I would say there is a place for pie charts. The article seems to have accuracy as ultimate goal, if you need that accuracy; fully agree, pie chart isn't the best visualization. But, data visualization is often about story telling, and the small percentage differences don't always matter.<p>For example, if you are telling a story about how one of the slices is much bigger than the other ones, pie (and donut) charts, are a very effective and visually interesting way to tell that story. The other case where I like pie charts (and I do prefer donuts btw) is when the data isn't very accurate and "hiding" some of the detail is actually a better representation of reality.<p>So yeah, pick the chart that works best for the situation (and if possible, give multiple options), but I do not agree with writing off the whole chart type. Radar charts are more questionable imo haha.<p>Data visualization is in essence trading accuracy for readability, how much you want and need depends on the goal, audience and data.
I had a college Statistics professor who would get triggered and go off on a (justified) rant every time he saw a 3D pie chart.<p>I am a big advocate or arranging the pie slices in order (largest to smallest, or smallest to largest).<p>I also really recommend <i>The Wall Street Journal Guide to Information Graphics: The Dos and Don'ts of Presenting Data, Facts, and Figures</i> <a href="https://www.amazon.com/Street-Journal-Guide-Information-Graphics/dp/0393347281/ref=sr_1_2?keywords=Donna+Wong&qid=1644268998&sr=8-2" rel="nofollow">https://www.amazon.com/Street-Journal-Guide-Information-Grap...</a>
The first example undermines the argument. It’s a demonstration that the author doesn’t understand a tool.<p>That pie chart tells me that the slices are approximately the same, which is a useful message to effectively deliver visually. Pie charts are great for understanding relative value.<p>Divining or comparing the precise values is not a good use for it.
Pie charts let you tell percentages at a glance, which is a good thing to be able to tell. If your data is not dividing up a pie, a pie chart might not be the best fit, and if you need to tease out questions like "which is the biggest" given several close numbers, it definitely isn't a good fit. It especially helps when you order the slices in a hostile way - if it was otherwise confusing, just order them clockwise by size instead of going with the (irrelevant in this format) data order.<p>Any chart format has a set of data for which it's a terrible fit. Different formats have
different information optimized to communicate at a glance, and if you're deliberately using the wrong one, you'll end up with a bad graph; that's not a takedown of the format itself.
As someone who has fought this battle many times:<p>People don't want accuracy.<p>People don't want details.<p>People want pretty.<p>They will reject any other type of plot that is better than pie charts on the other two metrics.<p>Hence, pie chart.<p>The higher up you go, the more important the appearance is, and the less important the details are. The role of most presentations is not to get people to understand, but to impress. Senior folks have given feedback that "Your presentation slides don't have enough details. If you make it easy for the audience to understand, they will undervalue your work."
Pie charts are useful when there are large disparities between some of the data items. Like, here are our total expenses for last year, we spent 6% on administration, 14% on facilities, 11% on R&D, 33% on manufacturing, and 36% on sales and marketing. You can see at a glance, even with human eyesight's poor judgement of areas, what the dominating two expense areas are. Or, you know, here's a breakdown of the OS our dev team members use: 85% use Windows, 11% use macOS, 3% use Linux and 1% use "other".<p>Oftentimes they will be labelled as well with the exact percentage numbers which helps. It's not really a scientific visualization tool, but it adds punch to a presentation when you want to show that one or more subsegments of a whole really dominate the rest.
This article is weird. It centres on the central premise that pie charts are useless because you can't tell exactly the relative sizes of slices that are very similarly sized. That's not the point of a pie chart, which is to show clearly which areas are big and which are insignificant.<p>The article then poses the question "Let’s See if it actually rains more on the weekend." Looking at the pie chart, it's immediately obvious at a glance, that YES, it does rain more at the weekend, as the two segments for Saturday and Sunday together account for nearly half the pie. Clearly the pie chart is perfect for answering this question. But then, the article launches into a whole spiel about how pie charts are useless because you can't tell which was rainier between Saturday or Sunday.<p>However, if you used an alternative presentation e.g. the lollipop charts as they suggest, then it's NOT obvious for answering the question of whether the 2 weekend days rained more or less than the 5 other days. Rather it's benefit is in determining which days were the rainiest.<p>The obvious take-home is to use the appropriate graph to illustrate the conclusion you're trying to make from the data, but also include the raw data so that others can analyse it if they think there might be other details that are important.
The author says the problem is that we have a hard time estimating areas. That is hard, but the problem isn't estimating the area. What you is segments of a circle's circumference. You can unroll a piechart into something like this, and it's still kind of difficult to figure out the proportions:<p>|----------|--------|---------------------|--|-----------|
<a href="https://www.perceptualedge.com/articles/visual_business_intelligence/save_the_pies_for_dessert.pdf" rel="nofollow">https://www.perceptualedge.com/articles/visual_business_inte...</a><p>- Stephen Few, August 2007
Yes. Yes. 1000x yes. Pie charts are bad and there is almost no justification for using them. Humans <i>are</i> bad at estimating area - and donut charts help fix this. I actually pretend donut charts are "curved line charts".<p>Also, engineers are notoriously bad at making visualizations. Sorry if this offends you. I would recommend <i>everyone</i> here spend some time looking at pretty visualizations and maybe reading about it