Who asked: Lena Brandt, finance lead at Fenwick Market, and the checkout team.
Why: the team ran a new one-page checkout (B) against the current one (A) and wants to ship it before the holidays.
The data: ab, one row per session: the variant shown, whether it ordered, and the revenue.
Done looks like: a checked split, the conversion rates and lift, a test and an interval, the revenue guardrail, the test's power, and a decision that follows the rule the team agreed before the test.
Work through the readout in the order a careful analyst would: trust the split first, then the primary metric, then the guardrail, then decide. Each step keeps the variables from the steps before.