What Heatmaps Actually Tell You (And What They Don't)
Click, move and scroll maps answer questions analytics can't — but only if you read them correctly. The four patterns worth acting on, and the misreadings that cost people redesigns.
By Niraj Kumar
A heatmap is the rare growth tool that shows you something you cannot deduce, argue about, or get from a chart. You look at it and you know — instantly, and often uncomfortably — that nobody has ever seen the section you consider the heart of your pitch.
It's also the growth tool most likely to be misread, because it's so visually persuasive. A blob of red feels like evidence. It might be a bug.
Here's how to actually read one.
Three maps, three different questions
They get lumped together as "heatmaps" and they answer completely different things.
| Map | What it records | The question it answers |
|---|---|---|
| Click | Where people tap or click | What do they think is interactive? Are they finding the CTA? |
| Scroll | How far down the page each visitor got | Does anyone reach the thing I care about? |
| Move | Where the cursor travels (desktop) | Roughly where attention goes — weakly |
Scroll maps are the most valuable and the least discussed. Click maps get all the attention because they look dramatic, but scroll depth is the one that reframes how you think about your page — and it's the one whose answer is most often "much worse than you assumed."
Move maps are the weakest. Cursor position correlates with gaze loosely — the correlation is real but far softer than the marketing for these tools implies, and it collapses entirely for people who read without moving the mouse. Treat move maps as a hint, never as evidence. Nobody should redesign anything on the strength of one.
The four patterns worth acting on
1. The scroll cliff
The single most valuable thing a heatmap will ever show you.
You'll see something like: 100% of visitors see the hero, 60% reach the features, 25% reach the pricing table, 8% reach the testimonials you spent a week collecting.
The important shift here is conceptual. A page is not a page — it's a funnel. Every section has an audience, and that audience is strictly smaller than the one above it. Your carefully argued objection-handling section, three-quarters of the way down, is being read by 12% of visitors, which means it is not handling objections. It's decorating the bottom of a page.
Two responses, and the order matters:
- Move the important thing up. Always try this first. It's free and it works.
- Then ask why they're stopping. A cliff at a specific point — not a gradual decay, a cliff — usually means something there is acting as a full stop. A section that looks like a footer. A big whitespace gap that reads as "the page has ended". A wide full-bleed image that visually terminates the flow.
The gradual decay is normal and you can't fix it. The cliff is a bug and you can.
2. Clicks on things that aren't links
You'll find a cluster of clicks on something inert. An icon. A bolded phrase. A product photo. A statistic.
This is a person telling you, unambiguously, "I expected this to do something." It's the most actionable signal in the entire discipline, because there's no interpretation required and the fix is trivial: make it do that thing.
People click product images expecting a bigger version. They click a stat expecting a source. They click a feature name expecting a detail page. Every one of those is a user showing you a link you forgot to build. Build it.
3. Rage clicks
Several rapid clicks in the same spot. Someone tapped a button and nothing happened, so they tapped it again, and again.
This is nearly always a bug, not a design problem: a slow handler, a dead button, a form that submits silently, a mobile tap target too small to reliably hit. It's the visual signature of the INP problem — the page taking 400ms to respond to a tap, which the user reads as "broken."
Rage clicks should be treated as a defect report, and they should jump the queue accordingly. This is a known failure with a known location, which makes it high-confidence and usually low-effort — the top-left corner of the triage grid.
4. The CTA nobody clicks
Your primary button is a cold spot while something adjacent is hot.
Usually one of three things, and the heatmap tells you which:
- They never saw it. Cross-reference the scroll map: if only 30% reach the button's depth, it isn't a copy problem, it's a placement problem, and no rewrite will save it.
- They saw it and clicked something else. You have two competing calls to action and one is winning. Often the wrong one — the "Learn more" link is stealing traffic from "Start free trial", because it's the lower-commitment option and you presented them as equals.
- They saw it and did nothing. Now it's a copy, offer or trust problem — and only now is it worth arguing about the wording.
Notice you can't tell those apart from analytics. The conversion rate is identical in all three cases. That's the whole value proposition of a heatmap: it disambiguates causes that a metric collapses into one number.
How to not fool yourself
Heatmaps mislead more than any other tool in this stack, precisely because they look like objective data. Four ways people get burned.
Segment before you look, or don't bother
An aggregate heatmap over desktop and mobile together is worthless. The layouts are different, so the coordinates mean different things, and you're averaging two unrelated pictures into one that describes neither.
Always split desktop and mobile. It's not a refinement; it's a precondition. Anyone showing you a single combined map is showing you an artefact.
Segment further if you can — new vs returning, paid vs organic — because a heatmap of "everyone" is a heatmap of nobody in particular.
Watch the sample size
Twenty sessions is an anecdote. A heatmap built from twenty sessions will still render beautifully, in confident colours, with hot spots and cold spots and a compelling story. It's noise with a gradient applied.
This is the A/B testing trap in a more seductive costume: the visualisation doesn't show you its uncertainty, so a tiny sample looks exactly as authoritative as a large one. Get to a few hundred sessions per segment before you believe a pattern, and be especially suspicious of a hot spot that appeared this morning.
The page changed underneath it
Heatmaps are recorded against a screenshot. Ship a redesign and your accumulated data now describes a page that no longer exists — coordinates land in the wrong places, and the map is confidently, invisibly wrong.
Reset the recording after any significant layout change. And be careful with personalised or dynamic content: if the page differs per visitor, the coordinates aren't measuring the same thing across sessions.
Hot ≠ good
The reddest thing on the page is frequently your navigation, and that's not a triumph — it might mean people can't find what they need and keep going back to the menu. Lots of clicks on "Pricing" in the nav can mean your homepage never made the value clear enough to get to pricing.
Always ask "is this hotspot the behaviour I want?" before celebrating it.
Where it fits
Heatmaps are the diagnostic layer between "the number is bad" and "here's the fix."
- Analytics says what: the pricing page converts at 1.2%. It cannot tell you why, and it never will — and it's undercounting, too.
- Heatmaps say where: 70% of visitors never scroll to the pricing table.
- Conversion checks say what to do: the CTA is below the fold at 375px — move it up.
And crucially, heatmaps are the tool that works when you don't have the traffic to A/B test. You need tens of thousands of sessions to detect a 10% lift statistically. You need about two hundred to see that nobody reaches your pricing table. For a small site, this is not a consolation prize — it is a strictly better use of the traffic you have.
That's the thinking behind our heatmap module: click, move and scroll captured by the same cookieless first-party pixel that powers traffic, rendered over a real screenshot of the page, with desktop and mobile kept separate — because combining them is the mistake that makes the whole exercise pointless.
What to do this week
- Put a scroll map on your most important page. Not a click map — scroll. Find out what percentage of people actually reach your primary CTA. Prepare to be disappointed.
- Split desktop and mobile. If you only ever do one thing from this article, do this one, because everything else you conclude depends on it.
- Look for clicks on non-links. Every cluster is a link your users are asking you to build. Build it.
- Treat rage clicks as bugs, not insights. Fix them this week.
- Wait for a few hundred sessions before believing anything. The map will look convincing long before it's true.
The value here isn't the pretty picture. It's that a heatmap will show you, in about ten seconds and with no argument possible, something about your own page that you were completely wrong about — and it will do it for the price of a few hundred visitors rather than a hundred thousand.