Reading a pricing experiment that looked flat

Context

A pricing test showed no overall movement and was about to be killed. I was asked for a second read before the team walked away.

Approach

1. 01Re-checked power and guardrail metrics before touching any segment cuts.
2. 02Decomposed the flat aggregate into pre-registered segments to avoid p-hacking.
3. 03Surfaced a clean, robust lift in one segment and a small drag in another.
4. 04Told the story as a trade-off the team could decide on, not a single verdict.