Heat Map

A heat map represents data as a grid of cells where color encodes the magnitude of a value. The two axes are usually categories or time buckets (day of week × hour, channel × month), and each cell's color shows the metric for that combination — darker or warmer meaning higher. It turns a dense table of numbers into an instantly scannable picture of where values are high and low.

Use one when you have a value measured across two dimensions and want to spot patterns, concentrations, or anomalies at a glance — peak activity hours, conversion by segment and device, correlation matrices. Heat maps shine when there are too many cells to read as numbers but the overall pattern matters. Avoid them when readers need exact values for every cell, since color is approximate; add labels in that case.

Match the scale to the data. For values from low to high, use a sequential scale (light → dark of one hue). For data that diverges around a meaningful midpoint (e.g. above/below target, positive/negative), use a diverging scale with a neutral center and two contrasting hues. Avoid rainbow scales — they create false boundaries and aren't perceptually uniform — and check the palette is colorblind-safe.

A heat map arranges cells on a fixed grid defined by two axes and uses color for magnitude, so it's about a value across two dimensions. A treemap nests rectangles sized by value to show hierarchy and part-to-whole. Use a heat map for a matrix of one metric over two dimensions; use a treemap for hierarchical composition where size shows share of a total.

Add them when precise values matter and the grid is small enough that labels fit without clutter — color then gives the at-a-glance pattern and the numbers give exactness. Omit them on large grids where labels would be unreadable, and rely on color plus a tooltip on hover. A clear legend mapping color to value is essential either way.