You can have some pie (charts)
One of the first things I learned as a data science student was that I should banish pie charts from my repertoire. Not only is it practical advice for novice data students, it’s a full-on meme. As new students, we spent an hour-long data visualization lecture sharing this inside joke about how awful pie charts are and quickly felt initiated into the data-visualization in-crowd. In preparation for writing this piece, I did a quick Google search to see what the more formal word on pie charts is. I got some pretty strong feelings, such as: “People, please stop making pie charts”, “Save the pies for dessert”, and even “How succinctly can I explain why pie charts are evil”. The message is clear: the cool data kids don’t do pie charts.
The rejection of pie charts in the articles mentioned above is very well-reasoned. The case against pie charts is essentially:
- Humans are bad at making visual estimations of angles and round shapes, and therefore have a harder time interpreting pie charts
- Pie charts tend to be a less efficient use of space, knocking points off from a design perspective
- Pie charts are often further obfuscated by adding 3D effects, layers of concentric rings representing additional dimensions of data, or lined up in an attempt to communicate a time sequence
These points are all excellent, and should encourage the vast majority of folks about to crank out another pie chart to opt for a simpler, clearer bar plot instead.
The gang of design-savvy pie-chart-bashers waiting in the wings to smash your visualization to pieces offer a few concessions about when pie charts are maybe less awful. They’re allowed if you have sufficiently few categories (3–4 seems ideal, but some were willing to go up to 8). Include a legend ordered by size, percentage labels to demystify segments, or other supplemental annotations, and your pie chart sin may be deemed less egregious. Even with these concessions, all the think pieces slamming pie charts ultimately conclude that pie charts are still always the wrong tool for the job. Anything a pie chart can do, a bar plot can do better.
So are pie charts ever the best solution? If so, when?
In “Why pie charts often suck”, Alex Quach describes why his personal finance application opted for bar charts over the ubiquitous pie chart: “A chart is a good fit when the chart’s strong suits align with the message we want to convey.” This is a powerful summation of why pie charts are overused, but it also points us to what pie charts’ hidden purpose is.
Pie charts describe how a whole breaks down into smaller segments. Using that definition, an eager Excel-wielder can shoehorn almost any visualization into the pie chart category and produce the confounding visualizations that birth these impassioned internet rants. It’s possible to argue that your pie chart showing how the whole of your survey respondents breaks down demographically fits into this category, when a bar chart would probably be a better way to convey your data. To get at when pie charts shine, one has to be even more restrictive.
Here is where pie charts pull ahead of the pack: pie charts are great for communicating allocation of a finite resource. If you want to increase one slice of the pie, it’s going to come out of another slice. Pie charts are useful for conveying budgets, because they demonstrate that any money added to a given category must come out of another. Illustrating how employees divide their time between tasks is a great job for a pie chart. Bar charts (often touted as the universal better option to pie charts) do not establish zero-sum relationships between the groups they represent.
Ultimately, pie charts’ fundamental strength is in conveying limitations. Talking about limitations is a daunting prospect; nobody wants to be a naysayer. Yet being able to do so, and do so with grace and tact, is an essential skill in any industry. It’s one that pie charts can help accomplish.
There should be lots of factors that go into your selection of a data visualization. If you trust that your audience will grasp the context of your data enough to understand the relationship between categories, then a pie chart may ultimately not be worth the sacrifice in clarity. If you do select a pie chart, make sure that it is as clear and unfussy as possible (skip the 3D effects). But there might be an occasion where you really want to hammer home that you’re talking about a zero sum game, and when that happens: consider a pie chart. If you’re looking for a data visualization to make a sassy meme about, consider the word cloud.
