A new matching model is ready for launch. Design an A/B test to determine success. Specify: (1) primary metric and guardrails for both sides of the marketplace (e.g., booking conversion, provider response rate, time-to-first-response, cancellation rate); (2) unit of randomization and how you will prevent interference/spillovers (e.g., geo or time bucketing, provider saturation caps); (3) power analysis with baseline rates, MDE, variance estimates, and test duration; (4) plans for CUPED or covariate adjustment, sample-ratio-mismatch checks, and sequential monitoring boundaries; (5) a pre-registration doc with stop/go criteria and a rollback plan; and (6) how to interpret heterogeneous lift by region and job_category without p-hacking.