Buy a thousand clicks at the UK ecommerce average CPC and you've spent £950. Around twenty of those people will place an order on that visit. The other nine hundred and eighty will look at something, form an opinion, and leave without telling you who they are.
You paid for all thousand. You can only see twenty.
Most founders I speak to have never put a number on the rest. They know the conversion rate. They know the ad spend. They have never multiplied the two together in the direction that hurts.
Why am I getting clicks but no conversions?
Because most of those clicks were never buying today. Around 98% of ecommerce sessions end without a purchase, and a large share of those people are browsing, price checking or looking at something they will decide on in six weeks. Your site didn't stop them. They just weren't decided yet.
For the last 20 years, everywhere I have worked, the answer to this question has been the same. Fix the site. Speed it up. Rewrite the product page. Add reviews. Move the button. Run a test, run another one, redesign the whole thing when the tests stop working.
I have run those programmes. Some worked. Plenty did not, and the ones that failed all failed for the same reason.
Conversion rate optimisation starts from an assumption nobody says out loud: that everyone who arrived came to buy, and something on your site stopped them. If that assumption holds, the work makes sense. Take the friction away and the sale appears.
It does not hold. Sometimes people are browsing. Sometimes they have three other tabs open and are checking your price against them. Sometimes they clicked the wrong thing and left in four seconds.
So you end up with a strange bit of arithmetic. The visitors most likely to buy today need the least help from you. Visitors who need the most help were never going to buy today, no matter what you did.
What does an anonymous visitor actually cost?
Multiply your non-converting sessions by your cost per click. A brand spending £20,000 a month on paid, at the UK ecommerce average of around £0.95 per click, buys roughly 250,000 clicks a year. If 98% of those sessions end without an order, about £233,000 of that annual spend lands on people who did not buy.
Use money rather than percentages, because percentages don't land and money does.
The £0.95 figure is the UK ecommerce average, with a typical range of £0.40 to £2.00, from Whito's Q1 2026 UK PPC cost research. The 2.08% conversion rate comes from a range of different studies and data across the web.
That £233,000 on its own is nothing to panic about. It is how ecommerce works. Nobody converts every session.
The question I would rather you sat with is this. Of those 245,000 people, how many left you anything at all? An email address, a phone number, a question, any signal you could act on later?
For most brands I look at, the honest answer is a low single-digit percentage, collected by a discount popup that fires the same way at everyone at the same moment.
So you spend the £233,000 once to bring them in. Then you spend it again next quarter, at full price, buying a retargeting audience made of the same people. And somewhere along the way that started getting called a strategy.
Can you tell who is browsing and who is buying?
Yes, and earlier than most brands assume. A study in Scientific Reports that analysed 443,652 real ecommerce sessions found purchase intent could be predicted from the first five clicks, with classifiers reaching an F1 score above 60% that early and above 70% by the ninth click. Sessions typically ran more than twice that long.
In plain terms: within the first few clicks, the difference between someone who will buy and someone who is having a look is already visible in what they do. And you usually have the rest of the session left to act on it.
Most Shopify stores do nothing with that. The browser and the buyer get the identical experience. Same popup, same offer, same timing, same silence when they go.
Treating a researcher like a buyer wastes both of them. The buyer did not need convincing. The researcher wasn't ready and got handed 10% off a decision they hadn't made yet.
Why doesn't a better popup fix it?
Because a popup asks for something before it gives anything, and it asks the same question of everyone. Someone comparing two winches on spec wants to know which one fits their vehicle. Someone reading a fitting guide at eleven at night wants to know whether the job takes two hours or a weekend. A discount code answers neither.
Answer the actual question and you get the email anyway, plus the far more valuable thing: what they were trying to work out. No analytics package will ever give you that. Analytics records what people clicked. It has never once recorded what they wanted.
The brands that get this right stop treating capture as a form to be filled and start treating it as a conversation to be started. The visitor gives you their intent because you gave them something useful first.
Do that with 5% of 245,000 non-converting sessions and you have 12,000 identified people a year with stated interest, sitting in your database rather than in a retargeting audience you rent from Meta.
What should you do this quarter?
Three things, in order, and none of them need a redesign.
Work out your own number first. Take last year's paid spend, divide by your average cost per click, multiply by one minus your conversion rate, then multiply by your cost per click again. That is what you spent on sessions that ended without an order. Most founders find this uncomfortable, which is rather the point.
Then find your capture rate on that specific traffic. Overall list growth will not tell you. What you want is the paid sessions that did not convert, and what percentage of them left any identifying signal at all. Very few brands have separated this out, and when they do, the number usually comes in under the guess.
Then pick one high-intent moment and make it a conversation. A product page for a considered purchase. A sizing or fitting guide. A back-in-stock request. Somewhere you already know roughly what the person is trying to decide, and where a straight answer matters more than a discount.
You do not have to fix your conversion rate to do any of this. Your conversion rate measures the people who were ready. Everybody else is a different problem, and at most brands it is a much bigger one.
The traffic is already paid for. The intent shows up in the first five clicks. Whether you ask is up to you.
Frequently asked questions
Is a 2% conversion rate bad?
No, it sits close to the ecommerce norm. The number worth more of your attention is what percentage of the other 98% left any identifying signal, because that one is usually fixable and rarely measured.
Does this mean conversion rate optimisation is a waste of time?
No. CRO works on visitors who arrived ready to buy. It has almost no effect on visitors who were browsing or researching, which is most of them. Treat it as one job rather than the whole job.
How do I measure my capture rate on non-converting traffic?
Segment paid sessions that ended without an order, then count how many produced an email, a phone number, or a logged preference. Divide one by the other. Most analytics setups will not show this by default, so it usually needs building.
What counts as a high-intent moment?
Any point where the visitor's question is obvious from their behaviour. Sizing and fitting guides, comparison pages, out-of-stock products, and delivery or compatibility questions are the usual candidates in considered-purchase categories.
Lee Marriott is the founder of Convo Commerce and a former ecommerce director who carried the revenue number at UK and US retail brands. Working out what your own non-converting traffic costs and want a second opinion on the figure? That is the conversation I have most weeks.