Pie charts are not always wrong
In a comment, Wade says “I’ll be the contrarian here and take the position that using pie charts is not always bad.” And he’s right. Pie charts are not always bad. There are times when they’re ok. As Wade says “If you have 3-4 datapoints, a pie can effectively convey what one is intending to present.” Which is true. But in every case I’ve seen, those situations are as well served with a small bar graph.
What’s the least contrived situation in which a pie chart is better than a bar graph or table? (Pac man [link to http://www.hemmy.net/2007/12/01/pacman-pie-chart/ no longer works] and pies are two obvious examples.)
when talking to upper management. bar graphs are too “complicated” and it takes too much time to read the tiny-little numbers in a table and actually make sense of them.
upper management understands eating pie: they know how large each piece of the pie is, what a fair distribution of said pie would be, and they know they want the biggest piece.
I use pie charts, but very rarely. As you say, bar charts are usually better.
A pie chart is better when you need to represent parts of a whole and one value is much larger than the others. A real example (from memory so values are slightly off): the cumulative disk footprint of 5 types of files in a directory with relative sized of 71% (PPT), 13% (PDF), 8% (DOC), 6% (XLS), and 2% (other). A bar chart does not convey the same visceral experience of “Don’t even bother looking for solutions unless you deal with PPT files.” This specific chart was needed to dispel any myths that many of the files could be easily deleted – because nobody was going to go through all those PPT files.
There’s another use of pie charts, it can be wielded a bludgeon, as shown here
http://www.cxo.eu.com/news/madoff-auction/ [link no longer works]
to hammer home the difference between $65B stolen by Madoff and the $2M raised by the auction
While bar charts can convey relative values, there is nothing intrinsic in their visual design that conveys the notion that the sum of the bars makes up a whole, nor any demand on them that they do so. Only labeling provides that information in a bar graph. Compare, for example two bar graphis, one where there is a bar whose length demonstrates how much of the world’s population lives on a particular continent at this moment, and another which shows the population on one of those continents each year. In the first, the bars are parts of a whole. In the second, they are not.
This is neither good nor bad. The only point I’m making is that pie graphs actually do actually have visual properties that other graphs don’t. I think Adam’s point is that, most of the time, those visual properties aren’t really the point of the graph (instead, “show the relative sizes of these things” is).
IMHO … the only use for a pie chart is to show comparison of the various parts of something relative to the sum of those parts. If it isn’t something that makes sense to do a percentage calculation on then a pie chart doesn’t fit!
A good example … we’ve done a security assessment to determine our maturity level relative to the ISO 27002 standard. Of course a lot of the objectives/questions are about technology … mostly because it is the most complex part, not because it is more important. The rest of the objectives are around process, facilities, services, environmental, etc. To let management know how this all fit together I used a pie chart to show that about 57% of the questions were technology related. Something that you can see better in an “analog” pie chart then in a “digital” number like 57%.
And if you use two pie charts side-by-side to show relative comparisons of two wholes then you have to be very careful to adjust the area of the pie charts based on the ratio of the two wholes … and not, as I’ve seen many people do, just increase the diameter … that way B being twice as big as A would have and area (and a perception) of four times the size!!