Metric Diagnosis: Tracing the Cause of a 10% DAU Fall on Google Maps

You are the product manager for Google Maps, and the daily active user count has fallen by 10%. Explain how you would figure out why it happened and what you would do about it. ```hint Locate before you explain Before guessing at causes, make sure the drop is real and pin down when it started, where, on which surface and for which users. Each of those cuts rules out whole groups of explanations at once. ``` ### Clarifying Questions - Which surface is affected: the consumer app on Android or iOS, the web, in-car use, or the developer platform that other businesses embed? - Over what window is the 10% measured (day over day, week over week, or year over year), and did it happen at once or as a gradual slide? - How is a daily active user defined: any app open, or an open plus a meaningful action such as a search or a started navigation? - Is the drop global or concentrated in a few countries, and are other Google products seeing a similar decline? ### What a Strong Answer Covers - Validating the number first: the active-user definition, logging or pipeline changes, and a cross-check against an independent data source. - Segmenting the drop by time, geography, platform, app version, entry point and user type to localize it. - A structured set of internal causes (releases, bugs, latency, outages, changes in other products that send people to Maps, marketing) and external ones (seasonality, major events, OS updates, competitors, regulation, press), checked in a sensible order. - A way to confirm the leading hypothesis with data rather than timing alone. - Remediation and communication: the fix, measuring recovery, stakeholder updates and a post-mortem with prevention. ### Follow-up Questions - The drop is limited to iOS users in one country and began the day a new OS version shipped. What do you do next, and whom do you involve? - Daily active users are down 10%, but navigation sessions per user went up. How do you read that? - After the fix, only half of the lost users return. How do you decide whether the rest are gone for good?