An Exception: Before finance and econometrics, I did a lot of public health work. The kinds of charts shown by Jeff Clark — which remnd me of radial heat maps (pardon the pun) and others ARE intuitive and useful, but only in my view to time-constrained technical, scientific, or field practioners such as physicians. It would be poorly applied in a business setting. I found they were best done with call-outs to the underlying text. Done this way they convey a lot of information quickly, and better, help guide the user to accompanying text without forcing them to digest the entire text. That newsletter, later taken over by the CDC, received considerable praise from practizing clinicians with little time to wade through text.

Jorge & Tim – Thanks for the link, I want to make sure I give credit where credit is due.

http://people.howstuffworks.com/prison4.htm

Perhaps they need a feature on “How Math Works.” :-)

BTW, there’s a really nice column chart on the same page.

http://people.howstuffworks.com/prison4.htm.

Jon, many people say what they need is a graph type that makes the 100% nature of the summed segments clear. 100% stacked bars do that, but suffer from some of the same problems as donuts. Your Excel panel graphs, as demoed elsewhere, are the ideal compromise: they stack the whole, but they also align the bases of all the parts.

Or, you could be silly like me and stack them back to back :-)

Ha, and I even commented on it. I knew I’d gotten the link from somewhere.

As I said, I don’t think it does anything that a simple pivot chart cannot handle, and a pivot chart doesn’t distort the display by forcing it into a circle.

While we’re on the subject of pies not adding to 100%, how about this chart which compares recidivism rates based on the original crime:

I don’t recall where I found this gem. If anyone recognizes it, please let me know.

I remember your chart now. I see what you were trying to do with the outer concentric arcs, but I have to say I think it is still harder to digest the data in this form than in a table. Maybe it’s a lack of familiarity with the thought process that went into it, but it takes too long to move one’s eyes back and forth, and try to internalize what the arcs are trying to communicate.

We shouldn’t stop trying to repurpose old tricks, even poorly performing ones, if we can make them perform new tricks.

I’ve come across a couple of other pie/donut variants lately that may merit full posts at some point, but now you’re making me comment about them :-)

The first pie/donut variant is a new way to display poll results, the visual output of a new computer package that allows users to change categories in their data easily. This is the brainchild of two researchers at the University of Utah, Geoff Draper and Richard Riesenfeld. I don’t have many details, just a press release, but it seems like it has all the functionality of pivot tables, without the “nice graphics” of pivot charts.

The second pie/donut variant is a sequentially broken down pie chart that becomes, in effect, a radial treemap. Jeff Clark of Neoformix has posted some of these, and I have inserted a couple below.

After doing this work, Jeff discovered that there is nothing new under the sun, or rather, under the SunBurst. The following is an illustration of hard drive usage generated by a program called Sunburst.

I’ve tried to exemplify how to create better pie charts, and that basically means using donuts charts this way.