How scroll depth is measured
Scroll depth is recorded as an event. Your site watches how much of the page has passed the top of the screen, and when the visitor crosses a set threshold, a small message is sent to your analytics saying so. Cross a further threshold and another message is sent. The result is a count of how many people reached each marker on the page.
GA4 does a simple version of this on its own through enhanced measurement, firing once when a visitor reaches near the bottom of the page. Anything more detailed — several markers, or markers tied to a particular section — is normally built in Google Tag Manager and sent through as a custom event.
One thing to understand about the number: it measures the page moving, not the reader reading. A visitor who flicks a thumb from the top to the bottom of a long article on a phone will register as having reached the end, having read nothing.
Why scroll depth matters
It answers a question page views cannot: was the content that mattered ever seen? If your enquiry form, your prices or your strongest proof sit low on a long page and most visitors never reach that far, you do not have a persuasion problem. You have a layout problem, and moving the section up is cheaper than rewriting it.
It is also the best evidence for settling arguments about page length. Long pages are not automatically bad; unread pages are. Scroll data tells you which of yours is which, and it tells you where readers give up so you can look at what sits at that point.
Where scroll depth goes wrong
The most common misuse is treating deep scrolling as success. Someone who scrolls the whole way down may be engrossed, or may be hunting for a phone number that should have been in the header. Depth on its own says nothing about satisfaction, so read it next to conversions rather than on its own.
The second is marking it as a key event so it counts as a conversion. That inflates every report, pollutes any audience built on it, and — if the event is imported into an ad platform — teaches automated bidding to buy scrollers instead of buyers.
The third is comparing pages of different lengths on the same thresholds. Reaching the foot of a short contact page and reaching the foot of a long guide are not the same achievement.
What to do about it
Turn it on, give it a few weeks, then look at the pages that matter commercially rather than the ones with the most traffic. Where readers stop short of your call to action, move the call to action. Where they stop at a particular section, read that section again — it is usually a wall of text, a heavy image, or a point where the page stops answering the question it opened with.
Pair it with a heatmap before you redesign anything. Scroll depth tells you where attention ends; a heatmap and a few recordings tell you why, and the two together are far more convincing than either alone.