Diagnose why average waiting time increased

# Diagnose Why Average Waiting Time Increased You are a Data Scientist supporting DoorDash logistics. Over the last 1 to 2 weeks, the business metric average waiting time has increased noticeably. Assume waiting time is defined as: `dasher_wait_time_minutes = pickup_time - arrival_at_store_time` You have event-level order data, including order created, merchant confirmed, dasher assigned, dasher arrived, pickup, and dropoff timestamps. You also have merchant attributes, dasher attributes, market-level supply and demand data, and experiment or product-change logs. ### Constraints & Assumptions - The goal is to diagnose the root cause before recommending fixes. - Include data quality validation before interpreting the metric movement. - Consider marketplace, merchant, dasher, customer, product, and operations explanations. - Assume any proposed fix must be evaluated with guardrails for cost, customer experience, merchant experience, and dasher earnings. ### Clarifying Questions to Ask - Is the increase global or concentrated in specific markets, merchants, hours, or order types? - Did the metric definition, instrumentation, or filtering change recently? - Are we looking at mean wait time only, or also percentiles such as p50, p90, and p95? - Were there recent changes to dispatch, batching, merchant prep-time prediction, promotions, or supply incentives? ### What a Strong Answer Covers - Reproduce the metric and validate timestamp quality, time zones, missing data, cancellations, and outliers. - Decompose the change by market, time, merchant, cuisine, order size, dasher segment, batching, and product surface. - Separate merchant prep delays from dasher arrival timing, assignment latency, routing, and demand/supply imbalance. - Identify change points and connect them to launches, promotions, weather, holidays, merchant onboarding, or operational events. - Propose levers such as better prep-time prediction, dispatch timing, merchant throttling, courier incentives, batching rules, and merchant operations tooling. - Evaluate fixes with experiments or quasi-experiments using primary metrics and guardrails. ### Follow-up Questions - How would you distinguish a true merchant-prep issue from an instrumentation issue? - If only a few high-volume merchants drive the increase, what would you do? - What guardrail metric would protect dasher earnings? - How would you design an experiment for a new dispatch-timing model?