Lab Safety for Audience Experiments

Measurement Readiness for Targeting Tests exists because pretty hypotheses age poorly once spend is live. Before you touch spend, write stop rules that a tired manager can apply at 10pm. If your rule requires a PhD to interpret, it is not a stop rule; it is decoration.

Holdouts do not need to be heroic. For SMB campaign sizes, modest holdouts that protect learning are better than theatrical splits that starve both cells. Document the compromise so external reviewers understand what you traded away.

UTMs and naming conventions are the unsung heroes of Social Media Audience Targeting tests. When naming drifts, results become astrology. Tie UTMs to creative versions with boring literal strings — no clever abbreviations your future self will not decode.

End every test with a retro, especially failures. The template is short: hypothesis, what happened, what you will change in the brief next time, and one kindness to the team that held the line.