Why Pie Charts are (Usually) a Bad Idea (Part 1)

My mentors have already written well on this topic. Here, I share some bits that resonated with me.

Pie charts are among the most sought-after and brandished charts in the corporate world, and, I have noticed, in the marketing materials of information graphic software publishers as well.

However, among serious information graphic designers the overwhelming opinion about pie charts is they are at best a poor choice. The reasons for this are amply documented in the in the blogosphere, and someday I will post references.

Meanwhile, I find the persistent popularity of pie charts to be profoundly strange. I will talk about this in two posts.

Three major reasons to avoid pie charts:

  1. They are Ineffective: Pie charts make it nearly impossible to compare relative values.
  2. They are Misleading: At best, pie charts suggest the entities shown comprise a “whole thing”, but this is often not really the case. At worst, they actually distort values.
  3. They are Distracting: For inexplicable reasons pie charts are strong attractors for “chart junk”–meaningless graphical effects that have nothing whatsoever to do with the data or the story you are trying to tell.

The Pie


Let’s address the three problems mentioned above with the pie chart on the right.

  1. Ineffective. Can you identify the company with the highest revenue? The lowest?
    • Probably, with some effort. Had I put the values (or percentages) on the chart, you could probably do it more quickly.
    • But with values you are forced to scan around to find the largest and smallest. In so doing, one must commit temporary results to short-term memory to make the comparisons. Values belong in tables, not in charts. That’s what tables are good at — displaying numbers.
    • You will notice Excel offers a lot of options for adding values (nominal and/or %) to pie charts — that’s probably because pie charts are ineffective without the values!
  2. Misleading. Are these the stories this chart should tell?
    • These five entities make up the entire universe of entity-space. In other words, these five companies are the only ones to worry about. This is probably not true, but the appearance of a “complete pie” lends to the perception that it is. This might not matter in all cases, but someone could make a bad assumption about completeness when they interpret a shape like this.
    • Innotech is bigger than AlphaBravo. The Innotech (red) slice seems to have an angle a little more than 90 degrees, while the AlphaBravo slice (darker blue) looks like about 60 degrees. That makes Innotech bigger, right? Maybe 50% bigger than AlphaBravo? This is patently false. Without values, this chart is not merely ineffective–it is also misleading. One could argue the brain can interpret the red slice as smaller than the blue one by processing the distortion due to the tilt–and you would probably be right. But you should ask yourself why you would want to force your viewers to go through such mental hoops to interpret such a presentation.
  3. Distracting. Not only does the edge-on perspective make it impossible to accurately compare the wedge sizes, it adds nothing to (and indeed, distorts) the message. The added “dimension” of height is also meaningless and artificial. These visual “enhancements” are just encumbrances and obscure the message.

A Better Way

Consider now then bar chart at right. It uses the same data as the pie above. Notice how it addresses each of the problems with the pie version:

Better than pie

  1. Effective*: The reader can immediately see who leads the revenue race. In one quick glance the eye can pick out the longest (and shortest) bars, and immediately correlate them with the name of an entity.
  2. Representation. There is no suggestion there that these five entities make up a whole anything. There is only the implicit fact that these are the five entities we have chosen to show. It is also clear Innotech is smaller then AlphaBravo, a fact that is not obvious in the pie.
  3. Succinct. This is only one click away from a default bar chart in Excel 2007  (I moved the legend to the bottom). In fact, there is no need for a legend at all since the title tells you what you are looking at. There are other small improvements that could be made to reduce “chart junk”, but this isn’t too bad for a default.

*An important part of effectiveness is making sure the question is well-formed. Only then can the message be effective. Those are topics for another day.

In Part 2 I will talk about a possible use for pie charts.

Advertisements

2 Responses to Why Pie Charts are (Usually) a Bad Idea (Part 1)

  1. Pingback: Why Pie Charts are (Usually) a Bad Idea (Part 2) « andyholaday

  2. Pingback: [Onbeantwoord] cirkel diagram % procenten naast de legenda van de diagram

Leave a Reply

Fill in your details below or click an icon to log in:

WordPress.com Logo

You are commenting using your WordPress.com account. Log Out / Change )

Twitter picture

You are commenting using your Twitter account. Log Out / Change )

Facebook photo

You are commenting using your Facebook account. Log Out / Change )

Google+ photo

You are commenting using your Google+ account. Log Out / Change )

Connecting to %s

%d bloggers like this: