Prefer to watch than to read? Watch a video version of this article on LinkedIn or YouTube:
tl;dr: When it comes to the maximum number of parts that should be shown as a pie chart, there are plenty of opinions floating around out there. In this article, I argue that the maximum number of slices depends on the situation and that, in certain situations, even a 30-slice (!) pie chart can be the best choice.
As usual, before saying anything about pie charts, I need to point to this article, which outlines why I don’t think that pie charts are Satan’s chart type and which describes the specific situations in which I think they can be the best choice. With that out of the way, on to the question at hand: How many slices can you put in a pie chart?
In the Pie Chart Wars, many pie-chart-hesitant people say that pie charts should only be used to show the breakdown of a total with two (or, according to some, three) parts, and that a bar chart should be used if there are more parts than that.
According to that guideline, the pie chart below would be a bad chart because it’s showing six parts (i.e., six slices):
I think this chart communicates insights like those in the blue callouts more clearly than a bar chart would, though:
So, a pie chart with more than two or three slices can be the best option in certain situations, I think. A bar chart would be a better choice in the previous scenario if the insight to be communicated were something like, “Mocha Marvels have the largest market share,” but that’s not the kind of insight that we needed to communicate in that scenario.
How far can we push it, though? What if we needed to show a total with, say, 15 parts?
We probably all agree that this is a pretty messy chart, however, insights such as those in the blue callouts are clearer than they would be in a bar chart of the same data. What to do?
Well, since the smaller categories aren’t very relevant to the insights in the callouts, we might be able to get away with grouping those into an “all others” category, which makes for a cleaner-looking pie chart:
What if we couldn’t get away with lumping less relevant categories together, though, and we needed to show them all individually, as in the scenario below?
I know many experienced chart creators will recoil when seeing a chart like this and, yes, I agree that it does look pretty messy. It also, however, communicates the insight in the chart’s title more clearly than a bar chart would:
While this bar chart does look cleaner, it also doesn’t look like a total with many parts, but that’s the main message in this scenario. The bar chart looks like a set of similar values (which may or may not be parts of a total), which isn’t the same thing. Yes, readers of the bar chart will figure out that these are parts of a total, but only after reading the (textual) title and (textual) horizontal axis scale and extrapolating from there, whereas the pie chart skips those cognitive steps altogether. If the main message were about how the parts compare with one another, the bar chart would be a better choice, but that’s not the kind of message that we needed to communicate in this scenario.
How about a treemap?
To my eye, this is pretty similar to the 15-slice pie chart, both in terms of “messiness” and how well it communicates the insight in the title, so I could see going with either option in this case. If there were fewer than six or seven parts, though, I wouldn’t use a treemap because a pie chart makes fractions of the total (1/4, 2/3, etc.) easier to perceive when there are fewer parts. I’d also favor a pie chart if my audience wasn’t very data-savvy, since treemaps are a less common chart type that might not be familiar to them.
What if we needed to show a total with even more parts? Like, say, 30?
Yes, at this point, a pie chart starts to look a bit ridiculous, but this is due partly to the fact that all the parts are similar in size and all of them are labeled in this particular example. There are, however, (admittedly rare) situations in which that wouldn’t be the case, and in which even a 30-slice pie chart could be the best choice:
Yes, a 30-bar bar chart would be cleaner looking, but it wouldn’t communicate the insights in the blue callouts as effectively, and neither would a treemap (which also wouldn’t be much cleaner looking with data like this).
So, what’s the bottom line, here? How many slices can you put in a pie chart? As with most other chart design decisions, a simple rule of thumb such as, “Never use a pie chart to show more than X number of parts,” is, IMHO, problematically oversimplistic and will routinely lead to poor design choices. I wish it were that simple but, as I hope I’ve illustrated, it’s not. It depends on both the nature of the data (the distribution and number of parts, in this case), and the specific purpose of the chart (what insight you’re trying to communicate, what question you’re trying to answer, etc.).
The best advice I can offer is (as always) to make sure you’ve figured out why you’re creating a chart in the first place (what specific insight you need to communicate, what specific question you need to answer, etc.) before you start designing it. Then, try visualizing the data in various chart types and ask yourself which one best communicates the specific insight/answer/etc. that you needed the chart to communicate. Note that a lot of this trial-and-error process can be avoided by using the decision tree below, which is included in my Practical Charts On Demand course:
Agree? Disagree? Awesome! Let me know in the comments below!