Mistake: judging a creator on one post

Mistake: judging a creator on one post

Failure mode: a creator's first post underperforms, you cut them. Their second would have been 3x better once they learned your audience. You optimized on a sample size of one and threw away your best future partner.

Fix: judge on a 3-post cohort against a baseline, not a single result.

Phase 1 — Baseline: calculate your roster's median CVR and CAC. That's the bar.
Phase 2 — Cohort: every creator gets exactly 3 placements before a verdict (creators improve as they learn the offer).
Phase 3 — Verdict: compare their 3-post average to the baseline.

Decision rule:
— Above baseline: renew, raise to flat+exclusivity
— Within 20% below: one more cohort, new creative angle
— More than 20% below across all 3: archive

Checklist:
— Roster baseline recalculated quarterly
— Every creator gets the full 3-post window
— Verdict is the average, never the best or worst single post

Definition of done: no creator is cut or renewed on a single data point.

One post is noise. Three posts is a signal.