pie charts are fine and you all need to calm down

I mean, yikes.

So look. I’m obviously no designer. And I agree that the pie charts in that blog post, which are clearly cherry-picked for maximum terribleness, are terrible.

But usually pie charts are fine. I’m not going to bother collecting examples that demonstrate this, because unless you’ve led an unusually brief or interesting life, you’ve seen thousands of them and understood them perfectly well.

But I realize this isn’t going to stop you. You have been to the Seminars. You have received the Texts. You have studied the words of the Prophet Tufte, and you have concluded that they demand jihad. I’m not going to convince you otherwise.

I’ll make my case anyway.

To me, pie charts are handy for conveying information about the relative components of a whole because they do so through visual cues of both area and angle. If you need to show the relative sizes of multiple wholes, you can do that by varying the pies’ size! Fun times, but I suppose it’s not for everyone.

But I would like to request that you not join the pie-haters I sometimes see advocating for donut graphs or treemaps as alternatives. These have all of the problems of pie charts, but they throw away one of the two visual cues. (Treemaps can be useful for expressing hierarchies, and donut graphs are useful for… er. Well. Sometimes you can put a number in the middle of them!)

If your graph is confusing, it’s probably a bad graph and you should try to make a different one. But let’s not pretend that there’s a Platonic Graph that we’re working toward just because some dweeb in an eye-tracking lab eked out a few milliseconds’ worth of statistical significance. Real graphs used for Science usually look like this (to be honest, even that red line is a bit decadent). It’s not great, but it gets the job done. Science is pretty much going fine despite a lot of shitty-looking graphs.

If you ask me (you haven’t) presenting information is about tradeoffs. I don’t really want to spend all day looking at things that are ugly, and that’s okay. Nicely designed graphs are pleasant! People like to look at them, which is important if you want people to look at your graphs. Even if you really think pies are inefficient, you’re kidding yourself if you think graphs are always purely made to shoot numerical truth into your readers’ brains (when was the last time you adjusted the stroke weight on an error bar?).

People have varying tastes. That’s a totally legitimate rationale for hating pie charts. And they’re not the best choice for everything. I’m on board with that, too.

(Building part of your self-conception around the fanatical endorsement of the work of the one widely-known name in information design is, I would suggest, a little sillier. But whatever floats your boat. I went through a big anime phase in college, for instance.)

Anyway: pie charts. All I really ask is that you please shut up about them.

10 Responses to “pie charts are fine and you all need to calm down”

  1. The best sentences we read today

    […] “You have studied the words of the Prophet Tufte, and you have concluded that they demand […]

  2. The best sentences we read today - Badly Spelt Words

    […] “You have studied the words of the Prophet Tufte, and you have concluded that they demand […]

  3. The best sentences we read today | TwentyfourSevenHeadlines: Breaking Local, National, World & Sports News

    […] “You have studied the words of the Prophet Tufte, and you have concluded that they demand […]

  4. Jeff

    This is a terrible rebuttal. Pie charts are almost never fine (although sometimes, as discussed in the other blog post). Personally, I think Tufte is overrated–I’ve enjoyed all his books, but I don’t consider him a prophet.

    I’m sure you, and many others, believe that you have understood perfectly well thousands of pie charts that you have seen, but you haven’t. Your ability to extract information from pie charts is severely limited, but outside of your awareness.

    If you wanted to convey “information about the relative components of a whole”, you can do that with a stacked barchart. People’s ability to compare areas is not very good, but will be much better in this situation because the shapes of all the areas are more similar than pie charts and the same width. People’s ability to estimate angles is substantially worse than their ability to compare areas, but people’s ability to assess linear distance (either vertical or horizontal) is quite accurate. So you would still get two visual cues, but both would work better than pie charts. There is no situation described in your post, nor can I think of any, where a stacked barchart could not do what people want from a pie chart and do it better. For example, “if you need to show the relative sizes of multiple wholes, you can do that by varying the” bars’ sizes.

    You certainly make a good point that donut graphs are terrible (worse than pie charts, actually), but treemaps are superior to pie charts. The linked graph at the Monkey Cage blog is quite good, in my opinion. I don’t argue that there is some single ideal graph, but for any situation, there will be some graphs that serve better than others; I challenge you to find one case where a pie chart would be best–I find it noteworthy that you don’t “bother” to show any.

    Note that none of my reasoning, pro or con, about any type of graph is based on Tufte’s “endorsement”.

    By the way, what is “the stroke weight on an error bar”? I’ve never heard of that.

  5. William Mello

    It’s not a rebuttal at all: “It’s a matter of taste” and “Please shut up” don’t respond to any of the arguments in the original article. Since Manifest Destiny seems deliberately unconcerned with data, I will concede that it probably doesn’t matter whether she uses pie charts or not. The rest of us would do well to follow the advice in the original article.

  6. Brandon

    points taken about the pie charts- not a fan myself, I think they are terribly space-inefficient compared to a horizontal-aspect stacked-bar, but fine. I do object to your characterization of the proffered objet d’art for “real graphs” for science — I think it’s a quite splendid, economical and immediately comprehensible example of a professionally-produced scientific chart. Tufte himself, I think, would not frown upon it. Science may indeed be “going fine despite a bunch of shitty-looking graphs” but that is not one of them.

  7. Marc P.

    Let’s say you are thinking about an issue coming up for vote in the European Parliament. Can you assemble a winning coalition if you get all of the votes of the four lefty parties: EU, S&D, Greens-EFA, and ALDE?

    Provided the slices are ordered along the left-right ideological spectrum, a quick glance at the pie chart tells you that no, you’ll fall just short. If the information is presented in a table, you’ll need a calculator and thirty seconds to get that answer. If all you’ve got is a bar chart, you’ll need a ruler and a lot of patience.

    It’s just not true that “the chart is only useful if we’re able to compare each and every element within it.” It depends on the question you’re trying to answer.

  8. virgin unrefined coconut oil

    Just what is the number one ingredient in those chips?
    Cancer has been shown to raise serum cholesterol and cause heart disease because inflammation is a contributing factor to
    these conditions. Due to caprylic acid, capric acid,
    antioxidants, vitamins, and amino acids. Virgin coconut oil
    is generally used in making skin and hair, and nails due to its cold-pressing extraction
    method. 1 Virgin coconut oil is high in saturated fat,
    coconut oil should be recommended as a treatment or medication,
    and you will start gaining weight again. Regular users of this oil will give you
    great benefits.

  9. Gg-Organs.Ru

    How often oil pulling are regular checkups needed?
    I have been looking for tartar buildup. This
    problem shows in difficulty with extreme hot or cold temperatures.

Leave a Reply