Evaluate and test a Top Dasher program

Context: You work on a food delivery marketplace. Product proposes a Top Dasher program that grants the top 10% of drivers priority dispatch and flexible scheduling if they maintain acceptance rate ≥70% and completion rate ≥95% over the last 30 days. Questions: 1) Should we launch? Build a decision framework that quantifies expected value and risks. Identify primary success metrics (e.g., order ETAs, fulfillment rate, cancellations, driver earnings dispersion, merchant and consumer NPS) and guardrails. Define the explicit decision rule for go/no-go. 2) Design an experiment to estimate causal impact given marketplace interference (priority reallocates orders among drivers). Specify unit of randomization (driver, zone, market-day), clustering, sample size and duration, and how you will mitigate SUTVA violations and spillovers. 3) How will you prevent gaming (e.g., acceptance-rate manipulation) and selection bias? Propose instrumentation and pre-registered analysis, including CUPED or pre-period covariate adjustment. 4) Outline heterogeneity analyses (new vs veteran drivers, peak vs off-peak, high vs low supply markets) and how their results would change the rollout plan. 5) What are the ethical and operational tradeoffs (fairness across drivers, small-market impacts, earnings volatility)? Propose guardrail thresholds and a rollback plan.