A product partner asks, “Is this feature successful?” The request does not specify the user population, behavior, comparison, or decision. Walk through how you would clarify the question and define a metric framework that can support a decision. ### Constraints & Assumptions - No single metric has been selected yet. - The feature may affect different user segments in different ways. - You must propose a useful framework before receiving a complete event schema. ### Clarifying Questions to Ask - What decision will the team make, and by when? - Who is eligible, who is exposed, and what user problem is the feature intended to solve? - What behavior should change, over what time horizon, and what harms must not increase? - Is the team asking about adoption, causal impact, retained value, or operational health? ### What a Strong Answer Covers - A precise decision statement, target population, unit of analysis, and time horizon. - A metric tree separating exposure, adoption, repeated value, and longer-term outcomes. - A primary decision metric plus diagnostic and guardrail metrics. - Definitions, denominator choices, segmentation, instrumentation checks, and an evaluation design. ### Follow-up Questions 1. How would you choose between per-user and per-session metrics? 2. What would you do if adoption rises but retention falls? 3. How would you evaluate the feature if randomization is unavailable?