Diagnose March Uber ride-volume drop

Uber observes a March ride-volume drop. After normalizing for days-in-month, completed trips are down 14% vs February and 9% vs the prior-year March. Design a diagnostic plan to (a) confirm the drop is real (not instrumentation/definition drift), (b) decompose it into demand-side vs supply-side drivers, and (c) quantify the top two root causes. Be specific: 1) Define the primary and guardrail metrics (e.g., request rate, conversion to accepted, pickup ETA, cancellation rate by initiator, completion rate, price indices, surge exposure, wait-time distribution, session-to-request funnel). 2) List the exact data points and cuts you’ll check first: by city/region, hour-of-day, weekday/weekend, rider/driver cohorts (tenure, loyalty), product type (UberX/Pool/Lux), price buckets, promo exposure, surge bins, weather, events/holidays (DST shift, spring break), airport vs non-airport, device/app version, major app releases. 3) Describe how you’d separate demand shrinkage from supply constraints (e.g., rides=requests×P(accept)×P(complete); use pickup ETA/cancellation elasticity, driver online minutes, active driver count, queue depth). 4) Specify statistical methods to establish “significant drop” and attribute causes (e.g., interrupted time series with seasonality/holiday controls, difference-in-differences across unaffected control cities, synthetic control). 5) Outline checks for measurement artifacts (schema changes, event loss, sampling, timezone/DST alignment, deduplication). 6) Propose 1–2 quick experiments or natural experiments to validate a suspected cause (e.g., restore prior promo budget in matched markets; temporarily cap ETA via supply incentives) including success metrics, MDE, and duration.