Heatmap

Colouring a grid of values so patterns across two categories appear at once.

Also known as: heat map, colour matrix, colour grid

A heatmap is a grid where each cell is a value drawn as a colour, so you can see patterns across two categories at once — weekday by hour, product by region, feature by cohort.

Why it works

The same numbers as a table are a grid of cells the reader must compare one at a time. Coloured, the eye finds the hot block in a second. It is the right form when both axes are categorical and the question is “where is it concentrated”.

Rules

One sequential ramp for one quantity, light to dark in value order — see colour scales.
Print the cell values when the grid is small enough. Colour is the fast channel and numbers are the accurate one; a reader who needs “13, not 14” cannot get that from a shade.
Anchor the two ends of the scale deliberately and say what they are.
Keep a blank cell and a zero cell visually distinct. They mean different things, and merging them is a missing data decision disguised as a formatting default.

Sorting changes what a reader sees

This is the part people forget. Order the rows and columns by total value and you get a clean diagonal and an obvious “top-left is hot” story. Order them alphabetically and the same data looks like noise. The ordering is an editorial act and it is usually the strongest claim the chart makes — sometimes the only one. Say which ordering you used, and never put two heatmaps ordered differently side by side and invite the reader to compare them.

Trade-offs

A heatmap has room for exactly two categories and one value. Anything that needs segment plus time plus a third measure is a different form.

The colour blocks hide the distribution inside each cell. When the question is spread rather than level, a box plot per row beats it.
A row that looks hot may be one product with a huge total sitting next to three flat ones. Aggregation artefacts live here — see Simpson’s paradox.

When the grid has a natural third measure, small panels of the same grid often beat one big block: see small multiples.