Heat map

A heat map (also written as heatmap) is a way to show data in two dimensions, using color to indicate how big or small each value is. The color can vary by shade or intensity. In fields like crime analysis or website click-tracking, color sometimes shows how densely data points are clustered, rather than a specific value tied to each point. Although the term "heat map" is fairly new, people have been shading matrices to show patterns for more than a hundred years.

The technique began with two-dimensional displays of numbers in a data matrix, where larger values appeared as small dark gray or black squares and smaller values as lighter squares. The earliest known example comes from 1873, when Toussaint Loua hand-drew and colored a shaded matrix to show social statistics across Paris districts. In 1899, Flinders Petrie introduced seriation—reordering rows and columns to reveal structure in a matrix. Louis Guttman developed the Scalogram in 1950 to order binary matrices and expose a one-dimensional scale. Peter Sneath in 1957 displayed cluster analysis results by permuting rows and columns to group similar values together. Robert Ling implemented this idea in 1973 with a computer program called SHADE, using overstruck printer characters to create gray shades, one character-width per pixel. Leland Wilkinson created the first computer program for cluster heat maps with high-resolution color graphics in 1994, called SYSTAT. The Eisen et al. display replicates that earlier SYSTAT design. Software designer Cormac Kinney trademarked the term "heat map" in 1991 for software showing real-time financial market data. SS&C Technologies, Inc. acquired the trademark in 1998 but did not renew it, so it was annulled in 2006.

There are two main types of heat maps: spatial and grid, plus over ten other varieties. A spatial heat map shows the magnitude of a spatial phenomenon as color, usually overlaid on a map. For example, temperature appears as a color range from blue (cold) to red (hot) across a world map. A grid heat map displays magnitude as color in a two-dimensional matrix, where each dimension is a category of trait and the color represents a measurement combining both traits. One dimension might be year, the other month, with temperature as the value—showing how temperature changed over years in each month. Grid heat maps split into two subtypes: clustered and correlogram. A clustered heat map, like the monthly temperature by year example, groups similar data. A correlogram is a clustered heat map with the same trait on both axes, showing how traits interact; it appears as a triangle because A-B and B-A are the same. In a grid heat map, cells are fixed in size and shape, and the goal is to detect or suggest clusters. A spatial heat map, often used on maps or satellite imagery, has no cells—colors vary continuously.

Heat maps simplify data and make analysis visually appealing, so they have many uses. In business analysis, they visually represent a company’s current functioning, performance, and areas needing improvement, helping teams and clients see growth and efforts. On websites, multiple heat maps are often combined to find the best and worst performing page elements. Mouse tracking (or hover) maps show where users move their cursor. Eye tracking maps measure where users look, including fixation volume, duration, and areas of interest. Click tracking (or touch) maps visualize where users click, covering both clickable items like buttons and non-clickable areas. AI-generated attention maps predict where a user’s attention will go on a webpage using a software algorithm. Scroll tracking maps show scrolling behavior and indicate which sections users spend the most time on. Heat maps are also used in exploratory data analysis for both small and large datasets.