Review this project's email campaign using marketing/strategy.md, marketing/measurement.md, and the aggregated results I provide. If results are missing, create a fill-in report and ask for counts; do not invent performance.
Show the reporting dates, campaign and cohort, eligible/enrolled contacts, unique delivered contacts, qualified replies or activations, paid conversions, bounces, complaints, and unsubscribes. Calculate rates only with named, valid denominators and distinguish missing data from zero. Do not include raw recipient addresses in committed reports.
Read anonymized reply examples if available. Separate targeting, offer, copy, delivery, and product-friction hypotheses. Do not infer campaign lift from before/after counts alone, and do not call a small sample conclusive.
Save marketing/review.md with what happened, what remains uncertain, and exactly one next experiment: hypothesis, change, primary metric, observation window, eligibility, stopping conditions, and a decision rule appropriate to the sample size. Investigate any delivery or suppression failure before proposing more volume.
Draft the proposed change locally and explain it. Do not activate it, send messages, or change the live campaign. Keep the original experiment record so the result remains interpretable.