an updated post on pies

Today's post is about one of my favorite dessert graphs: the pie chart. Those who follow my work know that I am not a fan. In fact, I've written posts with titles like "pie charts are evil" and given presentations called "death to pie charts." It can be fun to be a little provocative sometimes. Though one might argue this is taking it too far.

A wise person once said to me, "rather than ban pies altogether, teach people how to use them appropriately." This is sage advice. The challenge for me is that the appropriate use cases are few and personally, in pretty much every one of these cases, I'd opt for another approach.

Still, I do not want to spread misinformation. Pie charts are not inherently evil. Like pretty much any tool, they can be used well and they can be used not-so-well. Since my pie-bashing posts and presentations, new research has been conducted (by afore-quoted Robert Kosara and Drew Skau) that debunks some previously held common beliefs about pies. So I thought it prudent to write an updated post on the pie chart. While not exactly glowing, this will likely be at least a little more balanced compared to what you may have seen from me on this topic in the past.

The appropriate use case for the pie chart

Pies do a better job than probably any other visual out there at expressing the part-to-whole relationship. When you see a pie, you immediately have an understanding that it depicts a "whole" and can be sliced into pieces of that whole. It's also very easy for us to pick out a very large slice or a very small slice.

The limitation of the pie is that it's harder to say much more specific than that. When segments are close in size, it can be difficult to determine which is bigger or by how much. When that is an important goal, the pie chart breaks down. Ok, that didn't take so long—I'm already talking about what not to do with a pie chart. Let's shift next to more on that.

What not to do with pies

Pies seem to lend themselves more than other graph types to unnecessary—and often downright harmful—dressing up and embellishment. No other graph type is depicted in 3D or exploded as frequently as the pie.

A sampling of results from Google image search for "pie chart."

But the argument that it isn't fair to ban pies based on a bad example of a pie is a logical one. Due to their frequent misuse and in the spirit of teaching people how to use them correctly, here are some pointers on what not to do with pie charts:

• Don't use 3D effects or explode your pie. At best these add unnecessary clutter; worse, they can make it difficult or impossible to accurately understand the relative values in the pie. Here's a simple example.
• If the pie is depicting percents, it must sum to 100%. If it sums to anything other than 100%, something is wrong. If not percents, then the pie must sum to some meaningful whole.
• Don't have a ton of slices. There isn't a hard and fast rule here, but be reasonable. A pie showing a ton of tiny categories will be impossible to read (even if legible, hard to say much useful from, like this). Consider whether it might make sense to combine small slices into an "Other" category.
• Don't use a pie if the primary goal is to compare the size of the slices. The lack of alignment to a common baseline and area encoding of data makes this difficult. A bar chart will usually be a better option if comparing a quantity across categories is the primary goal.
• Don't use multiple pies and ask your audience to compare across them. This piece of advice may be controversial. But if the slices are different across the pies (which I'd expect they are if you have something interesting to say with them), the pieces shift around; this plus the spatial separation and lack of alignment to a common baseline make comparing slices across pies difficult. Perhaps you could get away with it if you're emphasizing a single slice across multiple pies, but if you want to do more than that, pies won't be a good approach. 

In conclusion

I will continue not to use pies. Does that mean you should follow suit? Not necessarily. What I advise in my workshops is, when you find yourself reaching for a pie, pause and ask yourself why. If you can answer that question, you've probably put enough thought into it to use the pie chart. Though when I step back and think about that advice—really, that's something we should do anytime we make any kind of graph. Think about what you want to enable your audience to do with the data you are graphing and whether the type of graph you choose is allowing for that in a straightforward manner. If you do that, you'll be well positioned to get your point across. And that's sort of the whole point, isn't it?

For more on this topic (and some varied perspectives on pies), check out the following posts:

• Understanding Pie Charts by Robert Kosara
• When Pie Charts Are OK (Seriously) by Ann Emery
• The Correct Way to Use Pie Charts by Randy Olson
• Save the Pies for Desert by Stephen Few 

The above isn't a comprehensive list—if you know of others worth mentioning, please do so in a comment.

I'll also note that one limitation of the pie study is that it included pies depicting only two segments. I hope to see further research expand on this and also look at pies having three or more slices.