When I first started studying Data Visualization, I began to see a diverse range of opinions on pie charts. Some seem to like them while some seem to hate them! So I set out to examine both sides and learn more about why people say they are either good or bad. I learned alot from examining the texts of several Data Visualization experts and have attempted to distill their expertise into a short blog post as food for thought. If this has whet your appetite, please dig into the sources cited because you will learn so much!
Opinion 1: Pie charts are bad
There seems to be a few reasons why people discourage pie charts.
First, academics have studied the perceptual tasks involved in interpreting data graphics and suggest that pie charts can be misinterpreted. Cleveland and McGill’s elementary perceptual task (EPT) (1) lays out the building blocks for visually comparing quantities. They suggest that there is a hierarchy of how people use charts to correctly interpret the underlying data. (2)
They suggest that as chart types move down the hierarchy, it becomes more difficult for users to correctly interpret the data.
Pie charts use angles thus they fall lower on the hierarchy and are more open to misinterpretation by the audience. They also use area, position, length, and curvature adding to the confusion about which task to perform.(3) Bar charts are suggested as a better alternative for comparison for the size of the bars on an axis compared to the angle of the wedges. (4)
The second reason is bad math. Pie charts present a proportion of the whole. (5) So the parts should sum up to the whole, or 100%. Some pie chart creators do not understand the math or data methods behind the numbers and show overlapping or mutually non-exclusive categories in pie charts. That is misleading. See this example below. (6)
The third reason for arguing against the use of pie charts is they are too easy to make with default settings. There is an abundance of bad examples brought about by bad defaults in dataviz tools. For example, default settings in Excel charts allow users to add 3D effects. (7) This distorts the data even further.
Here is another example. (8)
Finally, pie charts are discouraged when there are too many categories or the categories are too even in size. (9) Going back to the perceptual tasks, (10) when pie charts include too many categories, the angle gets more difficult to interpret. See example below. (11)
The inclination is to then add data labels, but that increases the clutter. See this bad example. (12)
Opinion 2: Pie charts are okay, when done correctly
Some people argue that the wedges are actually easier to read for proportion because the circular shape conveys the idea of “part of a whole”. (13) The bars in this case below do not suggest that they are part of the whole 100% as well as the pie chart.
This is specifically the case when the user needs to pay attention to a particular large wedge for comparison against smaller wedges. (14)
Also pie charts are good when there are only two or three categories or wedges for comparison. (15)
Especially the case with small multiples where the wedges can be compared across many categories. (16)
I think that both sides make good arguments. Read my sources for more on the debate.
For my part: I think that as long as I use the pie chart as intended (to show a portion of a whole), don’t clutter with unnecessary features (like 3D effects or too many wedges), and keep the focus on the data, pie charts are just fine.
(1) Cleveland W.S., McGill R. (1989). “Graphical Perception: Theory, Experimentation, and Application to the Development of Graphical Methods”. Journal of the American Statistical Association. 79(387): 521–534.
(2) Bergen, Silas (2017). On elementary perceptual tasks. https://driftlessdata.space/post/on-epts/
(3) Viguier, Clément (2018). https://medium.com/@clmentviguier/the-hate-of-pie-charts-harms-good-data-visualization-cc7cfed243b6
(4) Yi, Mike (2019). https://chartio.com/learn/charts/pie-chart-complete-guide/
(5) Wikipedia (2021). https://en.wikipedia.org/wiki/Pie_chart
(6) Mccready, Ryan (2020). https://venngage.com/blog/misleading-graphs/
(8) MacPherson-Krutsky, Carson (2020). https://theconversation.com/3-questions-to-ask-yourself-next-time-you-see-a-graph-chart-or-map-141348
(9) Yi, Mike (2019). https://chartio.com/learn/charts/pie-chart-complete-guide/
(10) Cleveland W.S., McGill R. (1989). “Graphical Perception: Theory, Experimentation, and Application to the Development of Graphical Methods”. Journal of the American Statistical Association. 79(387): 521–534.
(11) Hickey, Walt (2013). https://www.businessinsider.com/the-27-worst-charts-of-all-time-2013-6
(12) Foley, Katherine Ellen (2020). https://qz.com/1872980/how-bad-covid-19-data-visualizations-mislead-the-public/
(13) Randle, Tom (2019). https://www.geckoboard.com/blog/pie-charts/
(15) Emery, Ann K. (2015). https://depictdatastudio.com/when-pie-charts-are-okay-seriously-guidelines-for-using-pie-and-donut-charts/