Define a retention cohort with the right clock

Using [EVENT SCHEMA], [TIMEZONE] and [RETENTION QUESTION], define the cohort entry event, qualifying return event, observation window and exclusions. Provide SQL-like pseudocode or formulas, a tiny fictional worked example and a validation table for edge cases. Distinguish calendar weeks from rolling seven-day windows. Do not fabricate actual retention figures or use incomplete cohorts as complete. Check duplicate events, late-arriving data, multiple devices and users with no return. State which business decision each possible metric would answer before choosing one.