Why are people leaving your landing page? They are screening it
The risk of someone leaving is highest the second they arrive and falls from there, which is the shape of a screening test rather than an attention span. What a visitor checks before scrolling.
People do not leave your landing page because their attention ran out. The largest published analysis of page dwell time found the opposite shape: the risk of leaving is highest the second someone arrives, and it falls the longer they stay. You are not losing a timer. You are failing a screening test.
What the leaving curve actually looks like
Steep at the start, then flattening. Microsoft Research fitted dwell times across 205,873 web pages, each with at least ten thousand recorded visits, and found abandonment risk peaks immediately and declines from there. Survive the opening and your odds improve with every additional second.
- 205,873 pages analysed, each with 10,000 or more visits
- 98.5% showed a falling risk of leaving, not a rising one
- 70 seconds or less was the fitted dwell scale on 80% of pages
The authors named the behaviour screen-and-glean. A visitor skims to decide whether the page is worth reading, then reads. Your traffic dies in the skim, not in the reading.
Want the skim checked on your own page? Revslip reads it the way a stranger would.
Why the attention span framing gets it backwards
An attention span is a countdown. You get so many seconds before interest expires, which would mean the risk of leaving rises as time passes. The measured curve does the reverse. Risk starts high and falls, which is the signature of a test being passed, not a battery going flat.
That distinction changes the fix. If attention is a budget, you cut everything shorter. If it is a screen, you make the page pass the check, and length stops being the question.
Nobody is running out of patience on your landing page. They are deciding whether it deserves any.
Which pages get screened hardest
The serious ones. The same analysis split pages by category and found the early cull strongest on useful, less entertaining material. Education pages had a median shape parameter of 0.6506 against 0.7979 for vehicle pages, and lower means harsher. Business pages are triaged faster than fun ones.
That is the uncomfortable part for anyone selling software or services. Your page is not losing a fight with entertainment for patience. It is being judged more ruthlessly than entertainment ever is, precisely because it looks like work.
Where our two sources part company Nielsen Norman Group reports this as 99% of pages and turns it into a rule that you have ten seconds to state your value. The paper underneath says 98.5% and sets no deadline at all. The rounding is trivial. The deadline is an interpretation of a continuous curve, not a cliff at ten seconds.
What a visitor checks in those first seconds
Four things, in order, and all of them before scrolling. None is about your product in detail. Each is a reason to leave if it fails, and the screen is ordered so the cheapest check runs first. Fix them in this order too.
- Does this match what I clicked. The ad, the search result, the email. Different words for the same offer reads as the wrong page.
- What is this, in plain terms. Cover your logo and read the headline aloud. If it describes a category rather than an outcome, it says nothing.
- Is it for someone like me. Naming the buyer beats describing the product. A stranger should find themselves in the first screen.
- What happens if I act. One obvious next step, and what it costs in time, money and commitment.
Run all four past someone who has never seen the page. You cannot un-know your own product, which is why founders pass their own screen every time. The quickest way to borrow a stranger's eyes is the 5 second test on your website, and the research on what it can and cannot prove is worth reading before you trust the result.
To see which of the four your page fails, run a free audit.
Putting a number on the screen
Bounce rate alone will not get a fix funded, because it carries no currency. Convert it into money before you ask anyone for time or budget, and bring the arithmetic rather than the percentage. Revslip prices every finding with one published formula, and you can check it by hand:
traffic × conversion gap × average order value × mobile weight
Say a landing page takes 6,000 visits a month and converts at 1.4% while your other pages manage 2.3%. That gap is 54 conversions. At an average order value of €120 the page is costing roughly €6,480 a month, and it costs it again next month.
Where this evidence stops
Two places. The dwell time study measured general web browsing, not landing pages specifically, so the shape of the curve transfers but the exact numbers do not. And it was collected from browser logs of an era before most traffic went mobile, which is the one change most likely to have made the early screen harsher rather than gentler.
Where our own evidence runs out A crawl reads the page, never the visitor. Revslip can tell you the headline names a category instead of an outcome, and that the first screen has no next step. It cannot tell you that this particular person left after four seconds, because we were not watching them.
The 134 businesses Revslip has audited carry their own tilt. Every owner already suspected something was wrong before they arrived, so that group is a waiting room rather than a fair sample of the web.
Three things people ask once the bounce rate lands
The questions arrive in a predictable order. Whether the traffic was ever any good, whether the page or the source is at fault, and how to tell whether a change worked when the numbers are small. None of them is answered by staring at bounce rate on its own.
Is a high bounce rate always bad?
No. A page that answers the question in one screen and sends someone away satisfied will bounce and deserve to. Bounce only indicts a page that was supposed to produce an action. Judge it against the job, not against a benchmark.
Is it the page or the traffic sending it?
Split by source before you rewrite anything. If every source bounces alike it is the page, and if one source drags the average it is that source. Is it your traffic or your website runs the four checks.
How do I know a rewrite worked?
Not from a test you cannot power. Most landing pages never reach significance, and what to do on low traffic covers reading a change honestly instead. For the euro figure, how to calculate conversion loss works through every input.
If the bounce is one symptom among several, the conversion leak index ranks what usually breaks and what each fault costs.
Sources: the dwell time analysis is Liu, White and Dumais at Microsoft Research, and the ten second reading of it is Nielsen Norman Group.