Composition (part-to-whole)

Stacked bar and column chart

Up to 7 bars · at most 5 segments each

Use it to show how totals divide into parts across categories, for both the overall size and the contribution of each part. A 100% stacked bar compares proportions, and is the best choice when the data is already in percentages.

• Use horizontal bars for categories you can sort, and vertical columns for a time sequence.
• Avoid it when readers need to compare a segment’s exact value across bars. Only the first segment shares a baseline, so use a clustered bar or small multiples instead.
• Label segments directly, so readers don’t have to go back and forth to a legend.
• Keep the segment order the same in every bar, with the most important segment at the baseline.

Pie and donut chart

2–5 slices · at most 6

Use it for part-to-whole proportions of a few categories, especially data that is naturally expressed as percentages. A donut chart has room in the middle for a total or key figure.

• Avoid it when slices are similar in size (differences under about 5 percentage points are hard to see), to compare values across categories (use bars), and for trends or distributions.
• Start the largest slice at 12 o’clock and go clockwise from largest to smallest.
• Group small slices into “Other”, and place “Other” last.
• Label slices directly with the category and the percentage.
• Slices must add up to 100%.

Treemap

Up to 10 rectangles

Use it for how a whole divides into parts within a hierarchy, and to show which categories dominate a complex dataset.

• Avoid it when precise comparison matters, or when the data isn’t hierarchical. Use a bar chart instead.
• Label rectangles directly, and keep the labels short, because long labels don’t fit in small rectangles.