How would you investigate a metric decline?

## Scenario You are a Data Scientist supporting a consumer marketplace product. A key business metric (e.g., orders/day, conversion rate, revenue, profit, retention) has **declined noticeably** over the last 1–2 weeks. ## Task Describe a structured approach to investigate the decline. ### Requirements In your answer, cover: 1. **Clarify the metric** - Exact definition, numerator/denominator, filters, attribution logic. - Timezone, event time vs processing time, and whether the metric is cohort-based. 2. **Detect whether it’s real vs measurement** - Data pipeline issues, logging changes, bot/traffic anomalies, backfills. 3. **Localize the change** - When did it start? step vs trend? aligned to releases/experiments? 4. **Decompose and segment** - Break down by funnel stage and by key segments (new vs returning, geo, device, acquisition channel, product surfaces, supply vs demand side, etc.). 5. **Consider external and seasonal factors** - Seasonality/holidays, competitor actions, macro events, weather, outages. 6. **Form hypotheses and propose follow-ups** - Additional analyses, dashboards, and (if appropriate) experiments or mitigations. ### Output Explain the investigation plan and what you would do in the first day vs first week.