In the name of all that’s good in the world, please, please, please.
Our friend Stephen Few has been saying this for a while. It doesn’t seem like many of us are listening.
Pie charts are useful for one thing and one thing only: communicating that a series of numbers add up to 100%. That’s it.
Pie charts fail at everything else.
Generally we use pie charts to communicate a “part to whole” relationship. The problem is that they do a really bad job. That doesn’t seem to stop them from being ubiquitous; the fast food of the data visualisation world. They are tempting for a quick calorie hit, but ultimately lacking in nutrition.
The difficulty with pie charts starts not with the charts themselves, but with the limitations of human perception. The reality is that human beings are not very good at comparing the size of different areas. We are much better at comparing the size of different lines.
Look at these circles, imaginatively labelled “A” and “B”.
How much bigger is circle B than circle A? It’s more than 4 times bigger, but is it as much as 10 times bigger? We can tell that it’s bigger, but we do a poor job of saying by how much with any real confidence.
Now look at these two lines that we’ve called “C” and “D” (we’re nothing if not creative):
How much longer is line D than line C?
Most people find it easy to tell that line D is around 3 times as long as line C, but it is hard to say with confidence how many times larger circle B is than circle A.
How does this phenomenon play out when it comes to the humble pie chart?
Take a look at the following chart, showing visitors to this website from 4 different sources:
In this chart we can tell 3 things:
1. All parts add up to 100%
2. Organic Search accounted for the largest single share of traffic (somewhere just short of half of all visits)
3. Email marketing, Social media and Referrals were probably all about the same, but it’s hard to be sure and we certainly couldn’t tell with certainty which of those 3 is the largest, and which is the smallest.
So, in order to make this chart useful to readers, we have to give them additional guidance as to the relative size of the smaller parts, and tell them just how big the larger part is.
It’s hard to tell without labels, so we dutifully add the labels, and now the chart looks like this:
But when we stop to look at it, we realize that we have, in effect, created a circular table.
And worse than that, it’s a colour coded circular table. We have to look back and forward between the 3 smaller components in order to match the piece of pie to the label in the legend, and then, in our heads, try to rank them using the numbers. It’s not the hardest thing you’ll ever have to do in life, but you can probably think of more fun ways to pass your time. (Like doing this for example).
And that’s before we succumb to the temptation to add some “cool” 3D effects and perspective, thus making it truly impossible to make any meaningful visual comparison. Don’t even get me started on this mess:
Stephen Few suggests, and I whole-heartedly agree with him, that a more useful presentation looks like this:
I know, it’s not as “fancy” as that 3D exploding pie chart, but it actually communicates some useful information.
1. The items are placed in order for us; no effort is required to rank the components
2. The bar for each category is placed next to the label for that category, we don’t have to track back and forward to make sense of the chart
3. The chart takes advantage of the fact that humans are good at comparing lines – we can now see clearly the differences between the 3 smaller components
4. The data values are also presented in line with the bars, again making visual tracking much easier – no colour coding is required to relate the categories to the values
6. The chart still communicates that the numbers add up to 100%
So, let’s hear your arguments in defence of the pie chart; Stephen’s been saying this stuff for years, and pie charts don’t seem to be going anywhere.