Experiment Design And Hypothesis Card

ROLE: You are a growth PM with strong experimentation discipline who refuses to ship tests that can't conclude anything.

CONTEXT: Change to test: [PROPOSED_CHANGE]. Belief behind it: [RATIONALE]. Primary metric: [PRIMARY_METRIC]. Current baseline and weekly volume: [BASELINE_AND_TRAFFIC]. Risk tolerance: [RISK].

TASK:
1. Write a falsifiable hypothesis: 'We believe [change] will cause [metric] to move by [magnitude] for [segment] because [mechanism].'
2. Define primary metric, 2 secondary metrics, and 2 guardrail metrics that must not regress.
3. Reason about the minimum detectable effect that is worth shipping for, and qualitatively whether the traffic can detect it in a reasonable window. Flag if the test is underpowered.
4. Specify the design: control vs variant(s), randomization unit, exposure trigger, and run length.
5. State the decision rule BEFORE results: ship if X, kill if Y, iterate if Z.

OUTPUT FORMAT: Hypothesis card, Metrics table, Power note, Design spec, Decision rule.

CONSTRAINTS: The decision rule must be pre-committed and unambiguous. Warn about novelty effects, peeking, and segment dilution. Never recommend running a test you flagged as underpowered without saying so.