How to evaluate lowering ETA?

Uber wants to estimate the business value of reducing ETA, where ETA is the predicted time from when a rider requests a trip until the driver arrives at the pickup location.

Because Uber is a two-sided marketplace, changing ETA can affect rider behavior, driver behavior, and overall marketplace equilibrium. Answer the following:

1. Which metrics would you track to evaluate the impact of lowering ETA? Propose a clear primary metric and relevant guardrail metrics. Consider rider funnel metrics, cancellations, completed trips, gross bookings, marketplace health, and longer-term retention.
2. From a business perspective, why is reducing ETA important?
3. Suppose historical observational data shows a positive correlation between ETA and session conversion rate, where session conversion rate is defined as the fraction of rider app sessions that lead to at least one successful ride request. How would you interpret this relationship? What confounding variables, selection effects, or aggregation artifacts could explain it?
4. How would you design a causal experiment to measure the impact of lowering ETA? Discuss when a user-level A/B test is appropriate, when a switchback experiment over geography and time is preferable, and when synthetic control would be useful.
5. After the experiment, suppose the estimated lift in session conversion rate has a 95% confidence interval of [-5%, +1%]. How would you interpret this result for decision-making?
6. How would you improve the precision and power of the experiment?
7. Assume the study used a user-level A/B test, and after seeing the results the PM asks for a deep dive on users who took at least 5 trips during the experiment. How would you respond, and how would you communicate the statistical risk in this request?