The pie chart is the most popular chart type in the world and the least effective. This is not a new observation. William Cleveland and Robert McGill demonstrated in 1984 that humans judge angles and areas far less accurately than positions along a common scale. Jacques Bertin made similar arguments decades earlier. The evidence is overwhelming, and the world ignores it.
Why? Because pie charts feel intuitive. They represent parts of a whole using a metaphor everyone understands — the sliced pie. The problem is that this intuition is misleading. When a reader looks at a pie chart with five slices, they cannot reliably determine which of two similarly-sized slices is larger. They cannot estimate whether a slice is 23% or 28%. The encoding defeats the purpose.
The Perception Problem
Cleveland's hierarchy of perceptual accuracy, refined through decades of research, ranks visual encodings from most to least accurate:
Pie charts use angle and area — the bottom of this list. Bar charts use position and length — the top. This is not a matter of taste. It is a measurable difference in how accurately information is transmitted to the reader.
When Pie Charts Work
And yet. There are exceptions.
A pie chart with exactly two slices — showing a simple majority/minority split — works. The reader needs only to see that one part is larger than the other. A chart showing 73% vs. 27% communicates effectively as a pie because the judgment required is trivial.
A pie chart can also work when the exact values are labeled directly on the slices and the chart serves a purely rhetorical function — showing that one category dominates. In this case, the chart is not asking the reader to decode angles. It is using the circular form as a familiar frame for already-labeled numbers.
These exceptions are narrow. They apply to perhaps 5% of the situations where pie charts are currently used.
The Donut Variation
The donut chart — a pie chart with a hole in the center — is worse than the standard pie chart. It removes the central angle vertex, making angle judgments even harder. The only advantage is the center space, which designers fill with a large number. But if the goal is to display a single number, display a single number. The donut adds nothing.
What to Use Instead
For parts-of-a-whole comparisons, a horizontal stacked bar is almost always superior. A single stacked bar communicates the same information as a pie chart but uses length rather than angle. For more than three or four categories, a simple bar chart sorted by value is clearer still.
When the number of categories exceeds six or seven, no single chart type works well. Consider a table. Tables are often the most efficient way to display exact values for many categories. This is one of the arguments Tufte has made repeatedly: when the data is the message, show the data.
The Software Problem
Excel, Google Sheets, and most business intelligence tools place the pie chart prominently in their chart menus. Many make it the default recommendation for categorical data. This is a design failure at the tool level. It trains millions of users to reach for the wrong chart type.
Datawrapper, notably, includes a pie chart option but adds a warning when more than five categories are used. This is a responsible approach. The tool acknowledges user expectations while guiding toward better choices.
The goal is not to ban pie charts. The goal is to make them rare — used only when the specific conditions that make them effective are met, and never when a bar chart would serve the reader better.