Most people will not reach the bottom of this
Half your readers stop at the halfway mark, a quarter never get near the end, and the ones who share it are not the ones who read it. The usual answer to that is to write shorter. That is the wrong lesson, and here is the shape of the right one.
The number that everyone quotes and nobody acts on
In 2013, Chartbeat put scroll tracking on a large number of news sites and published what it found. On Slate, the median reader reached about the fifty percent mark of an article. Across the other sites Chartbeat was measuring, the median was about sixty percent. Only a quarter of readers on a typical page got past the sixteen hundredth pixel.
The finding travelled. It got quoted in every conference talk about attention for a decade. And then almost nobody changed what they built, because the obvious response to it, write less, is both easy to say and mostly useless.
The most uncomfortable part of that study is not the drop off. It is what Chartbeat found when it compared how far people scrolled against how often they shared. There was a very weak relationship between the two.
People were sharing articles they had not read. Not occasionally, and not as a scandal. As the normal case.
Two responses to the same finding
Cut the piece down until it fits above the drop off point. It is the obvious move and it works, in the narrow sense that a shorter page has a higher completion rate.
What it costs is everything that needed length to be true. You cannot walk somebody through a process, or a change over time, or a comparison at scale, in four hundred words. So the pages that survive this strategy are the ones that never had much to say.
Accept that attention is spent along the way rather than banked at the end, and build a page that pays out continuously. Something has to happen at the four hundredth pixel, and again at the twelve hundredth.
This is harder and it is the only version that lets a long argument stay long. It is also, awkwardly, exactly what the format we sell is for, which is why the rest of this page is written carefully.
What the scrolling format actually claims
The claim made for scrollytelling is that a page built to be moved through holds people longer and further than a page built to be read down. There are numbers attached to that claim and they are large.
An Infogram and DC Thomson study of articles carrying interactive data visualisations reported a sixty two percent increase in average dwell time and a three hundred and seventeen percent lift in scroll depth. Honda UK, relaunching a content hub in the format, reported a forty seven percent rise in click throughs and a six hundred percent increase in newsletter signups. The New York Times feature that started the genre held readers for about twelve minutes a session.
Thirteen years old, measured on news sites, and written before the phone became the default reading surface.
What the 2013 data is, and is not
The direction in that study has held up in every replication since, and the drop off is not controversial. The exact figures are another matter. Treat them as the shape of the problem rather than as a measurement of your own page, and then go and measure your own page.
What we take from all of it
Assume the bottom is optional
Put nothing load bearing below the point where three quarters of readers have already gone. If the conclusion only exists in the last screen, most people never read the piece you wrote.
Pay out on the way down
Every screen owes the reader something. A number that lands, a comparison that resolves, a thing that moves. Not a paragraph that exists to get to the next paragraph.
Stop treating shares as reading
They measure the headline. If the goal is that somebody understands the argument, shares are the wrong number to optimise and they will happily go up while comprehension goes down.
Measure your own page
Scroll depth is a few lines of code and almost nobody adds it. We are adding it to this site as we write this, which is later than it should have been: a format sold on engagement should have been checking engagement from the first page.
The part that cuts against us
There is a version of this argument that is too convenient, and it is worth saying out loud on our own site.
If the honest reading of the data is that attention is scarce and most pages are too long, then the honest conclusion for many businesses is not a scrolling page. It is a shorter page. A scrolling build cannot rescue a subject that a sentence already covers, and we would rather say that before taking the work than after.
The format earns its place where the thing being explained genuinely has a shape: a process, a change over time, a comparison at scale, an accumulation. Where it does not, the drop off numbers on this page are an argument against us, and they should be.
Sources
Chartbeat scroll depth data, reported by Slate in June 2013: How people read online
Scrollytelling case study figures, collected by DesignRush in 2025 and quoted via Metabole's scrollytelling guide, covering the Infogram and DC Thomson study, Honda UK, and the original New York Times feature.
Research current as at 26 August 2026. Figures marked as estimates or as case study reports are exactly that.