A conversion programme without an order is a queue of opinions. The team with the loudest voice sets the roadmap, and nobody can say afterwards whether it worked.
Every term used above is defined in our experimentation glossary, and three experiments taken apart step by step sit in the case studies. Longer pieces with the workings attached are in the field notes.
Funnel analysis first: how many arrive, how many reach each step, how many finish. The biggest drop is almost never where the meeting thinks it is, and it takes an afternoon to find rather than a quarter to argue about.
Do this: Draw your funnel with real counts at each step this week. The largest single drop is your first candidate, whatever anybody prefers.
Session recordings and form analytics find friction far faster than testing does. Five sessions surface most usability problems. They prove nothing about magnitude, which is what the test is for.
Do this: Watch five recordings of people who abandoned at your biggest drop. Write down what you saw, not what you concluded.
Send traffic and baseline rate. We reply with the smallest effect your test can actually detect. One email, no call.
Traffic and baseline rate give you a minimum detectable effect. If that number is larger than the improvement you realistically expect, the test cannot answer the question and building it is wasted work.
Do this: Compute the minimum detectable effect first. If it exceeds your expected lift, ship on judgement instead and say so openly.
Tests that show no difference are the cheapest knowledge you will ever buy, and they are the ones teams delete. Without an archive you repeat the same idea every eighteen months when staff change.
Do this: Write down every completed test with its result, including the flat ones. One row each is enough.
This is teaching material and our own reading of standard practice, not advice for your specific site. Check anything important with your own specialist before you act on it.
Traffic, baseline rate and what you changed. We reply with the minimum detectable effect and whether the result meant anything. No cost, no call.