Traffic counts arrivals. Engagement counts attention.
How far people read, how long they were genuinely there, and how many days they came back. Collected automatically, with every denominator stated instead of quietly chosen to flatter.
What are engagement metrics?
Engagement metrics measure involvement rather than arrival: how far down a page a visitor read, how long they spent actually paying attention to it, and how many separate days they came back. They are the numbers that separate a page which gets traffic from a page which does something with it.
They are also the numbers most easily faked. Time on page that counts a backgrounded tab, and scroll depth that counts a page which fit on screen, both produce healthy-looking charts that describe nothing. The definitions matter more than the dashboard.
Scroll depth, per page.
For each page, Mrkr shows what share of pageviews reached 25%, 50%, 75% and 100%, plus an average depth. It is measured by how far the bottom of the viewport travelled through the document, so somebody who saw a whole short page without scrolling is not recorded as having scrolled.
The denominator is pageviews that reported a threshold at all, not every pageview. Pages with less than 240px below the fold report nothing on purpose, and thresholds already visible on load are marked silently rather than fired as an event.

- scroll thresholds: 25, 50, 75 and 100 percent
- 4scroll thresholds: 25, 50, 75 and 100 percent
- of no interaction before engaged time stops accumulating
- 30sof no interaction before engaged time stops accumulating
- of content below the fold before a page is tracked for scroll at all
- 240pxof content below the fold before a page is tracked for scroll at all
- minimum engaged time before anything is reported
- 1sminimum engaged time before anything is reported
Geometry is re-measured on every pageview, every client-side route change, and on a debounced window resize, so rotating a phone mid-visit does not misfire a threshold against a stale measurement. None of these reports are metered: billing counts pageviews only.
What each number actually counts.
The definitions, written out, because every one of these is measured differently by different tools and the differences are large enough to change a decision.
| Metric | Definition in Mrkr | Denominator |
|---|---|---|
| Engaged time | Accumulates only while the tab is visible and there has been an interaction in the last 30 seconds | Pageviews that reported over one second |
| Scroll depth | How far the bottom of the viewport travelled through the document, at four thresholds | Pageviews that reported any threshold, not all pageviews |
| Stickiness | Distinct calendar days a visitor was active within the range | All visitors active in the range |
| Returning share | Visitors active on more than one day, as a share of all visitors | All visitors active in the range |
| Bounce rate | Sessions with a single pageview, available broken down by source | Sessions in the range |
Stickiness and returning share need cookie mode to be meaningful. In the default cookieless mode the visitor identifier rotates every 24 hours, so nobody can be observed on two days.
What engagement is good for.
Telling skimming from stalling
Most people stopping at 50% is a middle that loses them. Everyone hitting 100% in eight seconds is a skim.
Judging long-form honestly
A 3,000-word page with a 40-second average engaged time was not read. Useful to know before commissioning ten more.
Finding the section that loses people
A consistent cliff between 50% and 75% across several pages usually points at a shared component in the same place.
Backing a conversion hypothesis
If converters read twice as far as everyone else, getting people to scroll is the lever, not rewriting the headline.
Open the engagement view
The live demo has stickiness and per-page scroll depth on real demo traffic. No signup, no card.
Open the live demoWhere these numbers get misused.
Time on page is not engaged time
Most tools time the gap between two pageviews, so a tab left open over lunch is a two-hour visit. Mrkr counts presence only.
Scroll depth is not a quality score
A reference page that answers in the first paragraph should score low. Compare against its own history, not a site average.
A missing scroll row is information
A page with no scroll data is a page that fits on screen. Reading it as 0% is the commonest misinterpretation here.
Bots would flatten all of it
Automated traffic neither scrolls nor pauses, so unfiltered engagement is dragged towards zero. Bots are filtered at ingest.
Questions, answered.
Keep reading
- Engagement documentation
The exact stickiness definition and an honest read of the scroll denominator.
- Scroll and engagement tracking
What triggers each signal in the browser, including on single-page apps.
- Retention analysis
The multi-day view: who comes back, and what that needs.
- Core Web Vitals monitoring
Whether the page was slow before you blame the copy.
- Session replay without the cookie banner
Watch what a low scroll depth actually looked like.
- Every Mrkr feature
The full index of what the product measures.
Your first visitor is already here.
Drop in the script and watch them land. It takes about a minute.