Building a Launch Success Readout for a New Google Product From Log Data

You are a product manager at Google, and a new product has just launched. Explain how you would use logs to decide whether the product is succeeding. Cover how you would define success, which signals you would extract from the logs, and how you would turn those signals into a judgment. ```hint Success comes before signals Decide what the product must achieve for its users and for Google first, then work out which logged events can stand in for each outcome, and what the logs cannot tell you. ``` ### Clarifying Questions - Which logs are available: client-side usage events, backend request and error logs, or both? Can we add instrumentation, or must we work with what is already recorded? - What does the product do, who is it for, and did the launch plan set any success targets? - Was the product released to everyone at once, or gradually with a holdout or experiment group? - Is it a standalone product, or a feature inside an existing Google product whose users it shares? ### What a Strong Answer Covers - A definition of success tied to the product's purpose, expressed as a primary metric with guardrails. - A mapping from success dimensions (adoption, activation, engagement, retention, reliability) to concrete logged events, identifiers and computations. - Data-quality and privacy checks on the logs before any conclusion is drawn. - A way to separate the launch's real effect from noise: baselines, holdouts, cohorts and segment cuts. - A decision rule that turns the readout into scale, iterate or stop. ### Follow-up Questions - Adoption looks strong, but week-four retention is far below a comparable earlier launch. What do you investigate? - The only logs available are backend API calls with no user identifiers. Which success measures can you still compute, and which can you not? - How would you tell whether the new product is taking usage away from an existing Google product?