I think the example we put in wikipedia gets to the message--the bar chart rules over the pie chart, especially when the categories are close to 1/5th of the area. Some people like to see pie charts, or rather, a pie chart. Think of it as a variety in your sentence structure. Sure, you should not end a sentence with a preposition, but sometimes that is up with something you can put.

If anyone wants to start a draft of Dot chart in wikipedia, the position is open.

Very interesting. In fact, I think I have never used a pie chart in my life, but I did not have anything against it. Now I understand why it never seemed to me a good way to explain data.

Very insightful post.

I feel sorry for this little bastard, so let me play the Devil's Advocate role (it is not easy) .

People conveniently forget that some research (e.g. Spence & Lewandowsky, "Displaying proportions and percentages") actually shows that pie charts perform better than bar charts in some cases. Also, a pie chart has more anchor points (25%, 33%, 50%, 66%, 75%) than a divided bar chart, so it easier to estimate values.

Pie charts are in general misused. People ask them more than they can deliver by nature (they can't deliver much, but they are humble enough to recognize that). The difference between boys and grownups is that grownups know how/when to use them.

The *real* problem with pie charts is that Tufte doesn't like them. And he doesn't like them because they don't fit in his aesthetics agenda. Tufte's framework is just that, a framework, not the 10 Commandments ("Thou shall not use chartjunk..."). And you can use another one that fits your taste better. (This is a slippery floor, and I like Tufte, but sometimes his over-positivism seems to fall into Flatland.)

Also, it doesn't matter what you think and what your personal tastes are. What matters is what your audience thinks and likes and understands.

Perhaps you should ask your readers examples of best practices using pie charts. I am sure there are good examples.

The post shows so much Christmas rage :)

Now I risk to be instinctively slapped (and that seems sooo Tufte-esque...)

My role as Devil's Advocate ends here. In general I agree with the post.

[...] We hebben het er al eerder over gehad, de pie chart staat onder discussie. Juicy Analytics heeft gister nog een aantal nieuwe argumenten aangehaald, waarom de pie chart uit onderzoeksrapporten moet verdwijnen: [...]

I suppose there's a 12-step program for pie chart users; first step is admitting you have a problem -- and I apologize to all the readers of my pie charts in the past. Stacked bars (or tables!) are preferred in most cases. There may be at least a couple of capabilities of pie charts, though - 3-way comparisons, for one, and grouping by color/shading/proximity.

An example that combines these: a pie chart describing the composition of a legislative body with three major blocs, each of which comprises 1-3 political parties. A pie chart could display relative size of each of the parties, differentiate the blocs by using related colors within each bloc, and allow for fairly easy comparison of the size of blocs - the stacked bar could do the color grouping, I suppose, but the three-way size comparison might be handy to do with the pie chart. Just a thought.

I've been pie-chart free for six months, now.

I agree to an extent, but pie-charts are a form of chart that virtually everybody understands and can read.

I do personally prefer bar, colum and line charts wherever possible but I do also have pressure placed upon me to try 'mix up' the way charts look. For instance when viewing 6 reports over 2 pages people push for 'variety' in the presentation method. So even if say some basic bar-charts would be a better representation for the data they want variety. I think this reasoning, plus the fact that 99% of people I know/work with understand how to read a pie-chart this makes them worthwhile.

The more advanced-charts shown in this blog are fantastic but many of them require a lot of 'fiddling' to get them working (at least on the ones i've tried). The basic-charts that Excel give allow me to integrate auto-update and auto-manipulation to ensure that report production is rapid and reliable.

Now if I could find a way to implement these wonderful charts into my routines then I would be a happy man indeed, but working in a business with very standardised software/multi-computer use etc. and that becomes a little tougher. So whilst I agree they are maybe not the perfect thing, there is justification for their use.

Paul
There is a serious problem of graphicacy in the corporate sector. The managers are not aware of that because they are very illiterate themselves (information visualization-wise). They think that a general-purpose Excel training is enough to create some charts to put on a report, just to embellish it, not for anything serious (decision making, for example).

People like pie charts because they are linked to the everyday life experience since one can remember. People don't like scatter plots because they are too abstract and don't fit in the daily routine. This is solved by education and training only.

If your audience asks you for 'variety' perhaps you could quote Druker ("good management is boring").

We have larger and larger databases, but we look at them through pie charts... It puzzles me.

"Heard about the guy who fell off a skyscraper? On his way down past each floor, he kept saying to reassure himself: So far so good... so far so good... so far so good." (La Haine)

As far as I'm concerned, a pie chart may still be the best tool for the job when you need to show the relative contribution of parts (the slices) to a whole (the pie). Bar graphs don't cut it because they fail to explicitly convey the sense of the whole; they only show you the parts. You could replace a pie chart with a single stacked bar, but I only think that's necessary in instances where the slices in a pie chart would be too similar in size for the viewer to judge the differences.

The litmus test in deciding whether to use a pie chart or a bar chart shouldn't be the data-to-ink ratio or a knee-jerk admonition to avoid using "the dreaded" this or that, but rather whether the chart you've prepared communicates effectively. If you've got a pie chart with a zillion slices and you can't tell what bits are more important than the others, then don't use a pie chart. But if your data offer just a few slices that are clearly differentiated from each other, and if your point is to show how a whole divides out into its major components, the pie chart might communicate more clearly than anything else.

The problem with Brad's final sentence, as I see it, is that he's given the pie chart such an absurdly easy job to do that it stops making sense to use any graphical device at all. Why not just use a sentence or a table?

The point of resorting to info graphics, to me, is to communicate effectively <i>beyond the point where communicating with words and numbers alone is no longer effective</i>. And that is almost exactly the point where pie charts stop being effective.

The key question is: what is the purpose of the chart - what message is to be brought across.
I always learned pie charts are good for showing the distribution of a whole - the wh9le (circle) representing 100%. It is clear that when a large number of sections are necessary, a pie chart is meaningless - or at least difficult to read. Also, as in Zach's example of GA, with only a few slices there is no point using a graphical representation. I would say, empirically, that a division into say 4-8 sections could have a point for pies when relative distributions are important (% - e.g., market shares).
The bar chart is superior when the relative weight of distribution is the key point to make. A is larger than B, and the length of the the bar does a particular good job here, where the angle of the section (or surface for that matter, but I usually "see" the angle) is difficult to inteprete.
The stacked bar chart is to me about the same as the pie. It appears difficult to see the relative lengths in comparison; its only advantage over pie chart seems the lesser space it takes.
In conclusion: in almost every case the relative weight of the shares is what is of interest, even in classical cases where the books say a pie is indicated. E.g., what means a market share of 25% for A in itself? The relative position is important. The 25% has a total different meaning in case there are two others with respectively 65% and 10%, in comparison to the case that 10 companies fight for market share, A being #1, and B #2 with 17%. For this reason I believe the bar charts are more useful. But maybe this is just a personal preference.

I fully agree with Henk. And if you show pie chart on the web you can use different roll over effects, pull out on click and other... I would be glad if google analytics could use these pie charts - personly I like them very much: http://www.amcharts.com/pie/examples/1/

[...] Juice Analytics has an interesting article on pie charts and how they can sometimes break the basic rules of data visualization. There are some interesting points worth noting. I&#8217;ve personally always been a fan of line graphs and bar graphs, though pie charts have there times. [...]