1) Whether Pie charts are useful
2) Whether it makes sense to group the smaller items into one category.

Suppose you had data about days lost to accidents like

Fan Heater Fires 45%
Plumbing leak 38%
Exploded fridge 15%
Broken Window 5%
Screwdriver ailment 1%
Handkerchief mishap 0.3%
Tripping over carper 0.08%
Misreading Pie chart 0.025%
Burnt Toast 0.02%
Other Broken Window 0.01%
Different Screwdriver ailment 0.009%
Handkerchief mishap Again 0.006%
Tripping over carpet type 2 0.005%
Misreading a different Pie chart 0.004%
Burnt bread 0.004%
Hitting thumb with hammer 0.001%
Overloaded wall socket 0.001%

Now, if you are just trying to work out how to reduce days lost to accidents then it makes a lot of sense to trim the ‘noise’ at the bottom. This is true regardless of whether you are using bar or pie charts.

Abuse them as much as you like, they remain evolutionary winners.

Rather than curse them with megaphones from the rooftops, why don’t we dissect them and figure out why they survive and even thrive in human society?

Some here claim that pie charts tend to confuse and obscure the truth. A few pies do, and those should be stomped out or perhaps collected in glass jars for our amusement, as you do so well, Jon. (Especially those silly slanted 3-D ones, possibly with pies of pies.)

Many here point out that pie charts don’t allow very good resolution or accuracy. That is correct, of course: that is not their niche.

Effective pie charts are like sound bites. They provide a simple, memorable summary of the situation. Even if an audience is distracted and has other things they’d rather focus on, the pie chart delivers its big picture message in a matter of seconds or less. Even a child or busy executive can figure out whether one slice is a lot bigger, a lot smaller, or about the same size as another slice. That same child or busy executive intuitively grasps that the slices of a pie chart must add to 100% of a whole, or someone is cheating. Who knows at a glance whether a given bar chart will bother showing all the pieces, or have multiple selections that add to more than 100%?

Need to provide a quick and easily understood overview of a situation? The humble 2-D pie chart just might work fine. If you want to impress your audience with deep information content or visual effects, try something else.

I have stated on this blog that pie charts can be appropriate, but in a very small number of cases. In Political Pie Charts I cite the pies used by FiveThirtyEight to show poll and election results.

These pies work because there are only two or three data points in the chart, and the third, if present, has a value that’s out of whack with the other two (in this case, the third is very small). The effect is to compare two data points, and a pie makes it easy to see which one is greater than half, and roughly how much greater.

In my opinion, pies are great,

(1) when there are <4 slices (and labeled well)
(2) when one slice dominate the hell out of other (like may be Google vs. other searches, Windows vs. other OS')

But they get most criticism because they are the easiest kind of charts to abuse.

Hmm. I guess it depends on what you consider to be “best”. If your intention is to confuse, obfuscate, cloud* the issue, then we may be in the pie business. Of course, donuts and three-dimensionality should not be discarded lightly, neither should radars (a clarity-denying tool that is sadly-underutilised by unimaginative information-hiders).

So, is there ever a good use for a pie chart? (You’re allowed to say “no”)

* (addle, becloud, bedim, befuddle, bewilder, blind, blur, darken, daze, dim, eclipse, mist, muddy, mystify, perplex, puzzle are also suggested by a quick thesuarus check)