I appreciate your follow-up response: a fair and balanced treatment for pie charts.

Yes, your unique bar of bar charts is better than an spatially misleading pie of pie charts, no contest.

If you & Chandoo embrace diversity a bit and even teach your readers how to make better pie charts, that might mean a few less really bad pie charts in the world, and those people who learned to improve their pie charts can mull over your message that bar charts are generally more powerful and accurate than pie charts.

Obviously, your best 2-D bar chart still is better than the best 2-D pie chart for visualizing almost all small differences. Even in your 2-D pie, I admit that I still can’t be certain without reading the numbers (or noticing the polar ordering convention!) which of the two slices was biggest.

Which brings me to my final point this week. Discover Card likely uses 3-D pie charts because the people who glance at a pie chart rather than carefully read their statements really don’t care about 1% or 2% or even 5% differences in their monthly categorical spending: they are satisfied with a quick colorful approximate superficial view of it. Which seems pretty practical to me.

We might save the pies for dessert . . . or we might save the good bar charts for when the details really matter. (And when the details really matter, we both know a dot plot is often *technically* a better choice than a bar chart anyway.)

I know I’ve been on the topic of bad pie charts a lot lately, and I’m worried about saturating my readers. At the same time, lately I’ve become hyper-aware of the problem, and I keep seeing examples to talk about. And people keep sending me other examples: maybe they like to set me off.

I’ll try to write about something interesting next week!

There are two conventions for polar coordinates. The mathematical convention starts with zero angle pointing to the right (the Cartesian X axis) and increases in the counterclockwise direction. The everyday convention, embodies in every common analog clock you encounter, starts with zero angle vertically upwards (the Cartesian Y axis) and increases in the clockwise direction. Since pie charts are the graphical opiate of the masses, and since Microsoft is Lord of the Desktop, we’ll use the second convention.

The colors have morphed from the originals, but sure, this one shows the relative angles pretty well. Supermarkets is well over 90°, medical is a little over (I think), the combined last four are perhaps slightly over 90°, and services is well under. The pie beats the regular bar at comparing how the sum of one group of adjacent data points compares to another. If we want to make other comparisons, like medical plus automotive vs. supermarkets plus gasoline, we have to reorder the chart, losing the help that sorting gives.

I think the mixed bar chart in my previous comment beats the pie at comparing the group of four to the ungrouped three.

And keep in mind I only grouped the four because the original chart distorted them. There may be no financial basis for combining them.

Here’s that last pie displayed using Conventional Mathematician’s Polar Coordinates:

What you came up with is the horizontal version of this, which is pretty useless:

What I was thinking of in my comment to ikkeman is this, which may or may not help the analysis: