Pie charts: a neverending discussion

We all know how found of pie charts Tufte is:

A table is nearly always better than a dumb pie chart; the only worse design than a pie chart is several of them, for then the viewer is asked to compare quantities located in spatial disarray both within and between charts (…). Given their low density and failure to order numbers along a visual dimension, pie charts should never be used.(The Visual Display…)

Stephen Few shares the same feeling:

(…) allow me to declare with no further delay that I don’t use pie charts, and I strongly recommend that you abandon them as well.(Show me the numbers)

Of all the graphs that play major roles in the lexicon of quantitative communication, however, the pie chart is by far the least effective. Its colorful voice is often heard, but rarely understood. It mumbles when it talks. (Save pies for dessert)

They have some followers, like this one or this one. On the other hand, Stephen Kosslyn discusses the results of several scientific studies to conclude that:

Clearly, we need to temper Edward Tufte’s (1983) assertion (based solely on his intuition, as far as I can tell) that ‘the only worse design than a pie chart is several of them’ (p. 178). In some situations, this opinion is no doubt justified, but we should not make such a sweeping generalization about the value of any type of display, independent of the type of data to be displayed and the purposes to which the display will be put.(here)

Of course, Kosslyn is the author of Elements of Graph Design, “a book confirming that psychology is for graphics like ornithology is for the birds” (Tufte dixit). Do I sense some sarcasm?

You will also find things like this, that could be seen as the typical graphic designer approach:

Are you still using the traditional 2D graph for your reports and presentations? (…) create 3D Pie Charts with Illustrator CS2 at ease. The Pie Charts looks so real and professional. (here, emphasis mine)

Ian Spence writes a more conciliatory approach:

In my opinion, much of the adverse criticism of the pie has come from those who have wished it to do more than it could. The pie chart is a simple information graphic whose principal purpose is to show the relationship of a part to the whole. It is, by and large, the wrong choice as an exploratory device, and it is certainly not the correct choice when the graph maker or graph reader has a complicated purpose in mind.(here)

[fusion_builder_container hundred_percent=”yes” overflow=”visible”][fusion_builder_row][fusion_builder_column type=”1_1″ background_position=”left top” background_color=”” border_size=”” border_color=”” border_style=”solid” spacing=”yes” background_image=”” background_repeat=”no-repeat” padding=”” margin_top=”0px” margin_bottom=”0px” class=”” id=”” animation_type=”” animation_speed=”0.3″ animation_direction=”left” hide_on_mobile=”no” center_content=”no” min_height=”none”][Update: you must simplify your data to use a “simple information graphic”, and sometimes a simplified view is exactly what you need, according to Seth Godin. I agree with him, but there is much more to say about this.]

Now, let me rewrite this paragraph in light of my previous posts:

In my opinion, much of the adverse criticism of Crystal Xcelsius has come from those who have wished it to do more than it could. Crystal Xcelsius is a simple charting tool. It is, by and large, the wrong choice as an exploratory device, and it is certainly not the correct choice when the graph maker or graph reader has a complicated purpose in mind.

I would say that this is a classical divide between emotion and reason, a divide as old as can be. On one side we have pie charts, Xcelsius and many graphic designers. On the other, we have the scatter plot, Tufte, Excel. Don’t elude yourself, this discussion will never end. We could use some emotional intelligence, though.[/fusion_builder_column][/fusion_builder_row][/fusion_builder_container]

14 thoughts on “Pie charts: a neverending discussion

  1. Pie charts are one of the most commonly used tools, which is why they are useful: everyone understands the working of a pie chart at first glance. everyone can distinguish the shape of a “half”, a “third”, a “quarter” of pie without having to use much brainpower.
    however such charts loose much or their effectiveness as the number of slices rises. I spend much efforts in my organization trying to dissuade analysts to use pie charts with 10 datapoints, which is a blatant waste of time. A pie chart that shows just one datapoint, on the other hand, can be surprisingly powerful and really help make a point.
    finally I am very much against the use of 3D for pie charts. the use of perspective blurs our traditional perception of the angles and makes things confusing…

  2. Ever since I have seen and used treemaps [1], pie charts to me have no functional use, outside of maintaining the status quo. I would imagine as treemaps become more ubiquitous and understood, pie chart visualization will be relegated to the dust bin of visualization techniques.

    [1] http://en.wikipedia.org/wiki/Treemapping

  3. Jorge, this is a great visual to reinforce the message. Can anyone tell me by looking at the pie chart above, how the yellow and grey pie slices compare? Which is bigger? You wouldn’t have that problem using a y-axis bar chart. The things I can tell by looking at the pie: green is the largest and the parts add up to the whole. Great work Mr. Pie [hint of sarcasm]!

  4. I too am fond of conceding that a pie chart with only two slices is relatively harmless. But there’s a problem with conceding that they’re relatively useful, because the space available to them to be relatively useful does not extend all the way down to zero: at some point the usefulness of a simple pie graph drops below that of an equally simple table, or even a simple chunk of text.

    Who is unable to interpret the phrase “25%”, or “1/4”, or “3 to 1”? And if no-one who could interpret a pie chart with ease is unable to interpret such text with equal ease, what is a pie chart but pretty wallpaper?

  5. Why not use some science to answer this question? No one has mentioned Cleveland’s hierarchy yet, which reminds us that the perception of distance on a common scale is (much) more accurate than perception of angle or area (so applies equally well to treemaps too). If we want a graphic than can be perceived accurately, we should favour a bar chart (or even better, a dot chart) over a pie chart. While a bar chart loses the clear part-to-whole relationship of a pie chart, this can be included in a caption or with other graphical means.

    If you’re not aware of Cleveland’s pioneering work in this area, I would highly recommend that you look into it: W. S. Cleveland and M. E. McGill. Graphical perception: Theory, experimentation and application to the development of graphical methods. Journal of the American Statistical Association, 79(387):531–554, 1984.

  6. Jon –

    I am a little intrigued by the treemaps. However, I see some of the same pie chart problems with treemaps. Although, maybe it’s better than a pie, I still have difficulty judging the difference in scale between a verticle rectangle/square and one that’s horizontal. I end up tilting my head sideways to see if I can compare them, which I find inefficient and prone to neck muscle strains. I also agree with Derek that sometimes a simple table is sufficient!

  7. Jeff: thanks for the link. Like mine, is a balance view of the problem. By default I wouldn’t use multiple pies, specially as a map overlay

    jerjer: I updated the post to include a link to a post in Seth Godin’s blog that deals with too much data (also in pie charts). I like a pie chart with two slices when it is used as a visual clue to show a subset. At least one of the slices must be linked to a “proper” chart (in this case, the pie chart is more a “design device” than a real chart…

    Jon, Tony: Do you know that there is a (very basic) add-in from Microsoft Research Community to create treemaps in Excel? It’s called (surprise, surprise!). You can get very interesting results with a treemap if the dataset is right (high variability, for example).

    Tony, this is the default pie chart in Xcelsius…

  8. Hadley: Spence in the link above discusses Cleveland’s hierarchy, Simkin and Hastie findings and his own work. unfortunately there is no accepted scientific answer. There is some fragmented knowledge but the puzzle is far from finished.

  9. The wikipedia entry on pie charts points to Cleveland, and does try to take the tone that pie charts are to be avoided. My reading of treemaps is that they are just a little less cryptic than pie charts. I see pie charts as a necessary evil–like using a passive sentence in writing just to have some variety. As long as I am not making comparisons and don’t have more than a few slices, and the labels fit on the graph and not on a legend, and I have a bunch of bar charts in the work, then I will add a pie chart.

  10. Oh come on, there may be many reasons to prefer a pie chart over a bar chart or dot plot, but perceptual accuracy isn’t one of them. Can you find one peer-reviewed article that states otherwise?

  11. Oh, I see that Spece cite’s a few. It just doesn’t ring true to me – from experiments that I’ve seen pie charts are much inferior to pies – I find it very hard to accurately judge the angles. Maybe the difference is that bar charts normally have scales and pies don’t, but the experiments use scales for neither.

Comments are closed.