Design Rideshare Marketplace Causal Analyses

Imagine you are a data scientist working for a ride-hailing marketplace. Respond to the case prompts below as though you were advising leaders in product, operations, and marketplace teams.

1. High-risk drop-off warning for drivers: The company plans to add an in-app alert informing drivers when a passenger’s drop-off location is in a neighborhood classified as high-risk. How would you design an experiment to assess this feature? Which business, safety, marketplace, and fairness outcomes would you track? If the test could be run across many cities, how would that alter your experimental setup?

2. ETA importance and improvement: Describe why estimated time of arrival (ETA) matters for a ride-hailing product. Which metrics would you use to evaluate ETA quality? What product, data, or machine-learning methods would you consider for diagnosing and improving ETA accuracy?

3. Advance driver promotion for specific times and places: The company wants to tell drivers a week ahead of time that they can receive a promotion by going online and accepting trips in particular locations during particular time windows. What effects might this incentive produce? Why could a simple user-level A/B test be unsuitable? What issues would a switchback experiment create? If you used a synthetic control design instead, how would you estimate uncertainty or variance?

Overview: This question tests a data scientist’s abilities in experimental design and causal inference, marketplace metrics and diagnostics, fairness and safety impact assessment, incentive and policy evaluation, and ETA modeling and forecast quality measurement.