Hall of Shame: Pie Charts That Should Have Never Been Baked
Why are pie charts so loathed by the data analyst community?
Pie charts have certain inherent problems:
- The quantity for each of the segments is represented by relative slice size, but the human eye isn’t good at estimating quantity from angles.
- It’s especially hard for us to tell the differences between small percentages (small slices).
- It is difficult to match labels to the appropriate slice, especially if there are over 4 slices.
Even though pie charts are almost universally disliked by data analysts, they are still often used in news reporting.
Below are examples of pie chart “fails” I’ve found online. I will analyze each one and tell you which mistakes to avoid in your pie charts.
1. Marijuana Pie Chart
This pie chart shows the results of a poll about marijuana usage in the U.S. It has different slices showing the percentage of Americans who have tried marijuana today vs. last year vs. 1997.
Why is this a fail?
- The pie chart adds up to over 100% (51% + 34% + 43% = 128%). Pie charts are supposed to show each segment’s part of the whole (adding up to 100%).
- Without more context, we do not know why the time categories chosen were today, last year, and 1997.
- It makes no sense to use a pie chart to show marijuana usage over 3 time periods.
- Misleading category label: It says that 51% of Americans tried marijuana today, which makes it seem like there was a special event I missed.
- Unless you know the date that the pie chart was posted, you won’t know what “today” or “last year” means.
2. UMass Amherst pie chart: “Poll finds Ron DeSantis neck-and-neck with Donald Trump”
In this national UMass Amherst poll, Florida Governor Ron DeSantis is running neck-and-neck with former President Donald Trump for the 2024 Republican presidential nomination. The full press release is here.
Why is this pie chart a fail?
- The pie chart has over 5 segments — way too many for the human eye to be able to tell the difference in size between the segments.
- The colors of the pie chart are too similar. Pence, Haley, Cruz, and Scott are all varying shades of light pink, which makes it harder to read. Carlson, Cheney, and Others have similar shades of dark blue.
- It’s hard to tell which labels match up with which pie segments. I counted 9 category labels and 10 pie chart colors — meaning they may have accidentally included an additional pie slice, or forgotten to label it.
3. “A Real Pie Chart”: What are your three most favorite types of pie?
In this nationwide poll, people were asked to rank their three favorite types of pie.
Why is this a fail?
- Each person voted for up to 3 types of pies. This makes it hard to tell who voted for what, or exactly how popular each pie is. We don’t know how each one was ranked.
- The total adds up to more than 100% (271% by my count).
- “Chartjunk” makes it hard to tell the difference in size between the pie slices. “Chartjunk” refers to the unnecessary visual noise caused by having real graphics of pies. For example, the “lemon meringue” and “pecan” pie slices are both supposed to be 24%, but the lemon meringue has a differently shaped crust that shows more of the pan.
4. TechCrunch: Market Share of Publishing Tools
This TechCrunch article breaks down the market share of Twitter clients on desktop, saying that the most popular way to use Twitter is through the Website, followed by TweetDeck.
Why is this a fail?
- It’s 3D, which distorts the perceived size of each slice, making the chart less accurate (and harder to interpret.)
- There are more than 10 slices in this visual. The best practice is to avoid using pie charts when there are over 5 categories.
- It is hard to tell the difference in size between the smaller slices. It is also difficult to figure out which category label corresponds with which of the smaller slices.
- There are similar colors being re-used — for example, different shades of light green, gray, and black.
Final Words on Pie Charts
If you must use a pie chart, be very deliberate about the message you want to send. Reduce any visual noise or “chartjunk” that distracts from your message. Make your visual as easy to understand as possible.
The more bewildering your data visual, the more likely your coworkers will be confused, have follow-up questions, and misinterpret the data.
If you want to keep up with me and my Shady Stats posts, follow me on LinkedIn! I post #shadystats every week, focusing on misleading ads and data visualization fails.
—
Additional reading:
