Investigate Falling Successful Orders in LA

DoorDash observes that the number of successful orders per day in the Los Angeles market has declined materially over the last 2 weeks relative to the prior 4-week baseline. Assume a **successful order** is an order that is placed, accepted, picked up, and delivered without cancellation. You are asked to investigate the problem as a marketplace data scientist. Please answer the following: 1. How would you validate the metric drop and distinguish between a true business decline vs. a logging, definition, or seasonality issue? 2. Propose at least one plausible hypothesis from each side of the marketplace: - Customer - Dasher - Restaurant 3. If all three sides appear to be contributing to the decline, how would you **quantify the impact of each hypothesis** rather than just listing possibilities? 4. How would you structure a funnel analysis for this problem? What metrics would you inspect at each stage, and how would you segment the analysis to avoid misleading conclusions from mix shifts or confounding? 5. Propose one or two product or operational improvements. 6. Choose one improvement and design an A/B test. Specify the experimental unit, primary metric, secondary/guardrail metrics, interference risks in a two-sided marketplace, and how you would think about power or MDE. You should explicitly discuss tradeoffs such as demand vs. fulfillment, short-term metric recovery vs. margin impact, and risks like Simpson’s paradox, selection bias, and marketplace spillovers.