How to Read a Scroll Map: Fold Myths and Real Benchmarks
A scroll map reads from the top down: the warmest band at the top is 100% of visitors, and each color step below it is the share who made it that far before they stopped. To read one well, you find the percentage where the color drops off a cliff, check that number against a benchmark for that specific page type, and then throw out the combined map and look at mobile and desktop separately. That last step is where most people go wrong.
I'll get to the benchmarks. First, a word about why I like scroll maps more than almost any other behavioral signal I work with.
The least creepy heatmap you own
I spend my days thinking about what I call the creepy line — the point where measuring a user shades into surveilling them. Click recordings can cross it. Full session replays with unmasked form fields sail right past it. A scroll map, though, sits comfortably on the safe side, and that's worth saying out loud before anyone talks features.
Here's why. A scroll map is an aggregate. It doesn't need to know who you are, what you typed, or where your mouse hovered. It records one boring fact per session, the furthest pixel that entered the viewport, then averages that fact across everybody. GDPR Article 4(1) defines personal data as "any information relating to an identified or identifiable natural person." A pooled percentage across ten thousand anonymous visitors usually isn't that. You can't reverse a scroll map back to a person.
The caveat, because there's always one: if your tool builds the scroll map by replaying individually identified sessions, the underlying store still holds per-user records, and that is personal data even if the visualization on top isn't. So the map is clean. The pipeline behind it might not be. Ask your vendor whether scroll depth is stored as an aggregate counter or as a trail tied to a visitor ID. It changes your retention and consent posture more than the pretty picture suggests.
Right. On to reading the thing.
What the colors actually mean
Every scroll-map tool renders the same underlying measurement, even when the palette differs. Microsoft's Clarity documentation describes it plainly: the map shows the percentage of visitors who reach each point on the page. Hotjar, now part of Contentsquare, frames scroll maps as showing "how far down a page users scroll and where drop-off happens."
So the top of your page is always warm, always near 100%. Nobody loads a page and scrolls up. As you read downward, the percentage falls. The number you care about isn't any single band. It's the gradient. A gentle fade means people are working their way down at a steady rate. A hard edge, where the map goes from orange to blue in the space of one section, means you've hit a wall. Something at that pixel is telling people they're done.
Two more things the map won't tell you on its own, and both matter:
- Speed. Scroll depth is not attention. Someone can blow past 90% of a page in two seconds and absorb nothing. Hotjar's own guidance separates the two: scroll maps show depth, attention maps show real exposure time. If you want to know where people linger, you need the second map.
- Why. The map shows the wall. It never shows the reason for it. That's what session replays are for. You watch five people hit the same drop-off and you usually see the cause within a minute.
The fold myth, and the 80% that won't die
You've heard the claim: "80% of users never scroll past the fold." It gets repeated in design reviews as if it were a law of physics. It's wrong, and it's been wrong for over a decade.
The number has a real origin. Nielsen Norman Group's eye-tracking work found that back in 2010, roughly 80% of viewing time was spent above the fold. Somewhere along the way, "80% of viewing time" mutated into "80% of users," which is a completely different claim. One is about where attention concentrates. The other is about whether people scroll at all. They got welded together and the Frankenstein version stuck.
What people actually do is scroll, reliably, almost immediately. The design agency Huge ran a study, reported by Fast Company back in 2014, where between 91% and 100% of participants scrolled, and 91–92% of them started scrolling almost at once, whether or not the page gave them a visual cue to do so. More recent numbers hold the line. An analysis of over 20,000 sessions summarized by Sculpt found 85% of users scrolled past the first viewport, 55% made it halfway, and 45% reached the bottom.
Here's the honest synthesis, because these two findings sound like they disagree and they don't. Nielsen Norman's most recent study found users still spend about 57% of their page-viewing time above the fold, with 74% inside the first two screenfuls. So both things are true at once: people scroll freely, and their attention is heavily front-loaded. The fold isn't a wall. It's a slope. Your most important message belongs near the top not because people can't scroll, but because that's where the attention density is highest.
Design for the slope, not the wall.
Benchmarks by page type
The single most useful thing I can hand you is a set of reference numbers segmented by page type, because "what's a good scroll depth" has no universal answer. A 45% average bottom-reach is a triumph on a 3,000-word guide and a disaster on a short landing page where the CTA sits at 30%.
These ranges are drawn from the aggregated benchmarks circulating in 2025 CRO and analytics guides. Treat them as orientation, not a scorecard. Your own historical median beats any industry number.
| Page type | Healthy average scroll depth | The number that actually matters | Common misread |
|---|---|---|---|
| Landing page | 40–60% | % who reach the primary CTA | Panicking that only 45% reach the footer, when the CTA is at 35% and converting fine |
| Standard blog post (800–1,800 words) | 50–70% | % who reach the mid-article point / lead magnet | Judging it against a landing-page number |
| Long-form guide (2,000+ words) | 60–80% | Where the gradient first breaks | Assuming a smooth fade means everyone read it (speed, not depth) |
| E-commerce product page | 30–70%, highly variable | % who reach reviews / specs | Comparing a photo-heavy PDP to a text page |
| Docs / reference | Very high top retention, steep tail | % who reach the code sample or answer | Treating a steep drop as failure when people found their answer early |
A few of these deserve a sentence.
Landing pages get the most unfair criticism. A low bottom-reach is often good. It can mean people converted before they needed the footer. Read the depth at your CTA, not at the end of the page.
Docs pages are the opposite of a blog post. A steep drop-off high on the page frequently means success: someone searched, landed, found the answer, and left satisfied. On a docs site, a shallow scroll can be your best outcome. Context is everything, which is why a raw number without a page type attached is close to meaningless.
Why your averaged map is lying to you
Now the part from the top I promised to come back to, and the one that quietly wrecks the most analyses. If you're looking at a single scroll map that blends every device, you're reading a number that describes nobody.
The problem is geometry. A responsive layout shifts, stacks, and re-flows at each breakpoint, so the same content lives at wildly different pixel depths on a phone versus a monitor. Your "50% scroll depth" band lands on the pricing table on desktop and on a stray paragraph three sections earlier on mobile. Averaging those together produces a map that's precise and wrong.
FullSession put it bluntly in their 2024 guidance: mixing desktop and mobile data into one heatmap gives you a distorted picture, because the layouts, screen constraints, and interaction patterns differ fundamentally. Their observation matches what I see in every audit. Mobile visitors typically scroll more than desktop visitors but scan faster, while desktop users scroll less and click more precisely. Blend them and both signals cancel out into mush.
So the rule is simple and non-negotiable: never read a combined scroll map. Segment by breakpoint before you draw a single conclusion. If a tool won't let you split mobile from desktop, it's not giving you a scroll map. It's giving you a rumor.
Turning the map into a decision
Reading the map is step one. Doing something is the point. My working loop looks like this:
- Pick one page and one device. No blends.
- Find the pixel where the gradient breaks hardest. That's your wall.
- Overlay a click or attention map at the same depth. Depth plus engagement tells you whether people stopped because they were done or because they were stuck.
- Watch three to five replays that hit that exact drop-off. The map shows where; the replay shows why.
- Move one thing above the wall (usually the CTA or the single sentence that carries the value), then re-measure after a real sample builds up.
That fourth step is where scroll maps stop being decoration. Plenty of platforms render scroll maps beside click data, Hotjar, Microsoft Clarity, FullStory, and newer entrants like Kixo among them, and the ones worth paying for are the ones that let you jump from a cold band on the map straight into the sessions that produced it. If you're weighing those options, we compare the replay side of these tools in our Hotjar vs FullStory vs LogRocket vs Clarity breakdown.
A few honest limits
Scroll maps have failure modes, and pretending otherwise helps nobody.
They flatten intent. A 30% drop-off could be boredom or satisfaction, and the map can't distinguish them. They're blind to dynamic content: accordions, lazy-loaded sections, and infinite scroll all confuse the pixel math, sometimes badly. And they reward long pages unfairly, because a taller page mechanically produces a lower average depth even when engagement is identical. Always normalize against page height before you compare two pages.
None of that makes the tool weak. It makes it a starting point. The scroll map tells you where to look; it never finishes the sentence. Pair it with replays and attention data, segment ruthlessly by device, and drop the fold myth in the bin where it belongs. People scroll. The question was never whether they'd go down the page. It's whether you put the thing that matters where their attention already is. If you want the vocabulary that goes with all this, our behavior analytics glossary defines scroll depth alongside the other signals you'll be reading next to it.
FAQ
What's a good scroll depth percentage? There isn't a universal one. Aim for 40–60% on landing pages, 50–70% on standard blog posts, and 60–80% on long-form guides, but your own historical median for that exact page type beats any benchmark.
Is the "80% don't scroll past the fold" statistic true? No. It's a garbled version of a real Nielsen Norman finding that about 80% of viewing time was spent above the fold in 2010. Studies consistently show 85% or more of users do scroll past the first viewport.
Should I look at mobile and desktop scroll maps together? Never. Responsive breakpoints put the same content at different pixel depths per device, so a combined map averages two incompatible layouts into a misleading blend. Segment first.
Does a scroll map count as personal data under GDPR? The aggregated visualization usually doesn't, since it can't be traced back to an individual. But if your tool builds it from individually identified sessions, the underlying store does. Check how scroll depth is retained, not just how it's displayed.