Every conversion rate optimization best practices article is the same article. Add social proof. Simplify your forms. Improve page speed. Test your call to action.
None of it is wrong. That is the problem. A practice that everybody already follows is not an advantage, it is a floor. The list tells you how to stop losing. It does not tell you how to win, and it certainly does not tell you which item to spend your limited traffic on.
So here is the version I would actually hand a client: the practices that hold up, the widely repeated ones I would argue against, with the opposing view attached where there is one, and the times I made the argument and lost.
Key Takeaways
- Best-practice lists describe the floor. Diagnosis, not the list, decides what you fix first.
- Trust badges are not automatic. Sometimes they raise trust, sometimes they cheapen a premium design.
- Manufactured urgency is a negative signal, not a neutral one.
- Colour does not convert. Anchoring order does.
- Below roughly 200 to 250 conversions and 5,000 sessions per variation, you cannot test your way to an answer.
- The most useful thing in my own files is the list of arguments I lost.
The ones that hold up
Short list, because the honest one is short.
Three benefit bullets on every product card. This is the single most repeated tested result in my own corpus: a 10 to 20% conversion lift, A/B tested several times across different stores. Not on the product page, on the card in the grid, before the click. A shopper scanning a collection is choosing what to open, and a name plus a price gives them nothing to choose on.
Clarity of hierarchy beats everything cosmetic. Most stuck conversion rates are structural, not surface. The order of the page, what the eye meets first, what question gets answered before the next one is raised. That is the layer where the number moves.
Speed is not post-design polish. A slow first byte and a scroll that flashes white are conversion problems that arrive before any of your copy does.
One primary action per screen. Every secondary call to action that offers a lesser conversion siphons intent away from the one you want. A soft secondary that leads deeper into the same decision is the exception, and I build one by default.
That is roughly it for things I would recommend without asking about your store first.
The ones I would argue with
Most of these have an opposing view held by competent people, and where there is one I have written it down next to mine. If you only read the contrarian half, you are doing the same thing as the checklist, just in the other direction.
Trust badges are not automatic. The opposing view: the near-universal advice to add trust badges everywhere. Sometimes credit card icons raise trust. Sometimes they make a premium design look cheaper, and on one store where the design was doing the reassuring, I chose not to add them. Both outcomes are real, which means this is a test, not a rule.
Manufactured urgency is a negative trust signal. Not neutral. Negative. Buyers have been trained by a decade of countdowns that reset on refresh, and a fake deadline now costs you credibility on everything else on the page. Real scarcity still works. The distinction is whether the thing you are claiming is true.
Quick-add is a metric trap. The opposing view: fewer clicks means more conversions, which is usually sound. In my A/B tests, sending shoppers to the product page consistently beat quick-add tiles on both conversion rate and average order value. Quick-add optimises the metric named on the button. The product page is where the decision is actually made.
Colour does not convert. Anchoring order does. The opposing view: button and price colour testing, which is the most-run test in the industry. Colour barely moves the number. The order in which you present prices moves it a lot. If you have enough traffic for exactly one test this quarter, do not spend it on a shade of green.
Static beats a carousel in the hero. Carousels drain motivation at the moment you need it highest, because the visitor has to wait for content rather than consume it. You want them moving down the page, not standing still watching yours move.
Video usually loses to a static screenshot of the interface. A video raises questions the page then has to answer: how long is this, do I have to pause it to read the detail, is it worth the wait. It also costs you load time. There are products where video is the only way to show the thing. There are fewer of them than the advice implies.
More social proof is not automatically better. Dropping a testimonial block between two sections that were flowing into each other reads as a bolt-on, and a bolt-on can be net negative. Social proof placed where the doubt actually occurs beats social proof placed where there was room.
More whitespace has a hidden cost. More space means more scrolling, which means fewer people reach the end. Worse, generous whitespace can read as the end of the page, so visitors stop before content you spent weeks on.
Drop the trailing decimals. "$30" reads as a smaller number than "$30.00", which at a glance can even scan as three thousand. Small, free, and almost never on a best-practice list.
Three times I was wrong
The reason to publish this section is simple: nobody else does. Agency content carries wins. My own files carry fourteen documented arguments I lost or conceded, and they are more useful than the wins because they show where a general principle meets specific knowledge and loses.
The benefits section. A longevity supplement brand had what I considered a generic benefits image, and I proposed a cleaner replacement. The client showed me that my version hid the descriptors behind the titles, killed the ingredient animation that made the claim believable, and was harder to read. Their version stayed. I had optimised for tidiness and called it clarity.
My dark-background fix that failed. On a conversion agency's own site, I built a dark section I was pleased with. The client reported eye strain and visual clutter. I tried tightening the grouping first, which did not solve it, and then conceded that the font size on a dark background was the actual problem and reverted to the previous arrangement with larger icons. The lesson was that I had treated a legibility complaint as a layout complaint because layout is what I wanted it to be.
The layout I could not defend. A pain-relief supplement brand asked me for the conversion data supporting a sticky-gallery variant layout I had proposed. I did not have any. It is unconventional, we had not tested it, and the honest answer was that moving the text further down probably made more sense for now. Being asked to show your evidence is not an obstacle to good work. It is the job.
None of those changed my view that structure beats surface. All three changed how confidently I state a principle to somebody who sells the product every day.
The floor that decides whether any of this is testable
Before you test a best practice, check whether you can.
The working floor is roughly 200 to 250 conversions and 5,000 sessions per variation, more for split URL tests. Below that, even a large observed lift risks being a false positive, and you will act on noise with total confidence. My own upsell testing uses its own floor of 200 users per variant, because fewer than that rarely produces a winner that holds.
And when you do have a result, pick the winner on profit, acceptance rate, revenue per visitor or margin-adjusted average order value. Never on raw sales. Raw sales is the number that lets you buy a lift with margin and call it a win.
If you are below the floor, that is not a reason to stop working. It is a reason to stop calling it testing. Fix the friction you can document, measure a clean before and after over a window you set in advance, and be honest in writing that it was not a controlled test.
What to do instead of working the list
Diagnose, then decide, then fix one thing.
- Find where your store actually loses people. Funnel report, session recordings, and your own support tickets. The tickets are underrated: they are customers telling you in writing what confused them.
- Rank your problems, not everyone else's solutions. A best-practice list is a generic answer sheet. Your funnel is the question.
- Spend your traffic on the highest-leverage surface. The header is a reasonable candidate for most stores, since every visitor sees it on every page. I have run around fifteen tests on a single header bar for that reason.
- Write down what you did not test. The most credible sentence in this whole field is "we have not tested this yet."
The wider version of this, applied specifically to Shopify stores, is in shopify conversion rate optimization. The research-versus-reality version, covering the lifts that were real and unprovable, is in ecommerce conversion rate optimization.
And if what you want is not a list but a ranked set of your own problems, that is what an audit is for. Book a call and we will start with where your store is actually losing people.
