You're designing an experiment for a two-sided marketplace where rider-side changes can shift driver supply, acceptance, and wait times. The team is debating how to measure rider impact when treatment may spill over to the driver side.
How would you design an experiment for a marketplace product where treatment on the rider side may affect driver behavior?
Recognizing interference between riders and driversChoosing the right unit of randomizationUsing switchbacks for marketplace experimentsDefining primary metrics and guardrails