 faces were randomly drawn from a pool of 58 faces (half male). These faces were arranged along 3 levels of eccentricity circles. The size of individual faces and the number of faces was adjusted for each eccentricity, to account for cortical magnification and to avoid crowding effect. At the smallest eccentricity, 3 faces were rescaled to the size of 113-pixel diameter; at the middle eccentricity, 6 faces were rescaled to the size of 178-pixel diameter; at the largest eccentricity, 4 faces were rescaled to the size of 295-pixel diameter. The largest faces were positioned at 4 corners of the image, and the rest of faces were equally distanced along the eccentricity circle, with random jitters applied to individual face locations. Object arrays were generated using the same procedure. This step resulted in 20 face arrays and 20 object arrays. After making those base stimuli with 4 different image content (scenes, phase-scrambled scenes, face arrays, object arrays), we generated scotoma conditions by applying scotoma masks with 5 levels: 0 (i.e., no mask), 29, 58, 88, and 140 degrees in diameter. In total, 400 unique stimuli were generated across 20 conditions.

The main protocol was 6.9 min in duration (208 TRs), and used a block design, with 20 conditions presented twice per run. In each condition block (8 sec), five trials from one condition were presented. Each trial consisted of 1.1 sec stimulus presentation, followed by 500 ms blank screen. A fixation (black and white bullseye) was presented at the center of screen throughout an entire block. Among those five images in a block, four were unique images, and one of those images was randomly chosen and repeated twice in a row. Participants were asked to press a button when they detected the repetition. The presentation order of blocks in each run was randomized within each epoch. One epoch consisted of one block from each of 20 conditions and 5 resting blocks (8 sec). For each epoch, 20 unique images per condition were randomly split across 5 scotoma conditions. This procedure was repeated twice and concatenated with a constraint that the same condition cannot appear in two successive blocks. Thus, each of 20 condition blocks were repeated twice per run.

The classic category localizer was 6.9 min (208 TRs) and consisted of four conditions: scenes, faces, objects, and scrambled objects. Ten blocks per condition were acquired within a run. In each condition block (8 sec), four unique images were selected, and one of those images was randomly chosen and repeated twice in a row. Participants performed the one-back repetition task. Each image was presented for 1.1 sec and followed by 500 ms blank. In each run, the block order was randomized within each epoch, which consisted of one block from each condition and one fixation block (8 sec). This procedure was repeated ten times, and the block orders were concatenated across the epochs.

Additionally, the same retinotopy protocol from Experiment 2 was run. All stimuli presentation and the experiment program were produced and controlled by MATLAB R2020b and Psychophysics Toolbox (3.0.17) 67 , 68 .

Behavioral recognition task

To test whether participants can recognize a basic category of stimuli, a 2-alternative-forced choice (2AFC) was performed inside the scanner during an MPRAGE protocol. Only the face arrays and object arrays with scotomas were tested. Each array was presented for 1.1 sec, which was the same duration used in the main protocol. Then, participants were asked to indicate whether the stimulus was faces or objects, using a response button box.

fMRI data acquisition

All neuroimaging data were collected at the Harvard Center for Brain Sciences using the bottom half (20 channels) of a 32-channel phased-array head coil with a 3T Siemens Prisma fMRI Scanner. High-resolution T1-weighted anatomical scans were acquired using a 3D multi-echo MPRAGE protocol 69 (176 sagittal slices; FOV = 256 mm; 1 × 1 × 1 mm voxel resolution; gap thickness = 0 mm; TR = 2530 ms; TE = 1.69, 3.55, 5.41, and 7.27 ms; flip angle = 7°). Blood oxygenation level-dependent (BOLD) contrast functional scans were obtained using a gradient echo-planar T2* sequence (87 oblique axial slices acquired at a 25° angle off of the anterior commissure-posterior commissure line; FOV = 211 mm; 1.7 × 1.7 × 1.7 mm voxel resolution; gap thickness = 0 mm; TR = 2000 ms; TE = 30 ms, flip angle = 80°, multiband acceleration factor = 3, in-plane acceleration factor = 2) 70 , 71 , 72 , 73 .

fMRI data analysis and preprocessing

The fMRI data were analyzed with BrainVoyager 21.2.0 software (Brain Innovation) with custom Matlab scripting. Preprocessing included slice-time correction, linear trend removal, 3D motion correction, temporal high-pass filtering, and spatial smoothing (4mm FWHM kernel). The data were first aligned to the AC-PC axis, then transformed into the standardized Talairach space (TAL). Three-dimensional models of each participant’s cortical surface were generated from the high-resolution T1-weighted anatomical scan using the default segmentation procedures in FreeSurfer. For visualizing activations on inflated brains, the segmented surfaces were imported back into BrainVoyager and inflated using the BrainVoyager surface module. Gray matter masks were defined in the volume based on the Freesurfer cortex segmentation.

A general linear model (GLM) was fit for each participant using BrainVoyager. The design matrix included regressors for each condition block and 6 motion parameters as nuisance regressors. The condition regressors were constructed based on boxcar functions for each condition, convolved with a canonical hemodynamic response function (HRF), and were used to fit voxel-wise time course data with percent signal change normalization and correction for serial correlations. The beta weights from the GLM were used as measures of activation to each condition for all subsequent analyses.

Regions of interest (ROIs)

Experiment 2 did not have separate localizer runs. So, we split the main runs into two sets and used the half of runs to localize ROIs and the other half to extract data for subsequent analyses. We defined ROIs separately in each hemisphere in each participant, using condition contrasts implemented in subject-specific GLMs. Three scene-selective areas were defined using [Postcard Scenes–Faces] contrast ( p  < 0.0001). Specifically, the PPA was defined by locating the cluster between posterior parahippocampal gyrus and lingual gyrus, the RSC was defined by locating the cluster near the posterior cingulate cortex, and the OPA was defined by locating the cluster near transverse occipital sulcus. The FFA was defined using [Faces–Postcard Scene] contrast ( p  < 0.0001). The early visual areas (EVA; V1–V3) were defined manually on inflated brain, based on the contrast of [Horizontal–Vertical] meridians from the retinotopy runs.

In Experiment 3, independent localizer runs were used to define ROIs. We defined the PPA, RSC, and OPA using [Scenes–Faces] contrast ( p  < 0.0001). The FFA was defined using [Faces–Scenes] contrast ( p  < 0.001). The lateral occipital complex (LOC) was defined using [Objects–Scrambled Objects] contrast ( p  < 0.0001). Finally, the early visual areas (EVA; V1–V3) were defined manually on the inflated brain based on the contrast of [Horizontal–Vertical] meridians from the retinotopy runs. All ROIs were defined separately in each hemisphere of each participant.

Eccentricity preference map

To examine a topographic mapping of the eccentricity map, we calculated a group-level preference map. First, responses to each of 5 levels of eccentricities were extracted in each voxel from single-subject GLMs and then averaged over subjects. For each voxel, a condition showing the highest group-average response was identified as the preferred condition. The degree of preference was computed by taking the response differences between the most preferred condition and the next most preferred condition. For visualization, we colored each voxel with a color hue corresponding to the preferred condition, with a color intensity reflecting the degree of preference. The resulting preference map was projected onto the cortical surface of a sample participant. The same preference mapping procedures were used to generate individual subject preference mapping as well.

Reporting summary

Further information on research design is available in the  Nature Portfolio Reporting Summary linked to this article.

Data availability

Preprocessed fMRI data, design matrices, ROI masks, and all stimuli before the image transformation are available in the Open Science Framework repository ( https://osf.io/5hsbv ).  Source data are provided with this paper.

Code availability

Image transformation scripts are available on Github ( https://github.com/jpark203/FullField-ImageWarping ; https://doi.org/10.5281/zenodo.11113136 ) and on the method website ( https://jpark203.github.io/fullfield-neuroimaging ).

Strasburger, H. Seven myths on crowding and peripheral vision. i-Percept. 11 , 2041669520913052 (2020).

Google Scholar  

Epstein, R. A. & Baker, C. I. Scene perception in the human brain. Annu. Rev. Vis. Sci. 5 , 373–397 (2019).

Article   PubMed   PubMed Central   Google Scholar  

Dilks, D. D., Kamps, F. S. & Persichetti, A. S. Three cortical scene systems and their development. Trends. Cogn. Sci. 26 , 117–127 (2022).

Epstein, R. & Kanwisher, N. A cortical representation of the local visual environment. Nature 392 , 598–601 (1998).

Article   ADS   CAS   PubMed   Google Scholar  

O’Craven, K. M. & Kanwisher, N. Mental imagery of faces and places activates corresponding stimulus-specific brain regions. J. Cogn. Neurosci. 12 , 1013–1023 (2000).

Article   PubMed   Google Scholar  

Silson, E. H., Steel, A. D. & Baker, C. I. Scene-selectivity and retinotopy in medial parietal cortex. Front. Hum. Neurosci. 10 , 412 (2016).

Grill-Spector, K. The neural basis of object perception. Curr. Opin. Neurobiol. 13 , 159–166 (2003).

Article   CAS   PubMed   Google Scholar  

Dilks, D. D., Julian, J. B., Paunov, A. M. & Kanwisher, N. The occipital place area is causally and selectively involved in scene perception. J. Neurosci. 33 , 1331–1336 (2013).

Walther, D. B., Caddigan, E., Fei-Fei, L. & Beck, D. M. Natural scene categories revealed in distributed patterns of activity in the human brain. J. Neurosci. 29 , 10573–10581 (2009).

Article   CAS   PubMed   PubMed Central   Google Scholar  

Epstein, R. A. & Morgan, L. K. Neural responses to visual scenes reveals inconsistencies between fmri adaptation and multivoxel pattern analysis. Neuropsychologia 50 , 530–543 (2012).

Kornblith, S., Cheng, X., Ohayon, S. & Tsao, D. Y. A network for scene processing in the macaque temporal lobe. Neuron 79 , 766–781 (2013).

Park, J. & Park, S. Conjoint representation of texture ensemble and location in the parahippocampal place area. J. Neurophysiol. 117 , 1595–1607 (2017).

Oliva, A. & Torralba, A. Modeling the shape of the scene: a holistic representation of the spatial envelope. Int. J. Comput. Vis. 42 , 145–175 (2001).

Article   Google Scholar  

Bonner, M. F. & Epstein, R. A. Coding of navigational affordances in the human visual system. Proc. Natl. Acad. Sci. 114 , 4793–4798 (2017).

Article   ADS   CAS   PubMed   PubMed Central   Google Scholar  

Marchette, S. A., Vass, L. K., Ryan, J. & Epstein, R. A. Anchoring the neural compass: coding of local spatial reference frames in human medial parietal lobe. Nat. Neurosci. 17 , 1598–1606 (2014).

Persichetti, A. S. & Dilks, D. D. Dissociable neural systems for recognizing places and navigating through them. J. Neurosci. 38 , 10295–10304 (2018).

Kamps, F. S., Lall, V. & Dilks, D. D. The occipital place area represents first-person perspective motion information through scenes. Cortex 83 , 17–26 (2016).

Robertson, C. E., Hermann, K. L., Mynick, A., Kravitz, D. J. & Kanwisher, N. Neural representations integrate the current field of view with the remembered 360 panorama in scene-selective cortex. Curr. Biol. 26 , 2463–2468 (2016).

Ferrara, K. & Park, S. Neural representation of scene boundaries. Neuropsychologia 89 , 180–190 (2016).

Park, J. & Park, S. Coding of navigational distance and functional constraint of boundaries in the human scene-selective cortex. J. Neurosci. 40 , 3621–3630 (2020).

Ellis, C. et al. Re-imagining fmri for awake behaving infants. Nat. Commun. 11 , 4523 (2020).

Pitzalis, S. et al. Wide-field retinotopy defines human cortical visual area v6. J. Neurosci. 26 , 7962–7973 (2006).

Stenbacka, L. & Vanni, S. fmri of peripheral visual field representation. Clin. Neurophysiol. 118 , 1303–1314 (2007).

Yan, T., Jin, F., He, J. & Wu, J. Development of a wide-view visual presentation system for visual retinotopic mapping during functional mri. J. Magn. Reson. Imaging 33 , 441–447 (2011).

Wu, J., Yan, T., Zhang, Z., Jin, F. & Guo, Q. Retinotopic mapping of the peripheral visual field to human visual cortex by functional magnetic resonance imaging. Hum. Brain Mapp. 33 , 1727–1740 (2012).

Arnoldussen, D. M., Goossens, J. & van den Berg, A. V. Adjacent visual representations of self-motion in different reference frames. Proc. Natl Acad. Sci. 108 , 11668–11673 (2011).

Wu, J. et al. Development of a method to present wide-view visual stimuli in mri for peripheral visual studies. J. Neurosci. Methods 214 , 126–136 (2013).

Greco, V. et al. A low-cost and versatile system for projecting wide-field visual stimuli within fmri scanners. Behav. Res. Methods 48 , 614–620 (2016).

Mikellidou, K. et al. Area prostriata in the human brain. Curr. Biol. 27 , 3056–3060 (2017).

Levy, I., Hasson, U., Avidan, G., Hendler, T. & Malach, R. Center–periphery organization of human object areas. Nat. Neurosci. 4 , 533–539 (2001).

Silson, E. H., Chan, A. W.-Y., Reynolds, R. C., Kravitz, D. J. & Baker, C. I. A retinotopic basis for the division of high-level scene processing between lateral and ventral human occipitotemporal cortex. J. Neurosci. 35 , 11921–11935 (2015).

Baldassano, C., Esteva, A., Fei-Fei, L. & Beck, D. M. Two distinct scene-processing networks connecting vision and memory. Eneuro 3 , 5 (2016).

McCotter, M., Gosselin, F., Sowden, P. & Schyns, P. The use of visual information in natural scenes. Vis. Cogn. 12 , 938–953 (2005).

Park, S., Brady, T. F., Greene, M. R. & Oliva, A. Disentangling scene content from spatial boundary: complementary roles for the parahippocampal place area and lateral occipital complex in representing real-world scenes. J. Neurosci. 31 , 1333–1340 (2011).

Loschky, L. et al. The contributions of central and peripheral vision to scene gist recognition with a 180° visual field. J. Vis. 15 , 570–570 (2015).

Hasson, U., Levy, I., Behrmann, M., Hendler, T. & Malach, R. Eccentricity bias as an organizing principle for human high-order object areas. Neuron 34 , 479–490 (2002).

Harel, A., Kravitz, D. J. & Baker, C. I. Deconstructing visual scenes in cortex: gradients of object and spatial layout information. Cereb. Cortex 23 , 947–957 (2013).

Choo, H. & Walther, D. B. Contour junctions underlie neural representations of scene categories in high-level human visual cortex. Neuroimage 135 , 32–44 (2016).

Cant, J. S. & Xu, Y. Object ensemble processing in human anterior-medial ventral visual cortex. J. Neurosci. 32 , 7685–7700 (2012).

Stansbury, D. E., Naselaris, T. & Gallant, J. L. Natural scene statistics account for the representation of scene categories in human visual cortex. Neuron 79 , 1025–1034 (2013).

Van Essen, D. & Drury, H. Structural and functional analyses of human cerebral cortex using a surface-based atlas. J. Neurosci. 17 , 7079–7102 (1997).

Hasson, U., Harel, M., Levy, I. & Malach, R. Large-scale mirror-symmetry organization of human occipito-temporal object areas. Neuron 37 , 1027–1041 (2003).

Steel, A., Billings, M. M., Silson, E. H. & Robertson, C. E. A network linking scene perception and spatial memory systems in posterior cerebral cortex. Nat. Commun. 12 , 2632 (2021).

Pitzalis, S. et al. Human v6: the medial motion area. Cereb. Cortex 20 , 411–424 (2010).

Cardin, V. & Smith, A. T. Sensitivity of human visual and vestibular cortical regions to egomotion-compatible visual stimulation. Cereb. Cortex 20 , 1964–1973 (2010).

Malach, R., Levy, I. & Hasson, U. The topography of high-order human object areas. Trends Cogn. Sci. 6 , 176–184 (2002).

Baldassano, C., Fei-Fei, L. & Beck, D. M. Pinpointing the peripheral bias in neural scene-processing networks during natural viewing. J. Vis. 16 , 9–9 (2016).

Kamps, F. S., Hendrix, C. L., Brennan, P. A. & Dilks, D. D. Connectivity at the origins of domain specificity in the cortical face and place networks. Proc. Natl Acad. Sci. 117 , 6163–6169 (2020).

Arcaro, M. J. & Livingstone, M. S. A hierarchical, retinotopic proto-organization of the primate visual system at birth. Elife 6 , e26196 (2017).

Finzi, D. Differential spatial computations in ventral and lateral face-selective regions are scaffolded by structural connections. Nat. Commun. 12 , 2278 (2021).

Kay, K. N., Weiner, K. S. & Grill-Spector, K. Attention reduces spatial uncertainty in human ventral temporal cortex. Curr. Biol. 25 , 595–600 (2015).

Rima, S. & Schmid, M. C. V1-bypassing thalamo-cortical visual circuits in blindsight and developmental dyslexia. Curr. Opin. Physiol. 16 , 14–20 (2020).

Beltramo, R. & Scanziani, M. A collicular visual cortex: neocortical space for an ancient midbrain visual structure. Science 363 , 64–69 (2019).

Pitcher, D., Dilks, D. D., Saxe, R. R., Triantafyllou, C. & Kanwisher, N. Differential selectivity for dynamic versus static information in face-selective cortical regions. Neuroimage 56 , 2356–2363 (2011).

Pitcher, D. & Ungerleider, L. G. Evidence for a third visual pathway specialized for social perception. Trends Cogn. Sci. 25 , 100–110 (2021).

Frey, M., Nau, M. & Doeller, C. F. Magnetic resonance-based eye tracking using deep neural networks. Nat. Neurosci. 24 , 1772–1779 (2021).

Tootell, R. B. et al. Functional analysis of human mt and related visual cortical areas using magnetic resonance imaging. J. Neurosci. 15 , 3215–3230 (1995).

Cheng, K., Fujita, H., Kanno, I., Miura, S. & Tanaka, K. Human cortical regions activated by wide-field visual motion: an h2 (15) o pet study. J. Neurophysiol. 74 , 413–427 (1995).

Stenbacka, L. & Vanni, S. Central luminance flicker can activate peripheral retinotopic representation. Neuroimage 34 , 342–348 (2007).

Nasr, S. et al. In vivo functional localization of the temporal monocular crescent representation in human primary visual cortex. Neuroimage 209 , 116516 (2020).

Park, J., Josephs, E. & Konkle, T. Ramp-shaped neural tuning supports graded population-level representation of the object-to-scene continuum. Sci Rep. 12 , 18081 (2022).

Josephs, E. L. & Konkle, T. Large-scale dissociations between views of objects, scenes, and reachable-scale environments in visual cortex. Proc. Natl Acad. Sci. 117 , 29354–29362 (2020).

Park, J., Josephs, E. & Konkle, T. Systematic transition from boundary extension to contraction along an object-to-scene continuum. J. Vis. 24 , 9–9 (2024).

Bainbridge, W. A. & Baker, C. I. Boundaries extend and contract in scene memory depending on image properties. Curr. Biol. 30 , 537–543 (2020).

Ragni, F., Tucciarelli, R., Andersson, P. & Lingnau, A. Decoding stimulus identity in occipital, parietal and inferotemporal cortices during visual mental imagery. Cortex 127 , 371–387 (2020).

Dal Ben, R. Shine_color: controlling low-level properties of colorful images. MethodsX . 11 , 102377 (2021).

Brainard, D. H. The psychophysics toolbox. Spat. Vis. 10 , 433–436 (1997).

Pelli, D. G. & Vision, S. The videotoolbox software for visual psychophysics: transforming numbers into movies. Spat. Vis. 10 , 437–442 (1997).

Van der Kouwe, A. J., Benner, T., Salat, D. H. & Fischl, B. Brain morphometry with multiecho mprage. Neuroimage 40 , 559–569 (2008).

Moeller, S. et al. Multiband multislice ge-epi at 7 tesla, with 16-fold acceleration using partial parallel imaging with application to high spatial and temporal whole-brain fmri. Magn. Reson. Med. 63 , 1144–1153 (2010).

Feinberg, D. A. et al. Multiplexed echo planar imaging for sub-second whole brain fmri and fast diffusion imaging. PloS One 5 , e15710 (2010).

Setsompop, K. & Gagoski, B. A. et al. Blipped-controlled aliasing in parallel imaging for simultaneous multislice echo planar imaging with reduced g-factor penalty. Magn. Reson. Med. 67 , 1210–1224 (2012).

Xu, J. et al. Evaluation of slice accelerations using multiband echo planar imaging at 3 t. Neuroimage 83 , 991–1001 (2013).

Download references

Acknowledgements

Research reported in this study was supported by the Harvard Brain Science Initiative Postdoc Pioneers Grant awarded to J.P. and the National Eye Institute of the National Institutes of Health under Award Number R21EY031867 awarded to T.K. The content is solely the responsibility of the authors and does not necessarily represent the official views of the National Institutes of Health. This research was carried out at the Harvard Center for Brain Science and involved the use of instrumentation supported by the NIH Shared Instrumentation Grant Program (S10OD020039). We acknowledge the University of Minnesota Center for Magnetic Resonance Research for the use of the multiband-EPI pulse sequences. We also thank MCB Graphics at Harvard University for their assistance with the graphics in figures.

Author information

Authors and affiliations.

Department of Psychology, Harvard University, Cambridge, MA, USA

Jeongho Park & Talia Konkle

Center for Brain Science, Harvard University, Cambridge, MA, USA

Edward Soucy, Jennifer Segawa, Ross Mair & Talia Konkle

Department of Radiology, Harvard Medical School, Boston, MA, USA

Department of Radiology, Massachusetts General Hospital, Boston, MA, USA

Kempner Institute for Biological and Artificial Intelligence, Harvard University, Boston, MA, USA

Talia Konkle

You can also search for this author in PubMed   Google Scholar

Contributions

J.P. and T.K. designed the research, interpreted data, and wrote the paper. E.S. designed and constructed the physical apparatus for image projection and developed the computational warping algorithm. J.P. and J.S. collected data. R.M. performed an MRI signal quality assessment. All data preprocessing and experimental analyses were performed by J.P.

Corresponding author

Correspondence to Jeongho Park .

Ethics declarations

Competing interests.

The authors declare no competing interests.

Peer review

Peer review information.

Nature Communications thanks Jody Culham, Mark Lescroart and the other, anonymous, reviewer(s) for their contribution to the peer review of this work. A peer review file is available.

Additional information

Publisher’s note Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.

Supplementary information

Supplementary information, peer review file, reporting summary, source data, source data, rights and permissions.

Open Access This article is licensed under a Creative Commons Attribution 4.0 International License, which permits use, sharing, adaptation, distribution and reproduction in any medium or format, as long as you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons licence, and indicate if changes were made. The images or other third party material in this article are included in the article’s Creative Commons licence, unless indicated otherwise in a credit line to the material. If material is not included in the article’s Creative Commons licence and your intended use is not permitted by statutory regulation or exceeds the permitted use, you will need to obtain permission directly from the copyright holder. To view a copy of this licence, visit http://creativecommons.org/licenses/by/4.0/ .

Reprints and permissions

About this article

Cite this article.

Park, J., Soucy, E., Segawa, J. et al. Immersive scene representation in human visual cortex with ultra-wide-angle neuroimaging. Nat Commun 15 , 5477 (2024). https://doi.org/10.1038/s41467-024-49669-0

Download citation

Received : 01 September 2023

Accepted : 13 June 2024

Published : 28 June 2024

DOI : https://doi.org/10.1038/s41467-024-49669-0

Share this article

Anyone you share the following link with will be able to read this content:

Sorry, a shareable link is not currently available for this article.

Provided by the Springer Nature SharedIt content-sharing initiative

By submitting a comment you agree to abide by our Terms and Community Guidelines . If you find something abusive or that does not comply with our terms or guidelines please flag it as inappropriate.

Quick links

• Explore articles by subject
• Guide to authors
• Editorial policies

Sign up for the Nature Briefing newsletter — what matters in science, free to your inbox daily.

Help | Advanced Search

Computer Science > Computer Vision and Pattern Recognition

Title: llavolta: efficient multi-modal models via stage-wise visual context compression.

Abstract: While significant advancements have been made in compressed representations for text embeddings in large language models (LLMs), the compression of visual tokens in large multi-modal models (LMMs) has remained a largely overlooked area. In this work, we present the study on the analysis of redundancy concerning visual tokens and efficient training within these models. Our initial experiments show that eliminating up to 70% of visual tokens at the testing stage by simply average pooling only leads to a minimal 3% reduction in visual question answering accuracy on the GQA benchmark, indicating significant redundancy in visual context. Addressing this, we introduce Visual Context Compressor, which reduces the number of visual tokens during training to enhance training efficiency without sacrificing performance. To minimize information loss caused by the compression on visual tokens while maintaining training efficiency, we develop LLaVolta as a lite training scheme. LLaVolta incorporates stage-wise visual context compression to progressively compress the visual tokens from heavily to lightly, and finally no compression at the end of training, yielding no loss of information when testing. Extensive experiments demonstrate that our approach enhances the performance of MLLMs in both image-language and video-language understanding, while also significantly cutting training costs. Code is available at this https URL

Submission history

Access paper:.

• HTML (experimental)
• Other Formats

References & Citations

• Google Scholar
• Semantic Scholar

BibTeX formatted citation

Bibliographic and Citation Tools

Code, data and media associated with this article, recommenders and search tools.

• Institution

arXivLabs: experimental projects with community collaborators

arXivLabs is a framework that allows collaborators to develop and share new arXiv features directly on our website.

Both individuals and organizations that work with arXivLabs have embraced and accepted our values of openness, community, excellence, and user data privacy. arXiv is committed to these values and only works with partners that adhere to them.

Have an idea for a project that will add value for arXiv's community? Learn more about arXivLabs .