Pie charts: seldom or never?

So, the other day, I was casually browsing Twitter when I saw that Meg Duffy had posted some preliminary results from her latest survey on authorship over at Dynamic Ecology. “Oooh, pretty graph,” I tweeted of her pie chart. To which, I received this reply:

@margaretkosmala really? It’s terrible. Pie charts should just die @duffy_ma

— Gavin Simpson (@ucfagls) April 21, 2016

Which took me out of casual mode and made me stop and think. Because, to be honest, I am not all BFF with pie charts. I’m not even a minor fan of pie charts. I spent a bit of effort to try to convince my PI to get rid of an atrocious slide in his tenure talk that contained four pie charts. [1] He didn’t. And he got tenure anyway. So clearly pie charts aren’t make-or-break. Gavin and I exchanged several tweets about good data visualization (“data viz” for those in the know), and I think we’d generally agree on the basics. But I wonder if there might not be two pie chart camps among those savvy in data viz: those who are against pie charts at all costs, and those willing to tolerate their (limited) presence.

I’ve been interested in data visualization and graphic design for a while now, and back in the fall I read Tufte’s classic and influential The Visual Display of Quantitative Information. (Quick review: fun to read, lots of pictures, goes fast, covers all the fundamentals.) Tufte doesn’t mince words when it comes to pie charts:

A table is nearly always better than a dumb pie chart; the only worse design than a pie chart is several of them, for then the viewer is asked to compare quantities located in spatial disarray both within and between charts […] Given their low density and failure to order numbers along a visual dimension, pie charts should never be used.

And after reading this, I felt vindicated in my loathing for pie charts and swore them off forever. But then I kept reading. Next up on my data viz reading list was a pair of books by Nathan Yau: Visualize This and Data Points. (Quick review: no need to read both, if I had to pick I’d pick the latter, definitely more up-to-date than Tufte, but denser and not quite as fun.) And Yau, unlike Tufte, doesn’t believe that pie charts are inherently bad. He doesn’t love them, of course, and gives some nice alternatives to them. But he sometimes uses them himself.

Sometimes a pie chart makes sense.

All the arguments against pie charts are true. People are terrible at comparing sizes of pie pieces, so for comparing exact values they’re worthless. And people are also terrible at comparing angles. In any case, why would you use a two-dimensional circle to represent a set of one-dimensional values? A short table uses “less ink” — a Tufte tenet. But if you must make a graphic, there are several nice alternatives. The stacked horizontal bar graph, for example, is one of my favorites. It essentially unwinds the pie into a straight single dimension. And for pie charts with many slices, I think this is the way to go.

But sometimes a pie chart makes sense.

Sometimes pie charts make sense for the simple reason that people are familiar with them. If I have your attention for only a brief period (in this blog post, during a talk), and I show you a circle divided up into pieces of different colors, you know exactly what you’re looking at. If I show you a horizontal bar graph, I have more explaining to do: “The full width of the bar represents 100% and it’s divided up into parts that represent the fractions that make up that 100%.” Too much to say. Too little time.

That being said, I think pie charts serve us best when they’re limited to two major pieces. Let’s say I set up a survey that asks whether pie charts should never be used or pie charts should seldom be used. I could report the results like this, in a decidedly Tuftian manner:

I don’t know about you, but I’m lazy, and comparing those numbers means doing some (albeit gentle) mental gymnastics. I want a picture! Many pie chart loathers would recommend something like this:

Okay. That’s better. But I still have to stop and inspect that graph a bit. I need to read the axis carefully to understand that the width of the bar is 100% and I learn that “Never” is more than half by comparing it to the axis, not because it’s easy to see that it takes up more than half the distance from 0% to 100%. Least conveniently, my eyes need to flip back-and-forth between the legend and the data, because it turns out to be non-trivial to label a horizontal bar chart. (In Google Sheets, the program I made to make the charts, you can’t directly label horizontal bar charts. I also played with Plotly for 20 minutes and was likewise unable to make a labeled horizontal bar chart. [2] Twenty minutes! On a fake graph! What am I doing with my life?…)

So, let me just suggest: sometimes a pie chart makes sense.

You see what I mean? You get it immediately. Slightly more than half of people think you should never use pie charts. Most of the rest think they should seldom be used. Simple. To the point. Intuitive — because you are used to seeing pie charts.

So when exactly does a pie chart makes sense?

1. A pie chart might make sense if you have two major categories you want to compare (and few minor ones that can be lumped).
2. A pie chart might make sense if you don’t need your audience to grasp the specific values, but rather the the approximate ones. People are pretty good at inferring when a circle is divided into halves, quarters, or thirds.
3. A pie chart might make sense if your audience will only see your visualization for a short period of time, like during a talk or on social media.
4. A pie chart might make sense if you’re doing an exploratory analysis, because they’re quick to make with just about any software. Why would you spend twenty minutes trying to make a decent horizontal stacked bar chart when you could make a pie chart in just a minute or two?

Most of the time you can come up with a better visualization than a pie chart. But pie charts do have their limited place. I am decidedly in the “seldom” category.