Walk into most optimisation programmes and you will find a lot of motion. Button colours tested, headlines swapped, popups added, a steady stream of small tweaks that fill a roadmap and a status report. What you will rarely find is a meaningful change in revenue, because most webshop optimisation is busywork dressed as progress. It optimises metrics that are easy to move and ignores the decision that actually determines whether someone buys.
The distinction matters enormously. Optimising for a metric means chasing whatever number is convenient to improve. Optimising for the decision means understanding why a customer hesitates and removing that specific reason. The first generates activity and a tidy report. The second generates revenue. They look similar on a roadmap and produce completely different results.
Why the busywork feels productive
Small tests are appealing because they are safe, easy and endless. There is always another element to tweak, the work is never done, and every change can be reported as optimisation. It feels rigorous because it is busy. But activity is not the same as impact, and a programme can run for a year, ship dozens of changes, and leave the conversion rate exactly where it started.
The reason is that most of these tests address things that were never the problem. The customer did not abandon because the button was the wrong shade. They abandoned because they could not find the answer to a question, did not trust the returns policy, or hit friction at checkout. Testing the button while ignoring the question is optimising the wrong thing with great discipline.
Start from the customer's decision
Real optimisation starts by understanding the decision the customer is trying to make and what is getting in the way. That comes from evidence, not opinion: watching real sessions, reading where people drop out, listening to what they tell support, finding the moments where intent turns into hesitation. Those moments are where the revenue is, and they are usually nowhere near the elements teams instinctively test.
Once you can see where the decision breaks down, the work becomes obvious and the tests become worth running. This is hypothesis-led conversion work: each test exists to answer a real question about customer behaviour, not to fill a slot in the calendar. Fewer, sharper tests aimed at genuine friction beat a high volume of cosmetic ones every time.
A test for whether a test is worth running
Before any optimisation goes on the roadmap, ask three questions. First, what customer problem does this address, and how do we know it is real? If the answer is a hunch, it is busywork. Second, if this wins, will it move a number the business cares about, or just one that is easy to measure? Third, what will we learn even if it loses? A good test teaches you something about your customer either way. A bad one just consumes a fortnight.
Run every proposed change through that filter and the roadmap shrinks dramatically, which is the point. What remains are the few changes that target real friction at decisive moments, and those are the ones that actually move revenue. The discipline is in saying no to the comfortable tests, not in running more of them.
Opinion is the enemy of optimisation
The other thing that fills optimisation roadmaps is opinion. The loudest person in the meeting decides the homepage should look different, a stakeholder dislikes a colour, someone saw a competitor do something, and these preferences quietly become the test queue. None of it starts from the customer, and most of it is settling internal arguments with the conversion rate as referee. It feels collaborative and it is almost entirely a waste of testing capacity.
The discipline that fixes this is simple to state and hard to hold: every test starts from evidence about customer behaviour, not from internal preference. If a proposed change cannot point to a real signal that customers are struggling with something, it does not get tested, however senior the person proposing it. That rule is unpopular precisely because it takes the steering wheel away from opinion, which is exactly why it works.
This also changes the culture around being wrong. When tests are framed as experiments to learn from rather than opinions to vindicate, a losing test is useful information rather than a personal defeat. Brands that get there stop arguing about taste and start learning about customers, and the optimisation programme finally compounds because it is building knowledge instead of relitigating preferences.
Significance and patience
The other failure of metric-led optimisation is impatience. Tests get called early on tiny samples, noise gets mistaken for signal, and the programme lurches from false win to false win without ever compounding. Proper optimisation respects how long it takes to know something is true, especially for brands without enormous traffic, and it resists the urge to declare victory before the numbers have earned it.
This is unglamorous and it is exactly why it works. The brands that improve conversion durably are not the ones running the most tests, they are the ones running the right tests properly and believing the results. It is the kind of disciplined approach we brought to conversion work for brands like Grenade.
Traffic quality beats page tweaks
There is a deeper version of this mistake worth naming. A great deal of optimisation effort goes into squeezing the conversion rate of traffic that was never going to convert, while the question of whether you are attracting the right visitors in the first place goes unasked. No amount of on-page testing rescues a page being shown to the wrong people. The friction is upstream, in the targeting and the promise that brought them, not in the layout that greeted them.
This is why genuine optimisation cannot be walled off from acquisition. If a campaign is drawing price-led bargain hunters to a premium product, the page is not the problem and no test will fix it. The brands that improve conversion meaningfully look at the whole path, the promise made in the ad, the expectation it set, and whether the page honours it, rather than treating the page as an isolated unit to be tuned.
Seen this way, optimisation is less about the store and more about coherence: does what you promised match what you delivered, for the people you actually attracted. When those line up, conversion takes care of much of itself. When they do not, you can test forever and move nothing, because the gap was never on the page you were optimising.
Optimise for revenue, not motion
The shift is from measuring optimisation by how much you did to measuring it by what it earned. A programme that shipped three changes that moved revenue beats one that shipped thirty that did not, even though the second looks busier in every report. Judge the work by impact and the busywork falls away on its own.
If your optimisation programme is generating plenty of activity and little revenue, the problem is usually that it is aimed at metrics rather than decisions, and at internal opinion rather than customer evidence. A focused CRO consultation is the quickest way to point the effort at the friction that actually costs you sales, and away from the tweaks that only feel like progress.
