By presenting data effectively we can increase the number of people who actually read our data and reports. Harnessing the power of effective data reporting allows us to better communicate with peers, colleagues and clients, helping them make data-informed decisions.

At Evergreen Data, we teach a data visualization technique that is grounded in visual processing theory. Choosing the right visualization or chart is one of the most important steps in visualizing data: When we discuss data relationships, we need to pair that with an understanding of human perception. This can reduce clutter and emphasize the most important points of data, allowing us to capture our audience's attention with impactful visuals.

To start this conversation, we go back to the landmark work of William Cleveland and Robert McGill. Cleveland, one of the pioneers of graph design, worked with McGill to build a theory about the ways of showing data that are the easiest to interpret and the most accurate. They started by developing a theory of "elementary perceptual tasks," which are the most basic visual tasks humans perform when perceiving graphs. They ran experiments to determine a hierarchy of what tasks (and graphs) we are best at decoding.

Below are the elementary tasks tested in Cleveland and McGill's study in hierarchical order, ranked from tasks at which we are best to those at which we are worst. Next to them are chart examples that use the associated perceptual tasks.

As the ranking shows, judging position is our easiest task. People are best at judging position on a common scale – such as dots on a line. Whenever possible, designers should graph data like this. My favorite graphs that use position on a common scale are dot plots and lollipop charts. Both are less common visuals than others because they are not Excel default charts. Although they are a little more time-consuming to make, their payoffs are worth it. If you want to learn how to make these graphs, check out Stephanie Evergreen's blog posts here and here.

The thing we are second best at is position on different scales. When there are two graphs next to each other, we need to make sure they both use the same scale because people will make comparisons between them.

Next come, in this order: length, direction and angle. Length is shown in bar charts. Length and direction are displayed in line charts. Angle is comes into play in pie charts. Cleveland and McGill ran more tests specifically between length, direction and angle, and found that error rates were much higher for angles. Pie charts produced the most error.

This can be detrimental news to some pie-loving folks. Most people I hear from have two, maybe three go-to Excel charts they use. Often, one of those go-to charts is a pie. Pie charts are great at showing parts of a whole, but they easily get cluttered, and we are bad at decoding information from them. Still having a hard time giving up the pie? Check out the examples:

Pie charts can be a useful graph for the appropriate data. Here are some tips on how to make pie charts effective.

Graph fewer than three categories
Viewers must be able to tell which category is larger/smaller without data labels
Graph data from largest to smallest, clockwise around the pie
Directly label the slices
Use non-default Microsoft colors

Here is an example of a great pie chart:

Now, let's circle back to the Cleveland and McGill article, where some elementary perceptual tasks remain untested. Cleveland and McGill didn't test area, volume, curvature and shading because, based on the other experiments, they found that these elements were too hard to interpret. Back when this study was conducted in 1984, they used the ugly crosshatching technique for shading, and further research has shown crosshatching can produce optical illusions (Tufte, 2001). Luckily, nowadays we can use something much more effective: color.

These basic visual tasks form the foundation of how we represent data relationships. We should always strive to graph high up on the Cleveland and McGill hierarchy to aid the interpretability of our graphs. For more resources on chart choosing, check out Stephanie Evergreen's website and blog.

References

Cleveland, W. S., & McGill, R. (1984). Graphical perception: Theory, experimentation, and application to the development of graphical methods. Journal of the American statistical association, 79(387), 531-554.

Evergreen, S. D. (2016). Effective Data Visualization: The Right Chart for the Right Data. SAGE Publications.

Evergreen, S. D. (2013). Presenting data effectively: Communicating your findings for maximum impact. SAGE Publications.

Evergreen, S. (n.d.). Evergreen Data Visualization Blog. Retrieved December 09, 2016, from http://stephanieevergreen.com/blog/

Tufte, E. R. (2001). The visual display of quantitative information (2^nd ed.). Cheshire, CN: Graphics Press.