The hate of pie charts harms good data visualization…
How many articles on “how bad pie charts are” or “why should not use them”? On this platform only, if you type “pie charts” most of the entries will be against the use of those charts.
As we should do for everything, let’s stop having preconceived ideas and hate before questioning ourself on the matter.
I am not here to say if it’s bad or not to use such a chart, you can read about it here. I would like to change perspective, to see if we can have a better understanding of why they are so ubiquitous (despite the hate), why they work, and if we can better know how to use them from this thinking process.
The objective should be to better use them instead of limiting their use.
Why the hate?
The main reason why pie charts are the target of visualization critics is that they are everywhere! From TV news to web info-graphics and economic newspaper… It is the “by default” choice, or at least an easy-to-go solution, for visualizing multiple values. Therefore they are often misused.
If you use a default solution, probably you have not been thinking enough about your message or your question.
If you add to that the bad production, the wrong labelling, sum over 100%, the “fancy 3D” distortion (GOD! I hate when people do that) and a poor choice of colours, you got a pie chart indigestion (Buh dum, tsss !).
Everything mentioned above is a problem. And we should try to fix that by thinking more how and why we use pie charts, not by systematically limiting their use.
But many of people criticizing pie charts are also complaining about its intrinsic qualities. It is known that pie charts are made for proportions, yet many people attack them (in dedicated articles or in comments of articles defending pie charts) because they do not do well to compare absolute values. It’s true! But that’s not why we use pie charts! They continue by saying that it is easy to see if a group is 50% in other chart types (I should say that pie charts are pretty good for this) if you have proper axes and labelling, but you can also track other values. Also true, but quite often the message of a pie chart is to show one or two particular PROPORTION values (or rather “one against all others”), not to be able to make pairwise comparisons (changing bar charts into % does not change a damn thing). Let’s go beyond this unfair criticism and try to understand what makes them attractive enough to be that much overused.
Last thing before we go on this topic, that I have not mentioned yet: the criticism about how they are read. I will come back more in details onto that, but just to mention that too many articles attack the pie chart on the fact that it is mostly angles and area. First, they say that angles are very poorly processed by our brain, and it is true, but the angle is only one dimension of pie charts (as I said, I will come back to it later). Second, some people use the same argument but this time about the area. Once again “true”, processing direct comparison of two areas is not our cup of tea. It is the case for bubble charts (used as examples with reason in these articles to demonstrate their point), but they do not present differences in area the same way pie charts do. The way the dimensions of an object are presented impacts the way they are perceived.
In defence of pie charts
Despite all the hate towards pie charts, you can also find articles that defend the lovely shaped charts. Here are some examples of what you can find:
Why Tufte is Flat-Out Wrong about Pie Charts
Some critics have come crashing down hard on pie charts. Edward Tufte says “ the only thing worse than a pie chart is…
On how to use them properly: https://www.oreilly.com/ideas/in-defense-of-the-pie-chart and http://annkemery.com/pie-chart-guidelines/ and recently: https://blog.datawrapper.de/pie-charts/ from Lisa Charlotte Rost.
But maybe one the greatest defenders of pie charts is Robert Kosara (the author of the very good blog EagerEyes). What I like about what he does, is that he tries to understand and measure the performance of pie charts compared to other types of charts. In recent articles in tackled the question “how do we read pie charts ?”. And the answer is definitely not the angles (but maybe angles do count a bit). Robert does also discuss alternatives to pie charts with square charts. And according to him, they read well (IMO they are not as good (even if more precise) because they need a lot more of “computation” to be processed (and so, a lot more time), and are less appealing). However, it looks like Robert thinks (I might misinterpret his work), like people criticizing the angle/area dimensions of pie charts, that one chart type is read only by one dimension. I disagree with that statement and I think this one of the reasons why pie charts are particularly efficient and so omnipresent.
Let’s see now why we like pie charts and why they work sometimes.
Why do they work? The doughnut, the watch and the pie.
If some smart people take the time to defend pie charts, it’s because they work in some instances and are the best solution for visualizing proportions, quickly and potentially several at the same time. We know that, but the reasons why they work and are (over) used, are not always examined. I think understanding that could help us to better use them.
First, let’s see if we can see when they work and are used.
Pie charts are extremely popular within infographics, press, dashboard, powerpoint presentations and some data visualizations. Alone to show one proportion, in couple to show differences in proportion between two categories or in small multiples to show the variety of cases (on maps for example). In all these cases, they need to attract the reader’s eyes and give the information quickly rather than precisely. Most of the time the choice of pie chart works well because it’s appealing, it is self-explained and fast processed, and its encoding is efficient. Let’s examine these points:
The doughnut: the appeal
It is obvious to me that pie charts are appealing (even though it is hard to demonstrate/discuss). Their round shape is probably the first reason why we find them naturally appealing. It’s instinctive to perceive round shapes as soft, tasty, elegant… while pointy shapes and hard edges are perceived as colder, more dangerous… For this reason, if the pie chart reminds us the juicy berries (bonus: evolution argument to pie charts), it cannot compete against the doughnut chart. Indeed the doughnut chart does not even have pointy angles!
Also, no axis, tick, grid or number is needed. Only a good legend if the encoding is not self-explained (like in Arthur Buxton work above or any other pie chart using eloquent colours). It is only colour, and it is self-explained.
It addition to the self-explained encoding and appealing qualities, why are we so fond of pie charts? We might just be used to them. I think that if the choice is given between known and unknown solutions, we tend to pick the thing that we are used to instead of evaluating the different solutions.
The watch: the habit
So, another reason why we overuse pie charts is that they are effective and fast to process (even-though not always precise). And it might come from the fact that we are used to read them, we are trained. Every day, multiple times per day, we look at pie charts. I mean, not really pie charts, but watches. We might process them quickly because we read them like we read watches. I think that’s also why it helps to have the pie slice starting from the top, where the 00:00 is on a clock (even if doesn’t mean anything in the context of a pie chart), so we read the position rather than the actual angle.
(People wearing minimal dial watches are probably better than average people at reading pie charts. But I might over-thinking here)
But sometimes, even if they are harder to read, angles are useful. Indeed, some particular exact cases are easy for us to identify: quarters and halves. It’s because we do not look at them as angles, but as straight lines or positions (horizontal or vertical), things we are quite good at. Also, it is easy to spot small deviations from these perfect cases: slightly bent line can easily be interpreted as greater than 50%, or a non-perfectly horizontal line as 13%.
But this “position” dimension is only one way to look at pie chats, and all others give the same information.
The pie: the multi-dimensions
Pies are good because easy to make, look good, easy to share and both the pastry and the mixture are tasteful. Pie charts also work because multiple aspects match well together: they are consistent.
We, humans, are made to process quickly different clues about our surroundings and create a coherent image of reality from this information. At least that what I think about the functioning of the human brain.
The multi-dimensionality of pie charts is what first motivated me to write this post. I think too much criticism is focused on how pie charts are read and consider only one dimension. In addition to appealing aspects of its round shape, this chart type is characterised by a multiple encoding of the same information: angle, area, length of the arc, position (if align to the top). Consistency is key. That may explain why they work less well when they are reduced to a single dimension (arc only, or angles only). This is also the reason they stop working when distorted: consistency is lost and the brain cannot decide which dimension is the good one (who knows but the designer?). That’s why they work well and why you should not alter them: as said in the study of Robert.
And that’s why I like them (when they are well used and not distorted).
Take home message
Pie charts are everywhere because they work, and they work because we know how to read them, they are appealing, and they are quick to process. The consistency of their multiple dimension makes pie charts unique in the world of chart types.
Understanding why a chart form doesn’t work is helpful, but understanding why a chart type works is as important, if not more. Instead of telling you where not to go, it guides you to a better solution (and this is often more helpful considering the number of options you can have).
Don’t be dogmatic and look where good design solutions lie. Most of the time it won’t be pie a chart because of plenty of good reasons, but if it is the best solution, you don’t want to ignore it because of some king of dataviz snobbery. And this is valid for any kind of chart, any type of chart is created in the first place for good reasons, and we should acknowledge these reasons before banning a chart type from our design library.
Thank you for reading. Please tell me what is your favourite pie (chart) and if you agree with my analysis of pie charts.