Heatmap
Also known as: heat map, click heatmap, scroll heatmap, correlation heatmap, grid chart, density heatmap
A heatmap is a data visualization that represents numerical values as colors in a two-dimensional grid.
Higher values are typically shown in warmer colors (red/orange) and lower values in cooler colors (blue/purple), with intermediate values in a gradient between. Heatmaps excel at revealing patterns across two categorical dimensions simultaneously: website engagement by hour of day and day of week, sales by region and product category, correlation values between many variables. The color encoding makes patterns immediately visually apparent -- a cluster of red cells in a blue grid is immediately obvious -- without requiring readers to process individual numbers.
Heatmap, explained at three levels
At home
A heatmap shows numbers as colors -- darker or more intense colors for bigger numbers, lighter colors for smaller numbers. Imagine a grid showing website visits by hour of day (across the top) and day of week (down the side). Each cell would be colored by how many visits occurred at that hour on that day. The grid would immediately show you that Monday morning at 9am is bright red (lots of visits) while Sunday at 3am is pale blue (very few visits). You can see the whole pattern at once without reading individual numbers.
At work
Heatmaps appear in business contexts: website analytics (which areas of a webpage users click on or scroll to -- click heatmaps and scroll heatmaps), sales performance (revenue by salesperson and product), scheduling (customer volume by hour and day for staffing), and correlation matrices (how strongly each pair of numerical variables is related). The color-based encoding makes heatmaps particularly efficient for audiences who need to understand patterns across many combinations simultaneously -- it is often faster than reading a table of numbers.
At home
Correlation heatmaps in data exploration: a correlation heatmap shows the correlation coefficient between every pair of variables in a dataset as a color. Cells with high positive correlation are red; high negative correlation are blue; near-zero correlation are white or neutral. This quickly reveals: which pairs of variables are highly related (potential multicollinearity for regression models), which variables are redundant (high correlation means they carry similar information), and which variables are potentially useful predictors (high correlation with the outcome variable). Correlation heatmaps are a standard first step when exploring a multi-variable dataset before building predictive models.
At work
Website click heatmaps and scroll heatmaps: tools like Hotjar, Crazy Egg, and Microsoft Clarity record where website visitors click and how far they scroll, then display this as heatmaps overlaid on the webpage. Click heatmaps show which elements receive the most attention; scroll heatmaps show how far down the page users scroll (and implicitly, where they stop reading). These tools reveal: elements that attract clicks but are not links (suggesting users expect interactivity that isn't there), important content placed below where most users stop scrolling (invisible to most visitors), and differences in behavior between device types (mobile vs. desktop scroll depth differs significantly).
At home
Hierarchical clustering with heatmaps in bioinformatics and genomics: heatmaps of gene expression data across many genes and samples are typically displayed with hierarchical clustering applied to both rows (genes) and columns (samples), reordering both so that similar genes and similar samples are adjacent. This reveals biological patterns: clusters of genes that behave similarly across samples (potentially co-regulated), and clusters of samples that have similar expression profiles (potentially the same disease subtype). The combined heatmap-with-clustering visualization is one of the most powerful pattern discovery tools in genomics research.
At work
Color scale choice significantly affects heatmap interpretation: diverging color scales (blue-white-red, going through neutral white at zero) are appropriate when data has a meaningful zero point and values above and below are conceptually different (correlation coefficients: -1 to +1; performance vs. target: negative to positive deviation). Sequential color scales (light-to-dark single color) are appropriate for all-positive or all-negative data (sales revenue, click counts). Perceptually uniform color scales (like Viridis or Plasma) ensure that equal changes in data values produce equal apparent color changes -- unlike standard rainbow color scales, which create visual 'fake bands' where certain colors attract more attention regardless of the underlying data value.
Real-world examples
At home
Lena checked her email open rates across different send days and times in her newsletter. She created a simple heatmap with days of the week across the top and hours of day down the side, and colored each cell by average open rate. The heatmap showed immediately: Tuesday and Wednesday sends at 10am had the darkest colors (highest open rates), while Friday afternoon and weekend morning sends were pale (lowest open rates). She shifted her newsletter send schedule to Tuesday 10am based on this pattern and open rates improved by 18%.
At work
Marcus ran a retail chain. He built a heatmap showing sales per square foot by store location (geographic) and product category (shoes, clothing, accessories, etc.). The heatmap revealed: accessory sales were very high per square foot in stores in downtown cores (red cells) but average in suburban stores (neutral cells), while footwear showed the opposite pattern. This insight drove a floor space reallocation strategy: downtown stores increased accessory floor space; suburban stores increased footwear allocation. The heatmap made a complex spatial-product relationship immediately visible.
Common confusions
- Many people assume heatmap colors mean the same thing across different heatmaps — Heatmap colors are relative to the scale defined for that specific chart -- red means 'high' within that chart's data range, not 'high' in any absolute sense. A red cell showing 100 website clicks on one heatmap and a red cell showing 10,000 clicks on another heatmap both represent 'high' for their respective datasets. Always check the color scale legend before interpreting heatmap colors, particularly when comparing heatmaps from different sources or different time periods. Colors without a visible scale are uninterpretable.
- Some assume website click heatmaps are representative samples of all users — Click heatmap tools typically sample a percentage of sessions (often 10-30% depending on traffic volume and plan) and may exclude sessions from bots, users with certain browser settings, or users who load pages very quickly. The heatmap represents a sample of user behavior, not all users, and the sampling method affects which users are represented. Additionally, heatmaps aggregate behavior across all users -- they don't reveal whether different user segments (new vs. returning, mobile vs. desktop) have different click patterns that the aggregate heatmap masks.
For analysts
Heatmaps are effective for client presentations when the audience needs to understand patterns across two categorical dimensions simultaneously. The most impactful business heatmap applications: call center volume by hour and day of week (for staffing optimization), sales performance by salesperson and product or region (for performance management), website behavior by page section and device type (for UX optimization), and marketing response rates by campaign and segment (for targeting). When presenting heatmaps, always narrate the key patterns visible in the colors -- what is the most interesting red cluster? What is surprisingly pale (low when you expected high)? The color pattern alone requires interpretation for non-technical audiences.
History & origin
The term 'heat map' was coined by software developer Cormac Kinney in the early 1990s for financial data visualization. Early heatmaps appeared in financial trading software to show stock price performance across many securities simultaneously. The technique was adapted widely for website analytics (showing where users click on a webpage), geographic data (showing population or weather patterns across regions), bioinformatics (gene expression levels across samples), and general business intelligence. Website click heatmaps (tools like Hotjar, Crazy Egg, and Microsoft Clarity) became mainstream digital marketing tools in the 2000s-2010s. Modern business intelligence platforms (Tableau, Power BI, Looker) include heatmap chart types as standard.