I started with funnel breakdown which was the right instinct, but I jumped to hypotheses too fast without first confirming the data itself was clean.
Start by clarifying the metric definition, time frame, and expected vs. actual movement to ensure alignment. Then systematically rule out data issues, segment the metric to localize the change, and form hypotheses about root causes. Finally, prioritize hypotheses, test them with data, and recommend actions based on impact and confidence.
Pro tip: Always validate the data pipeline and metric definition first—many 'metric movements' are actually instrumentation or logging errors. Also, quantify the impact in business terms (e.g., revenue) to prioritize your investigation.
Confirm the metric definition, time period, and expected value. Check for data quality issues, pipeline failures, or tracking changes that could explain the movement.
Break down the metric by dimensions (e.g., platform, geography, user cohort, product category) to identify where the change is concentrated.
Generate plausible explanations based on business context, recent changes, and external factors. Consider internal (e.g., site changes, pricing) and external (e.g., seasonality, competitors) drivers.
Use data analysis, A/B tests, or causal inference methods to validate or refute each hypothesis. Prioritize by potential impact and ease of testing.
Based on findings, recommend corrective actions, further monitoring, or deeper investigation. Communicate impact and next steps to stakeholders.
AI-generated suggestions, not part of the candidate's original notes. May be inaccurate — verify before relying on them.