Blogs

Data visualization playbook: Using function to drive design

Data scientists must be selective when choosing what type of visualization to use with a data set. In particular, should we select a visualization before working with the data, or should the same type of visualization always accompany a particular type of data? To decide, let’s explore another, more foundational question: Which comes first—the data, or the visualization?

Putting a face on data

Consider a nonprofit organization that wishes to create a data visualization for an upcoming report. The members of the board decide to create a visualization depicting the distribution of funding for all initiatives, across eight different types of projects. To do so, they commission a graphic designer to create a distinctive icon to represent each project.

To emphasize the icons, the new visualization arranges them in the circular format shown in Figure 1. The icons are also ranked by amount of funding received, with each icon sized to scale.

Figure 1: The distribution of funding for projects across the organization’s major initiatives.

http://www.ibmbigdatahub.com/sites/default/files/data_visualization_playbook__5__-_figure_1.jpg

Identifying the problem

However, for several reasons, the new visualization failed to accomplish the board’s full purpose in creating it.

Issue 1: The more-is-better approach

Noticing that the visualization did not highlight information about the organization’s two most important initiatives, health and education, the board commissioned two additional visualizations for only those initiatives in the same format, as shown in Figure 2a.

Figure 2a: Two additional visualizations were created for the health and education initiatives, using the same format.

http://www.ibmbigdatahub.com/sites/default/files/data_visualization_playbook__5__-_figure_2a.jpg

Issue 2: Functional limitations

Each individual figure accomplished the board’s objectives in commissioning it, using specially designed graphics to display the amount of funding for each project category. However, the visualizations proved less effective than expected and did not effectively communicate information to readers.

Figure 2b: The three visualizations appeared separately in the report, making comparison of initiatives difficult.

http://www.ibmbigdatahub.com/sites/default/files/data_visualization_playbook__5__-_figure_2b.jpg

Once inserted into the report, each visualization filled half a page. Moreover, because the figures were separated by pages of text, comparison required readers to flip between visualizations.

What’s more, the images’ visual similarity led readers astray, creating the impression that the visualizations represented similar data sets. However, the visualizations for the health and education initiatives were merely spotlights on two important portions of the whole amount, whereas the first visualization depicted the total amount for all initiatives—including the amounts broken down in the other two visualizations.

Accordingly, some readers did not understand that the amounts given in the first visualization also included the amounts shown in the other two, a problem compounded because the largest icon in each visualization was sized the same as the largest icon in each other—yet represented a different dollar amount. Indeed, although two visualizations focused on health and education, no visualization similarly highlighted the remaining initiatives—employment, cultural and social.

Iterating toward a solution

When the board attempted to address the issue, several rounds of revisions ensued, each more closely approximating the board’s intent.

Step 1: Rethinking the design

The icons’ circular layout in the initial visualization worked well for a set of icons displayed in a single figure but frustrated comparison of icons across several figures. Changing the design from a circular format to a linear one reduced the amount of space required to display each series of icons and consolidated all three visualizations into a single image, allowing readers to easily compare icons across initiatives.

Figure 3: The revised visualization consolidated the three former visualizations by arranging series of icons in such a way as to allow their comparison.

http://www.ibmbigdatahub.com/sites/default/files/data_visualization_playbook__5__-_figure_3.jpg

Step 2: Redesigning by resizing

Though the new format helped readers compare icons, sizing each icon to reflect its associated percentage of overall funding decreased the figure’s visual appeal, as shown in Figure 3. The icons for the projects that received the least funding became unrecognizably small, rendering them useless. The new visualization didn’t capitalize on the newly designed icons, and it frustrated the organization’s attempt to associate particular icons with particular project types. However, although scaling the icons decreased the image’s visual appeal and reduced its usefulness, readers were able to easily understand the distribution of funds within initiatives.

Step 3: Choosing the right data

To provide users with information about all initiatives in the organization, the designer reworked the visualization once more, as shown in Figure 4, replacing dollar amounts for overall funding with figures representing combined funding for projects associated with the employment, cultural and social initiatives. The designer also sized all icons uniformly, taking full advantage of their distinctiveness. The resulting visualization provided a complete overview of the distribution of funding across initiatives and project types alike.

Figure 4: The final visualization showed the amount of funding received by eight types of projects across the organization’s five initiatives.

http://www.ibmbigdatahub.com/sites/default/files/data_visualization_playbook__5__-_figure_4.jpg

Using design to support function

Choosing design over function can lead to redundant visualizations that fail to take full advantage of their graphics. By welcoming changes to the original design, the organization created an effective visualization that preserved the most important features of the original design. The uniformly sized icons allowed readers to quickly identify issue areas, and the linear layout allowed comparison of sums without page-turning. Moreover, by arranging projects in descending order along a horizontal line, the final visualization preserved the relative rankings of amounts distributed for project types within an initiative.

Balancing design and function can produce a beautiful and effective data visualization, but finding balance can be tricky. In particular, overemphasis on visual design can diminish a visualization’s ultimate effectiveness. To create a powerful visualization, keep the big picture in mind—and remember that more isn’t always better.

Discover how the IBM advanced analytics portfolio can help you visually explore your data to find patterns and uncover insights.