The Psychology of Scroll: How We Engineer Landing Pages That Convert

“The psychology of scroll” makes the process sound more mysterious than it is. Most of what determines whether a landing page works is decided in the first fraction of a second, before any scrolling happens, and the rest is a fairly unromantic exercise in not making people work. Here is what the evidence actually supports, and — just as usefully — what it does not.

The verdict arrives before the scroll does

People assess a page absurdly quickly. Lindgaard and colleagues at Carleton University, writing in Behaviour & Information Technology in 2006, found that visual appeal ratings given after a 50 millisecond exposure were highly correlated with those given after 500 milliseconds. That is a judgement formed faster than a deliberate blink, and long before a headline has been read.

What produces a favourable verdict is well established too, and it is not what most landing page advice implies. Tuch and colleagues, publishing in the International Journal of Human-Computer Studies in 2012, tested first impressions of 119 real websites and found that low visual complexity and high prototypicality — a layout resembling what people expect that kind of page to look like — produced the most positive reactions, with the effect detectable at 17 milliseconds.

That is the single most useful finding for anyone building a landing page, and it is quietly deflating. The unusual layout, the asymmetric hero, the scroll-jacked animation sequence — all of them are working against the grain of how a page is first assessed. Save the invention for the content.

Reducing what the page asks of people

Once the first impression is survived, the constraint is attention. A landing page competes not with a rival’s page but with the reader’s willingness to keep going, and every element that has to be interpreted spends some of it.

The practical distinction worth keeping is between the effort a reader must spend to understand your actual offer, and the effort spent on everything else: the cookie banner overlapping the headline, the carousel moving on its own, the four calls to action of equal weight, the icon row that decorates rather than explains. The first is unavoidable. The second is a design decision that can be reversed this afternoon.

• One primary action per screen, at one weight — competing buttons are the commonest fault
• A single column of argument, not a grid the eye has to choose a path through
• Movement only where it means something; nothing that animates on a timer
• Space around what matters, because emphasis is created by isolation rather than size
• Every icon, badge and divider justified by what it explains, not by how it fills the row

The test is not whether the page looks minimal. It is whether a stranger, glancing at one screen, can say what is on offer and what to do next. Most landing pages fail that test while looking perfectly tidy.

They will not read it, so put the point first

Landing pages are usually written as arguments that build to a conclusion. Readers do not consume them that way.

Analysing 45,237 page views, Jakob Nielsen of the Nielsen Norman Group concluded in 2008 that on an average visit users have time to read at most 28% of the words on a page, with 20% the more realistic figure. The same group, drawing on a Microsoft analysis of more than 205,000 page visits, found in 2011 that users often leave within 10-20 seconds, and that holding someone for longer means getting the value proposition across within about 10 seconds. Both are old studies, and nothing since suggests people have become more patient.

So write the page in reverse. The conclusion goes in the headline. Each section heading states its point rather than teasing it. Benefits precede mechanisms, proof precedes elaboration, and the paragraph that restates the paragraph above it is deleted — every landing page has one, and it is always the cheapest improvement on the list.

Speed is part of the psychology, not separate from it

A slow page is a page that has already spent some of the attention it was going to need. It is also the one part of this where the link to money has been measured properly.

In Milliseconds Make Millions, research commissioned by Google and carried out by 55 and Deloitte across 37 brands and more than 30 million sessions, a 0.1 second improvement in mobile site speed moved 3.2% more shoppers from a product listing to a product page and 9.1% more from that page to the basket. Retail consumers spent 9.2% more and travel bookings rose 10%. Those are retail and travel figures, not a universal law — but the scale of the intervention is the point. A tenth of a second is less than most landing pages give away to a single unoptimised hero image.

What heatmaps can and cannot tell you

Here is where a lot of conversion writing, including the earlier version of this article, quietly oversells itself. Heatmaps, scroll maps and session recordings are useful. They are not evidence of cause.

A scroll map showing that most visitors stop before the second section tells you they stopped. It does not tell you whether the section was boring, whether the page had already answered their question, or whether they were on a train that entered a tunnel. A heatmap showing few clicks on a button tells you the button was not clicked, not that it was the wrong colour. Treating these tools as proof is how teams end up redesigning a page repeatedly and calling the resulting noise a result.

It is worth being blunt about which questions these tools can actually settle and which they only appear to.

Use them to generate questions, then answer the questions somewhere else — by watching five people use the page and describing what they think it does, or by reading the last twenty enquiries and counting how many ask something the page should already have answered. That is less impressive than a heat map in a deck, and considerably more informative.

A practical pass over a landing page

If you have an hour and a page that is underperforming, this is the order that finds the most for the least effort.

None of this is engineering in any impressive sense. It is removing the things that were never doing any work, saying the point sooner, and making the page arrive quickly. The psychology is real, but the practice it recommends turns out to be mostly subtraction.