A test that moves conversion by 0.4% needs an implausible amount of traffic to distinguish from noise. A test that moves it by 12% resolves in a fortnight. Most CRO programmes stall not because the statistics are hard, but because they keep proposing changes too small to be detectable.
The fix is not more tests. It is testing a bigger idea, and bigger ideas come from research rather than from a best-practice checklist.
Where the real objections hide
Five sources, in descending order of how often they are ignored:
- 01Support tickets. Every pre-purchase question that reaches a human is a question your page failed to answer. Read a thousand of them and the ranking writes itself.
- 02Post-purchase surveys. Ask new customers what nearly stopped them. People who bought will tell you what the people who left could not.
- 03On-site exit polls. One question, triggered on exit intent on high-value pages. Crude, cheap, and consistently revealing.
- 04Review sentiment, including competitors' one-star reviews. The fears expressed there are the fears your page must pre-empt.
- 05Session replay, watched with a question in mind rather than as ambient entertainment. Where does the cursor hesitate? What gets re-read?
Turning that into a ranked map
Group what you find into distinct objections, attach an estimated volume to each, then attach a confidence score to whether your page currently addresses it. The top three items on that list are your next three tests, and they will not be about button colour.
Then test it properly
Pre-register the hypothesis and the primary metric before launch. Run a power calculation so you know the detectable effect and the runtime in advance. Use sequential testing so you can stop early without inflating false positives. And read results in revenue per session, not conversion rate, because a conversion win that quietly sinks average order value is a loss in a costume.
One more discipline that pays for itself: keep a decision log of every test, including the losers, with the hypothesis and what you concluded. After a year that log is a genuine competitive asset, because it describes your specific customers rather than a generic buyer someone wrote a blog post about.