CRO advice assumes a broken funnel. When your store already converts well, the wins move elsewhere: segment gaps, high-value journeys and revenue per session.
Most conversion rate optimisation advice assumes a broken funnel. When your store already converts at or above benchmark, that advice stops working, and the remaining gains sit in four places: segment gaps, high-value journeys, revenue per session rather than rate, and the cost of the tests themselves.
This is the situation we are called into most often with established retailers. The obvious problems were fixed years ago. The checkout is short, the site is fast, the trust signals are there. Conversion rate has been flat for three quarters and the team has run out of ideas, because every list they read is a list of things they have already done.
Why the standard playbook runs out
The generic CRO checklist is a list of defects. Fix the slow page, shorten the form, add the reassurance, remove the surprise delivery cost. Every item on it moves the number when the defect is present and does nothing when it is not.
Once the defects are gone, the average gets you nowhere, because the average is hiding the thing you need. A store converting at 3 per cent overall might be converting at 6 per cent on returning desktop visitors in one category and 0.8 per cent on new mobile visitors in another. There is no site-wide change that improves both, and the site-wide metric will barely move whatever you do.
The shift is from optimising a page to finding the segment where the gap between what you get and what you should get is largest.
Where the remaining gains actually are
Segment gaps
Cut conversion rate by device, by traffic source, by new against returning, and by category. Then look for the cells where a segment converts far below a comparable one. That is your work queue, in order of revenue at stake.
The most common finding on a healthy store is a category or product type that converts well below the rest for a structural reason: the imagery is weaker, the sizing information is thinner, the delivery expectation is different, the price point crosses a threshold where buyers want more reassurance. None of that shows up in an aggregate number.
High-value journeys rather than high-volume ones
Testing usually gravitates to the highest-traffic pages, because that is where significance arrives fastest. On a store that already converts, the money is often in a lower-traffic, higher-value journey: the multi-item basket, the first-time buyer of a considered product, the customer who reaches a comparison page. Those journeys are slower to test and worth more per point of improvement.
Revenue per session, not conversion rate
Conversion rate is a proxy that stops being useful at this stage, because you can lift it while making less money. Discount prominence is the classic example. Optimise for revenue per session, and hold average order value and margin in view alongside it, or you will run a successful test programme that quietly trades margin for volume.
The cost of testing itself
At a 3 per cent baseline and realistic traffic, detecting a 5 per cent relative improvement takes a long time. A lot of teams run underpowered tests, call them at a fortnight, and act on noise. Two honest options: test fewer, bigger changes, or accept that some decisions will be made on judgement and qualitative evidence rather than statistical significance. Pretending a two-week test on 400 conversions settled anything is the worst of both.
What we look at first on a store that already converts
The order matters, because each step narrows the next.
Segment the conversion data before touching anything, and rank the gaps by revenue at stake rather than by size of gap. A five point gap on a category doing £8,000 a year is interesting. A one point gap on a category doing £400,000 is the job.
Then read the qualitative evidence for the worst segment specifically. Session recordings, on-site search terms, customer service tickets, returns reasons. On a healthy store the answer is rarely a usability defect and often an information gap: buyers cannot tell whether the thing will fit, arrive in time, or suit their situation.
Then check whether the problem is conversion at all. Traffic quality is frequently the real culprit. A category that converts badly because paid search is buying broad, poorly-matched terms is a media problem wearing a CRO costume, and no amount of page testing will fix it. That is one of several reasons we run organic and paid search as one programme rather than two.
Then, and only then, design the change. On a store with a sound baseline, the changes that pay are usually additive rather than corrective: better product information, comparison support, delivery clarity by postcode, sizing confidence, stock transparency.
Two patterns worth knowing
The first is that the biggest single lever on a converting store is often not on the site at all. It is what happens after the first order. Retailers who fix the second-purchase journey usually find more revenue there than in any on-site test, and it is measured with the same discipline.
The second is that the information that improves conversion is increasingly the same information that gets you cited by AI assistants. Specific delivery terms, stated in text. Sizing and fit detail. What the product does not suit. Both a hesitant buyer and a language model are trying to answer the same question, and pages that answer it plainly do better with both. The mechanics of that are in our AI search optimisation guide.
Questions we get asked
What is a good ecommerce conversion rate?
The honest answer is that benchmarks are close to useless at this stage, because they aggregate across price points, categories and traffic mixes that have nothing to do with each other. A £15 impulse product and a £900 considered purchase should not converge on the same number. Your useful benchmark is your own performance by segment over time.
How much traffic do I need to run a valid test?
It depends on your baseline and the size of the effect you want to detect, and the arithmetic is unforgiving. As a rough guide, detecting a small relative improvement on a low baseline needs tens of thousands of sessions per variant. If you do not have that, test bigger changes, run them longer, or make the decision on evidence other than a split test and be honest that you have done so.
Is CRO worth it if my conversion rate is already good?
Usually yes, but the framing changes. You are no longer buying a lift in a site-wide rate. You are buying revenue from specific segments and journeys that underperform their potential, and the business case should be written that way.
Should I fix conversion or buy more traffic?
Whichever is cheaper per pound of incremental revenue, which you can estimate rather than guess. On a store already converting well, more traffic is often the better marginal investment, which is an unusual thing for an agency to say and is true often enough to be worth saying. Our piece on SEO against PPC as investments covers how we weigh that.
How does CRO interact with SEO?
More than most teams assume. The content that helps a buyer decide is the content that earns the ranking and the citation. Pages that answer questions plainly convert better and rank better, which means the two disciplines usually agree. Where they conflict, it is normally because someone has added interstitials or hidden content behind interactions.
Where this leads
The pattern on established stores is consistent. The aggregate number moves slowly and the segment numbers move a lot. Teams that keep chasing the aggregate conclude that CRO has stopped working. Teams that go segment by segment keep finding money.
One retail partner we work with now sees over 63 per cent of total website revenue through the channels we manage, and £1.4m of tracked revenue in a single year came from that discipline rather than from any single test. On the ecommerce side, one beauty retailer grew online revenue by 400 per cent in the first year, with product-level detail doing much of the work.
If your store converts well and the number has stopped moving, that is usually a measurement problem before it is a design problem. Our conversion rate optimisation service sets out how we approach it, the CRO audit framework covers the diagnostic, and you can get in touch for a senior read on where your remaining gains sit.