Are pie charts half-baked?

Some couples argue about money, intimacy, religion, and things like that. But for me and my girlfriend, our most contentious topic of the month has been the value of pie charts.

Leila’s position is that any data that can be presented as a pie chart can be presented better in some other format. A leading faculty member in her department — a truly brilliant woman — agrees wholeheartedly. But I’m sure they’re wrong, for reasons explained elegantly by Bruce Gabrielle.

Gabrielle’s advantage #2 of pie charts, “Communicates parts-to-whole relationships better,” is the one I consider most important.

At a glance, you know a pie chart is splitting a population into parts.

Bar charts do not have the same meaning. You can signal to the reader the bars add up to 100%, by adding a column or an annotation. But this requires some extra mental gymnastics by the reader to understand the bar chart represents 100%. Nothing beats a pie chart for instantly communicating 100%.

I’m sure this is why pie charts are routinely used to introduce fractions in elementary school. The format is easily grasped by anyone who has ever divided up a pie (or a cake). Does any other type of graph connect so well to a common visual from everyday life?

While I lack Gabrielle’s experience in presenting data to executives, I do present data all the time. And while I rarely include pie charts, I claim that there is a time and a place for them.

About a year ago, I put together some slides about my website SingAboutScience.org for a potential collaborator. One of my points was that recent improvements in the website had not caused an appreciable increase in the fraction of website visitors who returned for multiple visits. A pair of pie charts helped make the point quickly and clearly.

pie chart

Or so I thought. When I described this example to Leila, she did not find it compelling.

At any rate, this beats arguing about money.

5 responses

  1. Friday links: in praise of “token” women, solving the two body problem, and more | Dynamic Ecology

    […] In praise of pie charts. Distance runner and scientist Greg Crowther of My Track Record disagrees with some scientific graphics gurus (and his own partner!) and says pie charts have their place. […]

    Reply
  2. Douglas Johnson Avatar
    Douglas Johnson

    Pie charts are fine when a pie is divided into two (or three or four) pieces and a look at the pie tells you which piece belongs to whom. They are pretty horrible when there are 20 slices and a reader has to discern which shade of magenta in the pie most closely matches the shade of magenta in the legend. In the example of new versus returning visitors, the two pie charts vividly show the similarity and would be suitable for viewing by an executive with a 15-second attention span. For presenting in a paper, however, the two percentage values are much more concise.

    Reply
  3. Le Manchot Avatar
    Le Manchot

    Hi Greg,

    I ran into your website after reading a bunch of your stuff on running/ultramarathoning. Thanks for your work/writing in this area, I find it very useful and informative.

    I could not resist commenting on the pie chart post as I fought this battle in academia and corporate R&D for over 30 years. I retired having lost the battle and resigned to continuous exposure to unsophisticated and information-sparse graphics, such as pie charts. The discourse continues (as you know) but I thought I might point you to a source that provides substantial reasoning as to why it is best to look for alternatives to pie charts: Stephen Few- an author and researcher who has written numerous books on the subject of the communication of quantitative information. Linked below is a recent post from his blog specifically about the effectiveness of pie charts:

    http://www.perceptualedge.com/blog/?p=1492

    In ancient times, as a career scientist (physicist), I took it upon myself to offer seminar courses in this area if for no other reason than to improve my own environment and enjoyment of reading the papers and reports of my colleagues. It was satisfying to see fine scientists become excited about optimal ways to graphically present their results and to see them apply their innovative abilities to this area.

    As our higher education system has evolved, the whole concept of teaching communication (as originally captured in the trivium (grammar, logic, rhetoric)) seems to have fallen by the wayside- and we see the product of this deficiency daily.

    Reply
    1. crowther Avatar
      crowther

      Le Manchot: Thank you for the comments and the link. It’s excellent that Stephen Few tries to use research data rather than intuition alone to address these issues, and his dissection of the Spence & Lewandowsky study seems on-the-mark. However, Few seems to acknowledge my main point in an article cited by Gabrielle (perceptualedge.com/articles/08-21-07.pdf). Here’s the relevant part: “Pie charts are not without their strengths. The primary strength of a pie chart is the fact that the message ‘part-to-whole relationship’ is built right into it in an obvious way. Children learn fractions by looking at pies sliced in various ways and decoding the ratio (quarter, half, three quarters, etc.) of each slice. A bar graph doesn’t have this obvious purpose built into its design.”

      Reply
      1. Le Manchot Avatar
        Le Manchot

        Hi Greg,

        Agreed, there are moments for pie charts. I have just found that those moments are not very momentous, i.e. the observation may be best stated in a text form or as a part of a more grand graphic that gives additional insight into the meaning of the data. For instance, your pie chart graphic above pertaining to changes in the ratio of new to returning visitors after ‘improvements’ in a website might be better communicated in a table (you have all the data for the table on the graphic, and would it not be more appropriate (and efficient) to communicate a two decimal place magnitude differential in a form that is simple and straightforward- like a table) or with a simple statement. If there were additional data to be communicated, such as, for example, any change in the origin of new visitors (state, country, etc.) with the new website, I would suggest that a table or a bar chart would be a better choice than a pie chart. I have found that if there is anything of substance and/or nuance to be communicated, a pie chart is typically of little utility.

        From a perceptual perspective I quote Few from his book ‘Show Me the Numbers’ (page 60) pertaining to ‘part-to-whole’ graphic communication and use of 2-D areas for such:

        “Our visual perception is not designed to accurately assign quantitative values to 2-D areas.”

        and this is amplified when trying to communicate quantitative changes in areas.

        But alas we are awash in a sea of pie charts and it is perhaps best to take them for what they are worth in each instance and to be prepared to re-imagine the data in another, more communicative, form when need be. This has been and continues to be my strategy. Thanks for the interchange!

        Bob Youngman

Leave a comment