Heatmaps Explained: What attention patterns can—and cannot—reveal about user intent

A website heatmap aggregates interactions such as clicks, scroll reach, cursor movement, or time spent over a page image and encodes concentration with color. It can locate patterns in a filtered sample. It cannot, by itself, reveal why a person acted, what was noticed without interaction, whether content was understood, or whether an element caused an outcome.

There is no universal heatmap formula or “good” hot zone. One product may aggregate clicks by page element, another by pointer coordinates, and another may label viewport time as attention. Every percentage needs its event definition, denominator, device segment, page state, and sampling rule.

That is why “the heatmap shows intent” is too strong. The map shows an observable trace. Intent is one possible explanation to investigate.

First identify what generated the color#

Microsoft Clarity’s heatmap overview distinguishes several map types. Hotjar’s documentation uses a related but product-specific set. The labels are similar enough to invite confusion, so read the collection contract rather than the color legend alone.

Clicks are actions, not explanations#

A hot call to action could mean that the label and placement worked. It could also be the largest interactive element, a misleading affordance, a repeated failure, or the only available escape. Rage-click and dead-click views can narrow the hypothesis, but even their product-defined thresholds are behavioral flags rather than direct reports of frustration.

A cold element is equally ambiguous. People may not have reached it, may have read it without needing to click, may have completed the task earlier, or may not recognize it as interactive. Scroll and click data need to be joined conceptually before interpretation: a low click count among all page views means something different from a low click count among views that exposed the element.

Cursor position is not eye gaze#

Cursor movement can sometimes help predict attention, but it is not a direct eye-tracking measurement. A Microsoft Research controlled study examined gaze and cursor behavior in web-search tasks and documented several cursor behaviors: people may leave the pointer alone while reading, use it as a reading aid, or park it near material they are examining.

That means a movement heatmap is a useful large-sample interaction trace under its collection conditions. It cannot be relabeled as “what users looked at” without a validated method for the task, device, and population. The boundary is especially important on touch devices, where there may be no persistent pointer between taps.

Aggregation can create a pattern no one experienced#

Heatmaps collapse many sessions into one image. That strength makes patterns visible, but it also erases sequence and context. The aggregate may combine:

• desktop and mobile layouts with different element order;
• new and returning visitors with different knowledge;
• paid and organic traffic with different promises;
• signed-in and anonymous states;
• experiments or page versions;
• dynamic menus, dialogs, and validation errors; and
• sessions with different tasks and outcomes.

Clarity’s FAQ documents sampling, dynamic-element, screenshot, and iframe limits in its own implementation. Its filter documentation shows why device, path, traffic, and session conditions should be isolated before comparison.

If the page changed during the collection window, version the analysis. If a URL represents several templates or states, split them. If the target element moved between devices, compare the correct rendered experience. Otherwise, a smooth red area can be an average of incompatible screens.

Convert the pattern into a bounded hypothesis#

What heatmaps are genuinely good at#

Heatmaps are fast locators. They can point a team toward unexpected noninteractive clicks, content few eligible views reach, differences between devices, concentrated interaction around one region, or a page state worth watching. They can also make a diagnostic conversation concrete enough to decide what evidence to collect next.

Their value disappears when color is treated as motive. The right outcome of a heatmap review is usually a sharper question, not a verdict about the user.