Reproducible Cohort Retention Analysis Recipe

CopyAct as a product analyst designing a cohort retention study. CONTEXT: the event data has columns [EVENT_TABLE_SCHEMA], a user is identified by [USER_ID_FIELD], the activating event is [ACTIVATION_EVENT], and the retention event is [RETENTION_EVENT]. The grouping I care about is [COHORT_DIMENSION: signup_month/plan/channel]. TASK, in order: 1) Pin down precise definitions: what counts as the cohort, the activation moment, the retention window, and whether retention is bounded or unbounded. 2) Write [SQL_DIALECT] to build a cohort-by-period retention matrix. 3) Specify how to render it as a triangle heatmap and which cells signal trouble. 4) Give an interpretation checklist: how to separate a real drop from a seasonality or sample-size artifact. 5) Draft three falsifiable hypotheses for any decline. CONSTRAINTS: state every assumption explicitly; never compare cohorts with fewer than [MIN_COHORT_SIZE] users; keep period math timezone-safe. OUTPUT FORMAT: Definitions, SQL (code block), Visualization notes, Interpretation checklist, Hypotheses.