I’ve talked a bit before about how I think data visualization is an important tool in the promotion of data use (for example, see this post illustrating my limited knowledge but vast appreciation of infographics). My view is that those who might make use of data must first understand it, and that many are more likely to understand data if it is not in huge tables or buried in paragraphs.
But oh the pie chart. Pie charts are so easily done so poorly. And so rarely done well. I feel like I still regularly see horrible (often 3D or greatly cluttered) pie charts. Most program evaluators, like all humans, fall into two camps — those who use and love the pie chart (sometimes naively — I hope I don’t fall into this camp, but fear at times that I do) and those who loath the pie chart. In the blogosphere, Flowing Data provides a good example of pie chart use here, while John Gargani asks whether the pie chart should be retired.
I’m not entirely ready to abandon using pie charts in my own practice, and here’s why. People understand what a pie chart is supposed to be at a glance (particularly when they’re simply) — that the slices are parts of a whole. This is not so immediately obvious with a column chart or the like. While it’s hard to compare the slices in one chart to the slices in another, it is possible to see the relatives size of the slices (especially if organized from largest to smallest and simply labeled. Certainly it is quickly obvious if there are very large or small slices, which is often the heart of the data being conveyed to begin with.
My answer, then, to John Gargani’s question is that we should definitely retire the bad piechart, and get better about making good pie charts. Frankly, the same can be said for almost every other kind of chart as well, in my opinion. And pie, frankly — some is do very good, some not so much (I prefer those with less traditional crusts though, so perhaps I am not the best authority on pie or pie charts).
Here is a pie chart from a recent report (on the 2012 Marylhurst University Alumni Survey):
Setting aside the awkward title language, I hope the message is clear — that responses are quite mixed, with the largest proportion of respondents reporting a weak connection to Marylhurst. It should be easy too, to add the various slices together as desired (i.e. 35% of respondents who are connected or strongly connected). That said, I’d LOVE to find another way to represent this information, given the challenges inherent (not the least of which is exhaustion) with the pie chart. What do you think? Suggestions welcome!
The bottom line for using pie charts as far as I’m concerned: never over-complicate. In other words, keep it simple smarty. Include only the information, text, and color, necessary to illustrate the data involved. The fewer colors, textures, etc., the better. The simpler the message the better. I’m not sure I’ve followed this rule perfectly with the above example of my own — I will likely try working with shades of one color for future pie charts — when pie charts are necessary at all.
What camp do you fall in — do you love or hate them? If you do see value in the pie chart, in what scenarios do you use or appreciate them? Why or how do you think they can be helpful, if at all? Do you know of other resources that might help those of us still attached to the pie chart? What is your favorite pie recipe?
Oh, and as a bonus for making it through this post — here’s another great somewhat related example of the power of simple data visualizations from Flowing Data that I can’t help but include — How simple charts tell a story.