Knowledge · Conversion Engineering
Scroll Depth + Engagement Tracking: What’s Signal, What’s Noise
Scroll depth is the most-cited and least-correctly-used engagement metric in mid-market CRO. Every agency reports “average scroll depth was 67%” like that’s a number. It isn’t. Scroll metrics are useful only when you decompose them by page, segment, and intent. Here’s the unsanitized version.
№ 01Why average scroll depth is a lie
If 50% of users scroll to 20% and 50% scroll to 100%, your average is 60%. That number is decorative. The two populations are completely different — one bounced, one engaged — and averaging them produces a number that describes nobody.
Report distribution instead: percentage of users reaching scroll-25%, 50%, 75%, 100%. The shape of that distribution tells you where the page is failing. A page with strong 25% but weak 50% has a hero or problem-statement issue. A page with strong 75% but weak 100% has a closing CTA problem.
№ 02Scroll-50% as the key threshold
If only 50% of visitors scroll past 50% of your page, then everything below 50% is invisible to half your audience. That’s the most actionable scroll insight available: it tells you where your effective page ends.
Practical implication: pricing, primary form, and proof should land above the scroll-50% line. Most mid-market sites violate this — the form is at 70-80% page depth, after a long value-prop and feature-list section. Half the traffic never sees it. Move it up.
№ 03Time on page: the most-misleading engagement metric
“Average time on page: 3:47” sounds like engagement. It isn’t. Browser tabs left open inflate the metric. Reading the page in the background while in another tab inflates the metric. A visitor who opens your tab, then leaves the laptop for lunch, generates 47 minutes of “time on page.”
GA4 introduced “engaged sessions” (sessions with 10+ seconds OR a conversion OR 2+ pageviews) to partially fix this. It’s better but still imperfect. The honest engagement signal is interaction events: scroll, clicks, hovers. Use those, not raw time-on-page.
№ 04Scroll + interaction = real engagement
The composite engagement metric we use: scroll-50% + 1 interaction event within the session. Interaction events: CTA hover, FAQ expand, video play, image zoom, calendar widget open. A visitor who scrolls to 50% AND interacts with something on the page is genuinely engaged.
This composite drops the false-engagement signal from time-on-page while picking up the engagement signal that scroll alone misses (e.g., a visitor who reads carefully for 90 seconds without scrolling). Build this as an audience in GA4 and use it for segmentation.
№ 05What scroll data does NOT tell you
Scroll data tells you reach, not comprehension. A user who scrolls to 100% may have skimmed the entire page in 6 seconds without reading anything. The 100%-scroll user isn’t necessarily more engaged than the 50%-scroll user; they’re just faster.
Combine scroll with reading-speed estimation (page word count divided by scroll-time) to filter signal. A 2000-word page consumed in 10 seconds is a bounce in disguise. A 1000-word page consumed in 90 seconds is real engagement. The raw scroll number can’t distinguish.
⚠What to avoid
- Reporting “average time on page” as engagement. Tab-leavers inflate it. Use engaged sessions or scroll+interaction instead.
- Bragging about 80% average scroll depth. Distribution matters; average hides bimodal patterns.
- Tracking scroll-25% only and skipping deeper thresholds. The interesting signal is in the drop between 50% and 75%, which 25%-only tracking misses entirely.
Related questions
Go deeper
Three Ways to Start · No Sales Pitch
Want this analyzed on your site?
$500 audit. 5-day delivery. Refundable on engagement.