Data Viz 101: Say Hello to the Pie Chart
Hello class, and welcome back to Data Viz 1010, Beutler Ink’s introductory course on the wonderful world of data visualization.
In our last installment, we looked at the line graph, one the oldest and most basic forms of data visualization. Today, we'll cover the pie chart, a type of graph commonly used in conjunction with percentages. Like the dessert from which it draws its name, a pie chart is large divided into pie slice-like sections, which the arc length of each sector proportional to the percentage of the whole it’s supposed to represent.
What is a pie chart?
A pie chart is best for making part-to-whole comparisons, like breaking down the percentage of all different colored M&Ms in a single bag—e.g., 16% of the M&Ms were brown, 22% of the M&Ms were red, 13% of the M&Ms were blue, etc. A pie chart includes every possible category (even if one category is a composite "other") and all of the percentages add up 100% (although in some cases there may be variance due to rounding). Again, this completeness is clearly indicated by the design, which clearly conveys fractions of a whole.
One knock on pie charts is that they're so simple and intuitive that they can actually be a little boring. Now, there's nothing wrong with "boring'" functionality in our book, but pie charts actually present ample opportunities to be visually creative, as this clever and eye-catching graphic from WIRED makes clear. You almost don't even need the call-outs to tell you this is about the relative popularity of Girl Scout cookies.
Here's another creative example—one of our own!—using a donut chart variation, in which the center of the pie chart is hollowed out. The empty space in the middle allows us to include a visual flourish while keeping the data relationships intact. We also use callouts to draw attention to segments within the two depicted categories.
When pie charts aren’t your friend
But there are certainly times when pie charts are not your friends—namely, if you're trying to depict portions that do not add up to a meaningful whole. Consider, for instance, this Fox News pie chart from 2009 that attempts to depict support levels for the three leading (and at that point, purely hypothetical) GOP presidential candidates for the 2012 race.
We don't need to tell you that 70 + 60 + 63 does not add up to 100! A simple bar chart would have been a far preferable data visualization in this case.
Another potential problem is including too many categories. Like a real pie, a pie chart could only be sliced so many times before things get a little sloppy. Consider this graphic from a (now updated) Wikipedia page listing U.S. states by population.
The graphic makes clear that the first few states are clearly larger than any of the rest… and after that the graphic becomes a cluttered mess of color slices. In general, pie charts should no more than five to seven different categories. Any more than that and the pie chart loses its functionality. With this graphic, even though the slices do represent parts of the whole, a bar chart would have been a more effective way to compare state population sizes.
Another thing to keep in mind with pie charts is the order of the slices. which ideally should be presented from biggest to smallest, proceeding clockwise starting at 12 o’clock. This will make the data arrangement and emphasis very clear to the viewer. You can also group some of the smaller slices together into an “other" category if they’re less necessary to the information you're trying to convey.
In short, the pie chart is a reliable data visualization—easy to interpret and pleasant to look at.
Pie charts' effectiveness comes in part from our familiarity with the format, but there are opportunities to be creative. Care should always be taken, though,not to overload the chart or depict data that does not represent parts of a whole.
