What Heatmap Analysis Reveals About User Behavior in Web Design

Heatmaps reveal where visitors click or tap, how far they scroll, and, on desktop, where the mouse cursor travels. They do not show where people look, and they never show why people did what they did. Read as behavior data with those limits, and collected from enough visits, they are a fast way to find the parts of a page worth investigating with recordings, tests, or interviews.

Which map answers which question

Map type	What it records	Answers	Does not answer	Devices
Click or tap map	Where people click or tap; some tools, such as Clarity, group clicks by page element	What draws interaction; what gets ignored; what people expect to be clickable	Why they clicked	Desktop and mobile
Scroll map	How far down the page each share of visitors gets	Where attention thins; whether key content is reached	Whether people read what they scrolled past	Desktop and mobile
Move map	Where the mouse cursor travels	What people hover over before acting	Where they were looking	Desktop only; touch screens have no cursor
Eye-tracking study	Where eyes fixate	What people look at and in what order	Why, unless combined with think-aloud	Lab or specialized setup

Contentsquare’s heatmap guide draws the key line: “eye-tracking measures where people look when viewing a website, whereas mouse-tracking measures where people move and click their cursors (or tap on mobile).”

Click maps: interaction and expectation

Microsoft Clarity’s heatmap documentation describes click maps built from “elements-based user clicks instead of the absolute mouse position”, and lists views for all clicks, dead clicks, rage clicks, error clicks, first clicks, and last clicks. Each view points to a different question:

• Clicks on things that are not links show an expectation mismatch: an unlinked product image that gets clicked may be one people expect to enlarge.
• Dead clicks (clicks that produce no response) and rage clicks (rapid repeated clicks) can signal a broken or unresponsive element. Check the element before concluding: a carousel arrow can produce rapid clicks in normal use.
• A main call to action with few clicks suggests the page’s hierarchy is drawing attention elsewhere, which is covered in visual hierarchy principles.

One limit to remember: a click map is drawn on a snapshot of the page, so menus, accordions, and other content that only appears after interaction may not show their clicks clearly.

Scroll maps: reach, and what the fold line means

Clarity describes its “Average fold” as the line that “indicates the average location visible on a webpage before the user starts scrolling.” It marks what visitors see on load, not where they abandon the page; look below it for any depth where the share of visitors drops sharply.

For a rough sense of a normal decline, Nielsen Norman Group’s 2018 scrolling study found that users spent about 57% of their page-viewing time above the fold and 74% in the first two screenfuls. A gradual loss is expected. A steep drop at one specific point, such as right after a full-width banner that looks like the end of the page, is worth investigating.

Compare devices against their own baselines. Content sits at different depths on phones and desktops, and the direction of the difference is not fixed: Chartbeat’s analysis of news homepages found the average mobile user stopped scrolling 25% of the way into the homepage, compared with 34% on desktop. That is one kind of page, so treat it as a reason to segment, not as a rule for yours.

Move maps: cursor, not gaze

Move maps are the easiest to over-read. In a CHI 2012 study at Microsoft, Jeff Huang, Ryen White, and Georg Buscher tracked 36 participants’ eyes and cursors during web search and found the cursor lagged behind gaze “by at least 250 ms and on average by 700 ms.” The distance between cursor and gaze was longest when the cursor was idle, shorter when people used it to help read, and shortest when they were performing an action such as a click. The authors concluded that “claiming that the cursor approximates the gaze is misguided” as a general rule.

Use move maps to see what people hover over just before acting. Do not use them to infer reading patterns such as the F-shape, which comes from eye-tracking research and, in Nielsen Norman Group’s reading, signals a page without strong formatting cues.

How much data before drawing conclusions

Small samples can produce hot spots that are just a few random clicks. Contentsquare’s guide gives a rule of thumb: “You should collect at least 2,000-3,000 pageviews before drawing conclusions from your heatmap data.” If you segment by device, apply that threshold to each device. Collect across full weeks, avoid editing the page during collection, and keep sale periods or campaign spikes separate from normal traffic.

Context decides what a pattern means

The same pattern can be healthy or a problem:

Pattern	On a homepage	On a single-goal landing page
Clicks spread across many links	People are finding their way, as intended	Attention is leaking away from the one goal
Heavy clicks on navigation	Normal wayfinding	Visitors may not find what they need on the page
Deep scrolling	Interest in the content	Possibly searching for missing information, such as price

Segment by traffic source as well as device: visitors from a paid campaign and visitors from search can behave differently on the same page.

Turn patterns into hypotheses, then check them

A heatmap tells you what happened. To learn why, combine it with other evidence:

1. Session recordings of visits in the hot or cold areas show the sequence of actions.
2. Moderated usability tests let you ask why someone clicked or stopped.
3. A one-question survey, such as “What were you looking for?”, captures intent.
4. An A/B test confirms whether a change prompted by the heatmap improves the outcome, covered in how A/B testing should guide design decisions.

When several sources point the same way, act; when they disagree, investigate further before redesigning.

Choose the map that matches your question, collect enough visits per device, read move maps as cursor data, and treat every pattern as a hypothesis to confirm before you change the page.