Steve Fenton

Can labels save pie charts?

I recently received this interesting question through the contact form. I always relish an opportunity to challenge my assumptions and opinions on pie charts – even though I appear to be a bit of a zealot. One day, I may even be convinced that I am wrong. Here’s the question:

“It seems to me that the telling difference isn’t the shape of the chart isn’t the problem, but the lack of labels on the pie chart. Compare a pie chart with percentages labelled with a column chart without percentages and wouldn’t you have similar problems?”

Good question. Here is my take on answering the question…

Ignoring how the charts are marked up momentarily (I will come back to that), the difference between a pie chart and a column chart is one of shape. The majority of people judge rectangles very well and triangles much less so. This is the fundamental problem with pie charts – they don’t play to our strengths.

In any situation where two charts are marked up equally well, the column chart (or some other alternative) is vastly superior. In fact, column-charts have natural positions for meta-data, whereas a pie-chart doesn’t. Because there is no natural place to place labels, they tend to be left out, or placed in a legend, or they float around the periphery with ugly lines to join them to the appropriate segment.

If we were to compare a labelled pie-chart with an unlabelled column chart, I feel that all we would prove is that labels are useful. That isn’t to say the pie-chart would necessarily win or lose, because that would depend on the question. For example, if the question was “what percentage of people voted for ‘blue'”, the label on the pie chart would probably answer the question the quickest (this is a win for the label, not for the pie chart).

To be scientific, I think it would be best to consider equally-well-labelled versions of a chart, or equally un-labelled versions of a chart – because if the chart is only good because of the labels, you have to wonder if a chart is required at all – perhaps a table would be sufficient. Whether a chart should be used and which type of chart should be used should be a considered decision. All too often, people just pick a pie-chart because they lack the judgement required to understand how the data should be presented. The pie-chart is the choice of the marketeer, not of the expert communicating important data.

If you want to join the great pie chart debate, please get in touch as I’m happy to answer questions and challenges if I have the time.

Written by Steve Fenton on