Verify the data pipeline first (instrumentation bug?), then segment by time, platform, geography, user type, and funnel step to localize the drop, then hypothesize causes and validate each.

Framework: (1) Is the drop real or a logging/tracking issue? (2) When did it start - align with deployments, campaigns, external events. (3) Where - narrow by platform, region, cohort. (4) What - funnel step where users drop, competing metric that rose. (5) Why - correlate with changes. Common causes: broken tracking, product regression, competitor action, seasonality, payment failure, or infrastructure degradation.