Scroll Depth Tracking Without Google Analytics: What It Measures, and Where It Lies

What scroll depth actually measures

Strip away the dashboards and scroll depth is one division. Take how far the page has been scrolled, divide by how far it can be scrolled, and keep the largest value seen during the visit. In GhostMetrics the calculation is scrollY / (documentHeight - viewportHeight), rounded to a whole percent and capped at 100.

Three details in that formula decide how you should read the result.

It tracks the top edge of the viewport, not the bottom. A visitor on a tall phone who has scrolled to 50 percent has the top of their screen at the halfway point of the scrollable range. The bottom of their screen is further down, so they have seen more than half the document. A reader at 100 percent has reached the very bottom of the page, nothing more.

The denominator is the whole document. Footer links, a comment thread, a newsletter box and a related-posts grid all count as page height. If the last 25 percent of your page is chrome, then the end of the article sits near 75 percent and a reader who finished it never reaches 100.

It is a maximum, not a path. A visitor who scrolls to the bottom and jumps back to the top still records 100. A visitor who jumps straight to an anchor link halfway down records 50 without ever having seen the first half. Scroll depth measures reach, not reading order and not attention.

Pages that fit inside the viewport have nothing to scroll. The tracker only measures inside its scroll listener, so a page that never fires a scroll event usually records 0 percent even though the visitor saw everything. Short pages can therefore drag the site-wide funnel down; compare like with like.

How GhostMetrics records it, with no cookies

The mechanics are small enough to describe completely, which matters if you are the one signing off on a privacy review. If you want the wider background on identifier-free measurement, our cookieless tracking explainer covers it.

1. A passive scroll listener samples the position. It runs at most once every 250 ms and only updates a single number in memory when the new percentage beats the old one.
2. The maximum lives in page memory. It is not written to a cookie, to localStorage, or to sessionStorage. The only things GhostMetrics puts in sessionStorage are a per-tab pageview counter and any first-touch UTM, both cleared when the tab closes.
3. A leave beacon ships the number. When the tab becomes hidden or the page unloads, the tracker sends one beacon carrying scroll depth, time on page and a bounce flag.
4. Single-page apps reset it per route. A pushState or popstate navigation counts as a new pageview, restarts the timer and starts the scroll maximum from zero. The previous route's depth and time are dropped at that moment, not sent, so in a single-page app only the route that is open when the tab is hidden reports scroll depth.
5. Browsers that send Global Privacy Control or Do Not Track are not measured at all. No beacon goes out, so those readers are absent from the funnel entirely.

On the dashboard side, the Scroll Depth Funnel panel counts, across all pageviews in the selected period, the percentage that reached at least 25, 50, 75 and 100 percent. The All Pages table adds an Avg Scroll column per URL. The funnel is site-wide for the period; if you want the same four rows for a single page, export the raw data (CSV from the Export data button, or JSON from the API; every plan) and group by page path.

What the funnel tells a content team

A funnel is a shape, and shapes suggest hypotheses. None of the readings below is proof; each tells you what to open next. The best habit is to look at the drop between adjacent thresholds, not at any single bar.

The last two rows are the reason scroll depth should never be read alone. Pair it with time on page (the dashboard has a distribution for it) and with entry and exit pages. A page with 40 percent average scroll and a two-minute average time on page is probably a reference page doing its job. The same scroll number with a six-second time is a page people abandon.

For editors, the highest-value use is comparative. Run the same template across ten articles and sort by Avg Scroll. The bottom three are candidates for a rewrite; the top three show you the structure that holds readers. Both are statements about your own pages relative to each other, which is the one comparison scroll depth supports well. For the broader picture of which numbers to trust, see how to measure website traffic.

Where the number misleads

Every scroll tracker has the same blind spots. Knowing them keeps you from rewriting a page that was fine.

Missing leave beacons count as zero. The depth is only reported when the tab is hidden or unloaded. If a browser kills a page before the beacon goes out, that pageview keeps its default of zero and drags the funnel down. Expect the 25 percent bar to understate reach somewhat, especially on mobile where tabs are backgrounded and discarded aggressively. The server also only attaches a leave beacon to a pageview recorded in the previous ten minutes, so a reader who keeps the tab in front for longer than that is recorded as zero too.

The beacon fires once per pageview. A reader who switches tabs after reading half the article, comes back and finishes, will have reported 50 percent when they first left. Anything scrolled after that point is not sent. Time on page has the same property.

Infinite scroll and lazy-loaded content move the goalposts. When the page grows as you scroll, the denominator grows with it, so the percentage can stall or fall while the reader keeps going. If a page loads more content on scroll, treat its percentages as unreliable and use section events instead.

Anchor links, sticky tables of contents and "jump to" buttons inflate depth. A visitor who taps "Pricing" at the bottom of your contents list records a high maximum having read nothing in between.

It tells you nothing about the person. There is no visitor ID, so you cannot ask "do returning readers scroll further?" and no per-visitor funnel across sessions exists. That is a design decision: no identifier means nothing to follow a person with.

How the other tools handle scroll depth

The question most teams actually have is what they would give up by moving off Google Analytics. For scroll depth the honest answer is: not much, unless you rely on custom thresholds.

GA4 is the one place where scroll tracking arrives bundled with a wider identity model. If your marketing team feeds engaged audiences back into Google Ads, that is a real reason to keep it, and nothing here replaces it. If the scroll event was the only reason you were reading GA4 reports, you can leave it behind; our GA4 alternative guide compares the broader options.

The bigger practical difference is consent. A cookie-based setup behind a banner gets its scroll data mainly from the visitors who accepted it. A cookieless funnel counts everyone who has not sent a privacy signal, so the sample is not limited to people who clicked "accept".

Add section-level milestones with a custom event

Percentages answer "how far". Editors usually want "which part". The tracker's custom-event API handles this, because an event carries the page path and one metadata field, section. The idea is to mark the sections you care about, observe them, and fire one event the first time each enters the viewport. Our guide to custom events without Google Analytics covers the API in more depth.

1. Install the tracker if it is not on the page yet: one <script defer> tag with your site ID in the <head>. No plugin, no build step.
2. Mark the sections you want to measure with a data-section attribute: the heading that opens the pricing table, the code sample, the conclusion.
3. Paste the observer below just before </body>. It fires seen-<section> once per section per pageview.
4. Load a page, scroll through it, and open the Custom Events panel. You should see one row per section, with a count and a last-fired time.
5. Compare the counts against pageviews for that page. The ratio is your per-section reach.

<h2 data-section="pricing">Pricing</h2>
<h2 data-section="conclusion">Conclusion</h2>

<script>
document.addEventListener('DOMContentLoaded', function () {
  // gm.js is deferred, so it has run by now. It is absent if the
  // visitor sends Global Privacy Control or Do Not Track.
  if (!window.ghostmetrics || !('IntersectionObserver' in window)) return;
  var io = new IntersectionObserver(function (entries) {
    entries.forEach(function (entry) {
      if (!entry.isIntersecting) return;
      var id = entry.target.getAttribute('data-section');
      window.ghostmetrics.track('seen-' + id, { section: id });
      io.unobserve(entry.target);
    });
  }, { threshold: 0.5 });
  document.querySelectorAll('[data-section]').forEach(function (el) {
    io.observe(el);
  });
});
</script>

Putting the section into the event name is deliberate. The Custom Events panel lists name, count and last fired, with no drill-down by metadata, so seen-pricing and seen-conclusion show up as separate rows you can read at a glance. The section field still travels with each event and lands in the raw export.

Two limits to keep in mind. The panel is site-wide, so two pages that both mark a pricing section share one row; use distinct names per page type or split the export by page path. And there is no numeric value field and no per-visitor funnel, so you can count how many pageviews reached a section but not follow one reader from section to section.

A weekly routine that fits in ten minutes

Scroll data earns its keep when it changes what you write, not when it sits in a dashboard. A routine that works for most small teams:

1. Open All Pages for a recent period and sort by views. Ignore anything with a trivially small count; a page with a handful of views produces a noisy average.
2. Read Avg Scroll next to Avg Time for the top ten pages. Flag only the pages where the two disagree, a high scroll with a tiny time or the reverse.
3. Check the site-wide funnel for your 75 percent row. Write that number down. When it moves by a few points after a template change, you have measured the change.
4. Pick one page and look at what sits at the largest drop. Edit that section, then leave the page alone for a couple of weeks so the next comparison is fair.

The discipline is changing one thing at a time. Scroll depth is an average over mixed audiences, and a rewrite that coincides with a new traffic source will move the number for reasons that have nothing to do with the rewrite. The referrers panel and the UTM breakdown will tell you if the audience changed under you.

All of the numbers above come from raw rows you can export, on the free plan as on Pro, so a skeptical colleague can recompute any figure from the data rather than taking the dashboard's word for it.

FAQ

Keep reading

Disclosure: GhostMetrics is built by Null Agency and we use it on this site, so we're not a neutral party about our own product. The product claims here describe what the tracker and dashboard do today; for GA4 and other third-party tools, check the vendor's current documentation before deciding.