If everything you have tried in CRO has not moved revenue yet, it is rarely because CRO does not work for your brand. It is usually because the work has gone into the wrong place: small tweaks instead of the offer, one page for every ad, ideas with no research behind them, tests judged on the wrong number, and a page that is slow before anyone reads a word.
1. You are testing tweaks, not the things that sell
Button colours, badge positions, a new font on the add to cart. These are the tests most brands start with because they are quick to set up. They are also the tests least likely to move revenue, because a shopper rarely leaves over a button colour.
Shoppers leave because they cannot see why your product beats the one in the next tab, because the offer is unclear, or because the page talks about the product and not about them. That is positioning and offer, and that is where the big swings are.
Penrose Skin is a good example. We removed the one time versus subscription choice from the product page, so the only decisions left were scent and offer. The same traffic added to cart more often, and the subscription take rate went up by 27.7%. That was not a tweak. It changed the decision the shopper had to make.
- Is the offer on your page different from what a competitor would show?
- Can a shopper say in one sentence why your product is the better buy?
- Are your next three tests about the offer or the message, not the styling?
2. The page does not match the ad that sent the traffic
Every ad makes a promise. When the page it lands on says something different, the shopper has to work out whether they are in the right place, and plenty of them do not bother. You can run perfect tests on that page and still lose most of the traffic in the first second.
Treat each ad angle as its own way in. Every persona arrives for a different reason and knows a different amount about you, so each one gets the page type built for that job: a product page whose first screen repeats the ad’s promise, or an advertorial or a listicle for colder traffic that needs convincing before it sees a price.
When we rebuilt Hound Nutrition’s product page, we connected the journey between traffic source and page, so the offer, proof and benefits became the focus above the fold. The published results were a 28.9% increase in conversion rate and a 39.4% increase in revenue per visitor. More on that in what belongs above the fold on a Shopify product page.
3. Your tests are guesses, not research
A test is only as good as the idea behind it. If the ideas come from a list of best practices or what a competitor changed last month, you are guessing, and guesses lose more often than they win.
The ideas that win come from your own customers: the words they use in reviews, the objections they raise in surveys and support tickets, the reasons they give for buying and for not buying. That language goes straight into headlines, benefits and the questions the page answers before anyone has to ask.
For Zoomie we went deep into their customer personas, pulling key language and objections from their real reviews and stories, then rebuilt the post click experience around them. After 9 days of testing, conversion rate was up 11.4% and revenue per visitor up 17%.
4. You are judging tests on the wrong number
Conversion rate on its own can lie. A change that turns more visitors into smaller orders can lift conversion rate and lower revenue. A stronger bundle can lift revenue while conversion rate barely moves. Judge tests on revenue per visitor as well as conversion rate, and you stop shipping winners that quietly cost you money.
Then give tests time. Stopping a test the moment one version pulls ahead is how good ideas get called losers. And judge pages on how long they take to win, not whether they do. Ours have a 100% win rate: a third win outright on the first test, just over a third within 30 days, and in rare cases some take a few more weeks. A test built on research tells you exactly what to change next. A test built on a guess tells you nothing.
5. The page itself is working against you
Page builders are quick to put up and expensive to live with. They add weight, they fight your theme, and they leave you renting the page your paid traffic lands on. Stack a few apps on top and the first screen, the part that has to load fastest, is the slowest part of the page.
No test result survives a page that loads late on a phone. That is why we hand code pages in custom Liquid in your theme, with no Replo, no Shogun and no PageFly, and build around the apps you already run, like Recharge, Loop, Seal, Kaching, Judge.me and Trustpilot.
- Does the first screen load quickly on a phone on mobile data?
- Is the page your paid traffic lands on built in your theme, or rented from a builder?
- Does every app on the page earn its place?
What else do people ask?
Long enough to cover at least one full buying cycle and enough orders to trust the result, which for most brands means weeks, not days. Stopping a test the moment one version pulls ahead is how good ideas get called losers and bad ones get shipped.
Testing needs volume. We work with DTC brands doing at least £250k per month, because below that the test volume is not there to move anything quickly. Smaller brands usually get more from fixing the foundations first: the offer, the match between ad and page, and a fast page.
Start with the first screen of the page your paid traffic lands on: the promise from the ad, the offer, the proof and the core benefits. Those four jobs decide whether anyone reads the rest of the page.
