This question is three questions stitched together and I didn't realize that until I was halfway through the diagnosis part.
Start by segmenting the problem to identify where cold food complaints concentrate (e.g., by restaurant, dasher, region, time, order type). Then define a metric tree that connects delivery time components to food temperature, and design an A/B test that isolates a specific intervention with clear success metrics and guardrails.
Pro tip: Emphasize that temperature is a proxy for delivery time and handling; use existing data to triangulate the root cause before jumping to solutions. Also, consider that a fix might involve trade-offs (e.g., faster delivery vs. cost) and design the experiment to measure both impact and cost.
Break down complaints by dimensions like restaurant, dasher, region, time of day, order size, and delivery distance to find patterns. Use funnel analysis to see where delays occur (e.g., food prep, dasher wait, transit).
Establish a metric tree: top-level metric (e.g., % orders reported cold), driver metrics (delivery time, time to accept, wait time at restaurant, transit time), and guardrail metrics (customer satisfaction, cost per delivery).
Based on segmentation, form hypotheses (e.g., long dasher wait times at restaurants, inefficient routing, lack of insulated bags). Prioritize by impact and feasibility.
Choose a specific intervention (e.g., incentivizing dashers to use insulated bags, optimizing dispatch). Randomize at the order or dasher level, define control and treatment, set sample size and duration, and specify primary and guardrail metrics.
Analyze results for statistical significance and practical impact. Check for heterogeneous treatment effects and guardrail metrics. If successful, consider rollout; if not, iterate on hypotheses.
AI-generated suggestions, not part of the candidate's original notes. May be inaccurate — verify before relying on them.