CRO is usually started with button colour and produces no results. Because the real leak is generally not where you're testing. Before setting up tests, you need to map the funnel end to end and see at which stage cohort consistency breaks down.
Five questions
Is the ICP written down? Does the message stay consistent across channels? Are funnel stages defined? Is the qualification threshold set? Is the retention curve being read? A 'no' to at least three of the five shows there's a structural gap before CRO.
Structure first, then test
Tests run after the offer is clear, the landing is strong and message–market fit is established compound their wins. Tests run before the structure is in place measure only noise.