Click a slice to isolate it · Hover for values · Toggle donut mode · Max 5 slices
A pie chart encodes quantitative values as arc length and central angle within a circle that represents 100% of the total. The viewer's visual system judges proportion by angle — the size of the "pie slice" — which is a weaker perceptual channel than position on a common axis (used by bar charts) but stronger than area alone. The key perceptual advantage of a pie chart is that it communicates the whole simultaneously: the reader can instantly see that one slice is "about a third" or "just under half" in relation to the complete circle, without needing to mentally sum a column of bars.
This part-to-whole relationship is what the pie chart is uniquely good at communicating, and it is the only circumstance where it should be used.
No more than five slices. Human angle discrimination degrades sharply below ~15°. At six or more slices, multiple segments become visually indistinguishable without reading the labels — at which point the chart is doing no perceptual work, and a sorted horizontal bar chart would be faster to read. This implementation enforces the five-slice limit by collapsing excess segments into "Other."
Values must sum to a meaningful whole. A pie chart is categorically wrong for data that does not constitute 100% of something. Survey responses that allow multiple answers, non-exhaustive category lists, and continuous measurements are all incorrect uses. The viewer's implicit assumption — "these slices account for everything" — must be true.
Avoid multiple pies for comparison. Comparing two pie charts side-by-side requires the viewer to mentally hold two sets of angles and compare them — a task humans perform badly. Two grouped bar charts with shared axes are dramatically easier to compare accurately.
The data structure is a complete, exhaustive composition of five or fewer categories that sum to 100%. The message is a single-distribution story: one category dominates, and the chart communicates that dominance at a glance. The reader needs to walk away with one proportion lodged in memory — not to compare seven categories precisely. That is the pie chart's job description.
A 100% stacked bar chart would allow more categories, enable multi-period comparison (one bar per time period), and be more accurate for precise proportion reading. It is the correct upgrade when either of these needs arise. A donut chart (togglable above) is perceptually equivalent to a pie chart but opens the centre for a summary statistic — useful when one number (total, selected %, dominant category) needs emphasis alongside the distribution.
Direct arc labels — percentage values placed on or adjacent to each slice — are strongly preferred over a detached legend. A legend requires the eye to travel off the chart, match a colour swatch, and return. A direct label eliminates that round-trip. The downside is that very small slices (under ~10°) cannot carry a readable label without a leader line — this implementation uses centroid-positioned labels and suppresses them on slices below 5% to prevent collision.
FT Visual Vocabulary category: Part-to-whole — "How a single entity is made up of its components." Abela quadrant: Composition (static, few periods). Tufte principle applied: the circle perimeter is data (100% of the whole); every arc is data (each category's share). The only non-data ink is the thin white stroke separating slices — necessary for figure–ground separation, not decoration.