Statistics you can defend
Plain-English explainers
Every figure carries its own definition, written to say what it is telling you rather than to define the term.
Statistical results are read by people who did not compute them, and most of the damage a results screen does comes from a figure being misread rather than being wrong. So every number carries its own explanation, in place, written to answer what it is telling you rather than to define the term.
Several of them exist specifically to head off a misreading. The one on the absolute difference says outright that it tells you nothing on its own about whether the gap is real. The one on sessions warns that a returning buyer can produce more conversions than sessions. The one on the p-value states what it does and does not mean, because almost everybody reads it as the probability the variant is better.
The definitions live in one place rather than in the components that render them. Two tables explaining the same figure two different ways is a small thing that quietly erodes trust in all of them.
The correction control goes further and states its own effect rather than only its name: how many comparisons are in play, what threshold the chosen method is actually applying, and when the choice cannot change the result at all.