TL;DR
- A sale is not one decision. It is four or five small ones, and you lose people at every join.
- The biggest documented lever between the homepage and the sale is not persuasion. It is how fast the pages load.
- Most people who leave were never going to buy today, which is normal and not a failure.
- Measure the joins, not the overall conversion rate — a single number cannot tell you where the leak is.
Most businesses watch two numbers: how many people came, and how many bought. Everything in between is a black box, so when the second number is disappointing the response is usually to change something visible — the homepage headline, the hero image, the colour of a button — and hope.
The middle is knowable, and it behaves in a fairly consistent way. A visitor does not decide to buy. They decide to keep going, four or five times in a row, and each of those decisions has its own failure mode.
The journey is a series of joins, and every one leaks
One of the more useful pieces of public evidence here measured exactly that. In Milliseconds Make Millions, commissioned by Google and carried out by 55 and Deloitte across 37 brands, the researchers did not report a single conversion figure. They reported progression rates between individual steps: for retail, a 0.1 second improvement in mobile site speed moved 3.2% more people from the product listing page to the product detail page, and 9.1% more from the product detail page to add-to-basket. Retail consumers spent 9.2% more. Travel booking rates improved by 10%, and in lead generation there was a 21.6% improvement in the number of users who progressed to the form submission page.
Two things are worth taking from that. The first is that the effect was not evenly spread — some joins moved a lot, others barely at all. The second is that you cannot see any of this in an overall conversion rate. A site can lose most of its visitors at the second step and look identical, in the headline number, to one that loses them at the fourth.
The joins, in order
- ArriveDoes the page say what this is and who it is for? If not, they leave in seconds and nothing else you built ever gets seen.
- Look furtherDo they click through to a service, a product or a case study? This is the first sign of genuine interest rather than a stray click.
- Check you outPrices, examples of your work, who you have done it for. This is where a considered buyer spends most of their time.
- Start the actionOpening the contact form, adding to the basket, beginning the booking. The commitment is small but the intent is real.
- Finish itSending the form, completing the purchase. Almost everything that goes wrong here is friction rather than doubt.
Each of those has a different cause when it fails, which is why one number cannot diagnose it and why redesigning the homepage does nothing for a leak at step four.
The first join: the ten seconds you get on arrival
Jakob Nielsen of the Nielsen Norman Group, writing in 2011 on an analysis of more than 205,000 pages, found that users often leave a page within 10-20 seconds, and that holding someone for several minutes means communicating your value proposition within about 10 seconds. He described the opening moments as a search for a reason to leave, not a reason to stay.
So the first join is not a persuasion problem, it is a comprehension one. Can a stranger tell, without scrolling, what you sell and who you sell it to? Most homepages that fail here fail because they lead with a claim about the company rather than a description of the service — “delivering excellence since 2009” instead of “commercial electricians in Brighton and Hove”.
Nobody abandons a page they understood. They abandon the ones they had to work out.
The middle: where speed does more than copy
This is the part that surprised us in the research and changed what this article says. Every version of this advice — including the one we started with — treats the middle of the funnel as a persuasion problem to be solved with better messaging and stronger calls to action. The best-evidenced lever in that middle is duller than that. It is page speed, and the figures above came from a tenth of a second.
It makes sense once you picture the behaviour. Somebody comparing three suppliers on a phone on the train is not weighing your value proposition against a competitor’s. They are opening tabs, and the tab that has not painted anything yet gets closed. The persuasion never happens because the page never arrives. Note that those figures come from retail and travel brands on mobile, so treat them as an order of magnitude rather than a promise for your sector — but the direction is not in doubt.
None of that is a redesign. It is a morning’s work, and it acts on every join at once.
The last join: the form, and why most people leave
The final step leaks hardest, and this is where expectations need adjusting. The Baymard Institute, which averages the documented figures across 50 studies, puts the average online shopping cart abandonment rate at 70.22%. That is ecommerce, and most businesses reading this are not ecommerce — but the shape carries over to enquiry forms, and so does the lesson.
Seven in ten is not a catastrophe to be fixed. A large share of those people were browsing, comparing, or saving something for later, and no amount of design would have converted them today. What you can act on is the smaller group who intended to finish and did not: the form that asked for a phone number they did not want to give, the field that rejected their postcode, the step that demanded an account. Take a worked example: a building firm whose contact form asked for budget, timescale, project type and a description before it would send. Cutting it to name, contact and “tell us what you need” does not persuade anyone — it just stops turning away people who had already decided.
What we cut, and why
The original version of this article recommended countdown timers, “limited availability” notices and timed promotions. We took them out. For a considered purchase — an agency, a builder, a solicitor, a piece of B2B software — we could not find a credible source showing they help, and the risk is asymmetric: a fake deadline on a professional service reads as a trick, and a buyer who spots one has learned something about you that no amount of design can unlearn.
Genuine constraints are a different matter. If you really do have two slots left in October, saying so is useful information. The test is simple: would you be comfortable if the visitor could see how the number was calculated? If not, it does not belong on the page.
What to measure instead
Stop watching the overall conversion rate as a health metric — it moves for reasons that have nothing to do with your site, mostly changes in the mix of traffic arriving. Watch the joins instead.
A month of useful measurement
- Define your five steps
Write down what "looked further", "checked you out" and "started the action" mean as specific pages or events on your site. If you cannot name them, that is the first problem.
- Set up the events
Most analytics tools will track page views and form starts out of the box. You need the step-to-step percentages, not the totals.
- Find the worst join
One of them will be visibly worse than the rest. That is where your attention goes, regardless of which page you thought was the problem.
- Change one thing
Fix the speed of that page, or cut the form, or rewrite the section above it. One change, then a month of data before the next one.
It is slower than a redesign and far more likely to work, because you are acting on the step that is actually losing people rather than the one that is easiest to look at.
Good to know
FAQs
Four or five separate decisions to keep going: understanding what you do, looking further, checking you out, starting the action and finishing it. Each fails for a different reason, which is why one overall conversion figure cannot tell you where the problem is.
Usually page speed. Google’s work with 55 and Deloitte measured progression between individual funnel steps and found meaningful improvements from a tenth of a second on mobile — a bigger documented effect than most copy changes.
In ecommerce, yes — the Baymard Institute’s average of documented studies puts cart abandonment at roughly seven in ten. Many of those people were browsing or comparing and were never going to buy that day. Focus on the smaller group who meant to finish and were stopped by friction.
We would not use them for a considered purchase. We could not find credible evidence that they help in that setting, and a deadline a buyer suspects is invented does lasting damage. Real constraints, stated plainly, are fine.
The percentage of people moving from each step to the next, rather than the overall conversion rate. The step-to-step figures tell you where to work; the headline number only tells you whether last month was better.
Next step
Want to know where your site is leaking?
We help businesses across Brighton and London find the step that is losing people, and fix that one rather than rebuilding everything.