Scroll depth benchmarks: what has actually been published, and a calculator for your own pages
Last reviewed: 27 July 2026. Every external figure below was read from the publisher's own page and is attributed inline with the date it was read.
The short answer
Only four scroll measurements have ever been published, none of them a DTC benchmark, and the 53% figure most pages quote is not one of them. Three of the four are at least eight years old, they report four different statistics, and not one is cut by ecommerce page type. So get your own distribution from the calculator below, then compare your product page against your product page last month. That is the only comparison that holds up.
Scroll depth calculator
Five numbers from one page or template. Worked example loaded — replace it with your own.
Views that reached
10,000 views · desktop
Reach curve — share of views surviving each threshold
Depth of the page
Estimated average scroll depth
55.9%
Estimated from banded midpoints. Bucketed inputs cannot produce a true mean.
Biggest drop-off band
You lose 24% of desktop views between 25% and 50% of the page.
Your number, beside every scroll figure anyone has published
- 55.9% Your pages Estimated average scroll depth, desktop Your own views, today
- 50.5% Contentsquare Scroll rate, desktop 99 billion sessions · 2026 like for like — same unit
- 50% Chartbeat, in Slate Median scroll of a Slate article — 60% across other Chartbeat sites Slate pages + Chartbeat aggregate · 2013 editorial, 2013
- 57% Nielsen Norman Group Share of page-viewing time above the fold 120 participants · 2018 gaze time — not comparable
- 50% Chartbeat, pixel study Readers reaching 1,500 pixels on content pages 25 million sessions · ~2013 pixel row — not comparable
What scroll depth actually measures
Scroll depth is the share of a page's height that a single view reached. It is per view, not per session: a shopper who opens your product page three times produces three numbers, not one.
It gets reported two ways. A threshold reach rate is the share of views that got past a fixed line — 25%, 50%, 75%, 90%. An average or median depth is one number standing in for the whole distribution. Contentsquare's published scroll rate is the second kind: their own metric documentation defines it as the last line of pixels displayed divided by the pixel height of the page (support.contentsquare.com, Scroll Rate, read 27 July 2026).
Which is why what follows is four different measurements, not one benchmark.
Every published scroll-depth benchmark, and what each one measures
Four real measurements exist. The fifth row is the number you were probably quoted.
| Source | Headline figures, verbatim | Sample and vintage | What the unit actually is | Safe to cite for |
|---|---|---|---|---|
| Contentsquare, 2026 Digital Experience Benchmark | "Scroll rate is 5 percentage points higher: 50.5% on desktop vs. 45.2% on mobile." And "Scroll rate decreased –2%" year on year | "99 billion sessions analyzed across web and mobile" from "6,500+ websites across 9 industries"; "Year-on-year comparisons from Q4 2024 to Q4 2025". Benchmark published 25 February 2026; the engagement chapter carrying the scroll figures published 9 March 2026 | Scroll rate — the last line of pixels displayed ÷ the pixel height of the page, per Contentsquare's own metric documentation (support.contentsquare.com, Scroll Rate, read 27 July 2026). That is scroll depth, and it is the one external figure directly comparable to your calculator output | A current cross-industry figure and the device gap. Not a DTC target, and not a page-type target |
| Nielsen Norman Group, "Scrolling and Attention" | "57% of their page-viewing time above the fold" · "74% of the viewing time was spent in the first two screenfuls" · "more than 42% of the viewing time fell within the top 20% of the page" · "more than 65% of the time was spent in the top 40%". Their 2010 baseline was 80% above the fold | 120 participants, 130,000+ eye fixations, 1920×1080; published 15 April 2018. Tasks spanned "news, ecommerce, blogs, FAQs, and encyclopedic pages" | Share of gaze time by page position. Not how far anyone got | Where attention concentrates, and the fact that it moved between 2010 and 2018 |
| Chartbeat, published in Slate | "Chartbeat's data shows that most readers scroll to about the 50 percent mark, or the 1,000th pixel" at Slate; across other Chartbeat sites "most people get to 60 percent of the article". "About 5 percent of people" who land on Slate pages never scroll at all; on the wider set, "10 percent" | Slate article pages plus Chartbeat's aggregate; published 6 June 2013. Analysis by Josh Schwartz of Chartbeat, written up by Slate's Farhad Manjoo | Median scroll length as a share of the article | A real median scroll depth, in the same unit as your calculator output — but thirteen years old and editorial only |
| Chartbeat, "Scroll behavior across the web" | "just under 70% of visitors saw the very top of the page they were viewing" · "The most viewed area of the page is just above the fold, at about 550 pixels, with just over 80% viewership" · "About 50% of readers see 1500 pixels down the page on content pages, while on home pages and section fronts 50% of readers make it to pixel 1000" | "a random sample of 25 million user sessions"; undated on the page, internal references place it around 2013; news and editorial sites | Share of visitors who saw a given pixel row | The pixel-level shape of a long editorial page. Not ecommerce, and there is no device split |
| "An average of 53% scroll is typical" | "An average of 53% scroll is typical. Maybe higher for shorter content, lower for longer." | One blog post — giveitanudge.com, 14 October 2022, read 27 July 2026. No study, no dataset, no sample size, no outbound citation | Unstated | Nothing. No source is given, and there is none behind it |
That last row is not a straw man. It is the verbatim sentence you get if you search this term, and it is now repeated as though it were established. Two more figures circulate with Chartbeat's name attached — "average scroll depth is 27%" and "34% desktop / 25% mobile." Neither appears on either Chartbeat page. We read both on 27 July 2026.
Three things follow. A scroll rate, a median share of an article, a share of gaze time and viewership at a pixel row are four different statistics; averaging them produces a number that means nothing. Only one source is current. And none of the four is cut by ecommerce page type, which is the cut that decides anything: a collection page and a cart are not the same reading problem.
One concession, and it matters. For a cross-industry figure in a board deck, cite Contentsquare's. Ninety-nine billion sessions from 6,500-plus websites is a sample no DTC-only dataset will match. How Mtrix and Contentsquare differ.
Your GA4 scroll depth is a yes/no flag at 90%
Google's own documentation on the percent_scrolled dimension:
"The percentage down the page that the user scrolled. If someone scrolls at least 90%
of a page, the value '90' populates the dimension. Otherwise, the dimension is
blank." (support.google.com/analytics/answer/9143382, read 26 July 2026.)
So out of the box GA4 gives you a flag, not a distribution: the shopper who stopped reading at your third section looks identical to the one who bounced off the hero. Finer buckets are a custom event someone has to build and keep alive. When a page quotes an average scroll depth, ask which tool produced it — if the answer is GA4, the number is either not scroll depth or not from GA4.
The DTC benchmark nobody has published yet
Mtrix draws scroll maps — depth, attention and exit points, by device — on 100% of sessions. Sampling is an opt-in control here, not the default, so the aggregate isn't a sample of a sample.
No publisher has that cut by device and ecommerce page type. Contentsquare's published scroll figures are split by device only — its page-type cuts, like the PDP figures below, are bounce and traffic share. Nielsen Norman Group's study included ecommerce pages but published no ecommerce figures; the one category it breaks out separately is search-results pages. Chartbeat's two datasets are editorial. Nobody has published median scroll depth for a product page.
What to actually do with a scroll number
Scroll depth is a diagnosis, not a verdict. It tells you where attention stops, never why, and never what a fix is worth.
Read it against yourself, split by page type and device. Same template, same week last month. A product page at 48% and a blog post at 48% are not the same result, and the two devices need separate targets — Contentsquare's own figures sit five points apart. Contentsquare also reports that "1 in 3 visits now starts on a product detail page (PDP)" and "61% of visits to these PDPs bounce" (read 27 July 2026). The PDP is where the reading happens now, and it is the page no benchmark covers.
Then go and look. A scroll map gives you the row where attention dies. A session replay gives you the shopper who stopped there — stuck, distracted, or waiting on a review block that loaded three seconds late. In Mtrix the heatmaps and those sessions sit in the same platform on the same subscription, not in two.
Then change something, and test it. Moving your add-to-cart above the drop-off band is a hypothesis until an experiment settles it.
If all you want today is a free scroll map, install Microsoft Clarity. Free forever, no traffic limit on the account, recordings and heatmaps included — with Clarity's own cap of 100,000 page views per heat map. Microsoft's wording: "Clarity is a free service forever. You never encounter traffic limits or are compelled to upgrade to a paid version" (learn.microsoft.com/en-us/clarity/faq, read 27 July 2026). Better than paying for a platform you don't need yet. Mtrix has no free tier: one month free, then paid, flat across three tiers and not metered on sessions or orders. It is built for brands where a percentage point of conversion is worth more than the subscription.
Frequently asked questions
What is a good scroll depth?
Is 53% the average scroll depth?
Why does GA4 only report 90% scroll depth?
scroll event fires once, the first time 90% vertical depth becomes visible, and the percent_scrolled dimension is blank for everyone below that. Finer thresholds are a custom event you build and maintain yourself.