The 5 Types of Heatmaps and When to Use Each
There are five heatmaps worth knowing, and each answers exactly one question. Click maps tell you where people acted. Move maps tell you where the cursor wandered. Scroll maps tell you how far down they got. Attention (or engagement) maps tell you where they actually lingered. Rage maps tell you where they lost it. Pick the wrong one for your question and you'll confidently redesign the wrong thing.
Let me show you why the distinction matters before we go type by type.
I was reviewing a pricing page last month. In the replay, watch what she does at 0:14: she lands, her eyes go straight to the plan cards, she hovers over "Business," reads for maybe four seconds, then scrolls right past the "Talk to sales" button and keeps going, looking for a number. The team had a gorgeous click map showing almost nothing happening on that button. Their read was "the CTA is invisible, make it bigger." But I pulled the same page across 340 sessions and the pattern held: people weren't missing the button. They were hunting for a price the page never showed. A bigger button fixes nothing. That's the whole problem with heatmaps in one story. The map is only as smart as the question you bring to it.
Click and tap maps: where people committed to an action
A click map (tap map on mobile) aggregates every click or tap on a page and burns the busy spots red. Per Hotjar's own taxonomy, this is the most literal of the five, and it's the one most teams start with. It's the closest thing to intent you'll get from a heatmap, because a click is a decision, not a glance.
Best question it answers: are people using the thing I built for them to use? If your primary CTA, your nav, your filter controls all show up cold while some non-interactive headline runs hot, you've learned something concrete.
Worst way to misread it: treating a hot spot as approval. I've watched teams celebrate a bright-red region that turned out to be people clicking an image they thought was a link. Those are "dead clicks," and on a click map they look identical to happy, productive clicks. Same color, opposite meaning. A click map counts clicks; it doesn't know whether the click did anything. In one flow I audited, the second-hottest element on the page was a static product photo. Nobody had shipped a lightbox. 60-odd sessions a week were poking a picture that never opened.
So click maps are where you start, and also where you fool yourself first.
Move (hover) maps: where the cursor went, which is not where the eyes went
Move maps track mouse movement on desktop and shade the paths the cursor took. They feel like a window into attention. They mostly aren't.
Here's the thing I wish someone had told me early: the mouse follows the eyes loosely, and only sometimes. A 2012 Google Research study modeling gaze from cursor activity hit roughly 67% accuracy predicting which result a person was actually looking at, and the authors concluded plainly that mouse tracking is a weak proxy for eye tracking and can't replace it. Sixty-seven percent sounds decent until you remember you're making layout decisions on it. A third of the time the cursor is parked somewhere the person isn't even reading.
Best question a move map answers: is there hesitation or hunting around a specific control? If cursors swarm and stall near a form field or a confusing toggle, that clustering can flag friction worth watching in replay.
Worst way to misread it: calling a move map an attention map. It isn't. People rest the cursor in dead space while they read elsewhere. Left-handed users drift the pointer to a different edge. On trackpads, the cursor barely moves at all between clicks. If you want to know where attention actually landed, a move map is the flimsiest of the five, and I'd reach for scroll plus attention data before I trusted it.
One aside: move maps produce the prettiest visuals of the bunch, which is exactly why they end up in slide decks. Pretty and load-bearing are different jobs.
Scroll maps: how far down the page people actually got
A scroll map shows the percentage of visitors who reached each vertical depth, usually as a warm-to-cold gradient that cools as you go down. It's the antidote to the oldest bad assumption in web design, the idea that nobody scrolls.
They do scroll. In an analysis of around 20,000 sessions summarized by the agency Sculpt, 85% of visitors scrolled past the first viewport, 55% made it halfway, and 45% reached the very bottom. So the fold is real but not a wall. What's also real: Nielsen Norman Group's classic research found people spend roughly 80% of their viewing time above the fold and only about 20% below it. Both things are true at once. Users will scroll, but the top of the page is still where most of the looking happens.
Best question a scroll map answers: is my critical content sitting where people never reach it, or above where they still look? If your primary CTA lives at a depth only 30% of visitors touch, that's a placement bug you can fix in an afternoon.
Worst way to misread it: reading depth as engagement. Reaching 90% depth doesn't mean someone read the page. It might mean they scrolled fast, hunting for something and not finding it (see my pricing story above). Scroll depth is exposure, not attention. Which is the exact gap the next type exists to close.
Attention and engagement maps: where people actually spent time
An attention map (some tools call it engagement) weights each region by how long it stayed in view and how much it drew real interaction, not just whether a cursor passed over. Contentsquare's engagement-zone map, for instance, fuses click, move, and scroll signals into one exposure-time view. This is the map I trust most for the question "what did people care about," precisely because it's the hardest to game with a fast scroll or an idle cursor.
Best question it answers: given that people reached this section, did they linger or bounce off it? Attention data separates "seen" from "scrolled past," which scroll maps can't do on their own.
Worst way to misread it: assuming long dwell time is always good. Sometimes people linger because a table is confusing and they're re-reading it three times. Dwell is a signal, not a verdict. I always pair a hot attention zone with a handful of replays to check whether the lingering looks like interest or like struggle. Usually it takes five or six sessions to tell which one you're looking at.
Rage-click maps: where people lost the plot
A rage-click map surfaces spots where the same person clicked rapidly in the same place in a short window, the digital equivalent of jabbing an elevator button that won't light up. It's the only heatmap of the five that's unambiguously about emotion, and it's the one I check first when conversion is bleeding.
Rage clicks cluster where something looks interactive but isn't, or where a control is slow, or where an action silently failed. The cost is real. Contentsquare's rage-click guidance describes checkout flows where frustrated, rage-clicking users convert far below the smooth-path baseline, in their illustration something like under 1% versus roughly 4%. Even if your exact numbers differ, the direction never does. Frustration and completion move in opposite directions.
Best question a rage map answers: where is the interface actively breaking trust right now? A rage cluster is rarely a false alarm. Somebody wanted something to happen and it didn't.
Worst way to misread it: treating it as a cosmetic nice-to-have instead of a bug queue. When I find a rage hotspot, I don't file it under "UX polish." I file it next to the error logs, because half the time it maps to an actual failed request or an unresponsive element. In one checkout I looked at, a rage cluster on the "Apply promo" field lined up perfectly with a validation call that timed out. That wasn't a design problem. That was a broken endpoint wearing a design problem's clothes.
Quick reference: which map for which question
| Heatmap type | Best question it answers | Worst way to misread it |
|---|---|---|
| Click / tap | Are people using the controls I built? | Hot spot equals approval (dead clicks look the same) |
| Move / hover | Is there hesitation around a specific control? | Treating cursor position as gaze position |
| Scroll | Is my key content above where people stop? | Reading depth as if it were attention |
| Attention / engagement | Did people linger once they reached it? | Assuming long dwell always means interest |
| Rage-click | Where is the UI actively frustrating people? | Filing it as polish instead of a bug |
How to actually use them together
No single map is a strategy. The workflow that's served me: start with a scroll map to confirm people even reach the section you care about, overlay an attention map to see if they stayed, check the click map for whether they acted, and keep a rage map running as your smoke alarm. Move maps I treat as a tiebreaker, not a source of truth, for the accuracy reasons above.
And every heatmap is an aggregate. It tells you where and how often, never why. The "why" lives in the individual sessions underneath. That's why I never close a heatmap finding without watching a few of the replays that produced it. If you want the shared vocabulary for the signals these maps are built on, our behavior analytics glossary lays out rage clicks, dead clicks, and the rest in plain terms.
Most replay suites bundle several of these map types over the same session data, so you can jump from a rage cluster straight into the recording of it. Hotjar, Contentsquare, FullStory, Microsoft Clarity, and Kixo all pair heatmaps with replay, and they differ mostly in how deep the analytics goes, how aggressive the pricing gets at volume, and how much native-mobile support you need. If you're weighing the replay-first tools specifically, this session-replay comparison walks through the trade-offs with real numbers.
Pick the map that matches your question. Then go watch the sessions and find out why.
FAQ
What are the five types of heatmaps? Click/tap, move/hover, scroll, attention (engagement), and rage-click. Click maps show actions, move maps show cursor paths, scroll maps show depth, attention maps show dwell time, and rage maps show frustration.
Is a move map the same as an attention map? No, and this trips up a lot of teams. Move maps track the cursor, which follows the eyes only loosely. Google Research put gaze-from-mouse prediction around 67% accuracy and called mouse tracking a weak proxy for eye tracking. Attention maps weight actual dwell time and are the better read on where people looked.
Which heatmap should I check first when conversions drop? The rage-click map. It points straight at spots where the interface is breaking, and rage clusters often line up with real technical failures, not just design issues.
Do heatmaps tell me why users behave a certain way? No. Every heatmap is an aggregate of where and how often. For the why, open the individual session replays that produced the hot spot and watch what people actually did.