Data Visualization Case Study: Pie Charts are Evil?
Pie Charts are evil? Well, given the correct situation, pie charts can be used to your advantage. The following case study uses Power BI samples to review examples and provide recommendations to get the most out of an often misused visualization.
I have seen many examples of poorly used pie charts and shiver each time I come across one in a report. The goal of a pie chart is to show how each segment relates, not only to each other, but the entire data set.
Issues with Pie Charts
Most of the problems you come across are illustrated by the following example published by Statistics Canada. This shows the distribution of Aboriginal population by province over two surveys. You would want to compare the distribution between provinces and compare that to the values 16 years later.
Too Much Data Makes for too Much Noise
The more data you have on any chart makes it difficult to see differences between the various data points. As illustrated in the above example, the meaning behind the data can get lost because of the amount of data on the graphic. Pie Charts make this more difficult because each segment gets divided up between the 360-degree nature of the visualization. Large differences are easier to see, but if there are groupings of like values, differences are hard to make out.
Cannot Determine the Difference in the Data
When you look at the different segments, it is tough to see the differences between each element, as they are all in a circle and the eye finds it difficult to relate each segment to each other. When comparing the differences between the years we are faced with two separate charts which can make it difficult for the eye to compare the results.
The Story you are Trying to Convey
Given the two issues above, you can see how pie charts can let the story you are trying to tell get lost in the data. This can lead to misinterpretation and loss of confidence in your argument. The charts below, show the same information using PowerBI, however, demonstrates a horizontal bar chart that displays the data and trending more effectivly than the original.
Make Pie Charts Work: A Case Study
A group had a survey that asked their organization which month they would like a corporate event. The survey’s author and management thought September would be the preferred month for the event.
The survey results, pictured below with the original chart (1), with the vast majority selecting March. This created a problem for the committee, as they now needed to convince management that March was the preferred month and needed a strong visualization they could include in their report.
Example 1 – Original Pie Chart
For the report, the main story was the overwhelming choice of March for the event. There is an abundance of information not required for the view. If the month and percentage are supplied, the eye is drawn to the clumping of information at the top of the chart. This takes away from the message as the grouping takes the eye’s focus away from the result. The eye is drawn to the more detail at the top of the chart first.
Example 2 – Final Pie Chart Version
The chart clearly conveys the survey result that March was the preferred month. With a value so much higher than the others, denoting only the main response fits in this situation. The point of the topic is displayed without letting the reader get lost in all the other information.
Additionally, the value for March is more than half which shows the reader that no matter how many responses are provided, this one the majority. This makes it safer to group all the other responses to Other.
Example 3 & 4 – View with all values
You still may want to provide a summary of all responses. Having a horizontal bar chart is one that I prefer. Example 4 would be my final version if all the data was required. This has the data bar you want to draw attention to darker than the others. All the information is displayed and yet your eye is brought to the point you want to illustrate.
A pie chart shows the relationship of the parts of a whole. When data points are easily discernible, differences in those relationships can be demonstrated by using pie charts to your advantage. Every chart has a purpose. With the right question and the right data, you can make pie charts work for your data visualization project.