What problem does it solve? Product decisions are often made from dashboards whose numbers were never audited: wrong denominators, bot traffic counted as users, events that fire on the wrong action, and percentages reported without populations. This Skill runs a structured audit of a funnel or analytics export so you know which numbers can carry a decision and which cannot. ## Core Features & Use Cases - Premise audit and identity resolution: Recomputes every number the user provided from raw data, then defines what counts as one person (device, account, or human) and applies it consistently. - Population cleaning and leak analysis: Removes bots, internal accounts, duplicates, and test data with every filter's row count reported, then ranks funnel losses by people lost per month rather than percentage gaps. - Source reconciliation and tracking plan: Compares two data sources by ID sets, grades every conclusion as MEASURED, OBSERVED, or ASSUMED, and converts every unanswerable question into a concrete event to log. - Use Case: You have an analytics export and suspect your activation rate dropped. Run the audit to discover the drop is mostly one week of bot traffic, that the funnel mixes device IDs and account IDs, and get a tracking plan for what to instrument next. ## Quick Start Ask the assistant to run /product-hygiene on the analytics export in the current folder and report which funnel numbers can be trusted.
