Evaluate ETA Impact on Conversion

You are employed as a Senior Data Scientist at a ride-hailing platform, such as Uber. ETA is the estimated pickup time displayed to a rider before they choose whether to request a ride. A product manager wants to lower ETA and has asked you to assess the business impact. Respond to the following related prompts:

1. Why does reducing ETA matter? Describe the anticipated benefits for riders, drivers, and the marketplace as a whole from shorter ETAs.
2. In historical observational data, you observe that higher ETA is positively correlated with rider conversion. Why could this occur even if longer waits are not actually causing higher conversion? Give several plausible explanations.
3. How would you set up an experiment to estimate the causal effect of lowering ETA on conversion?
   • State the treatment and control conditions.
   • Select the primary success metric and at least 2-3 guardrail metrics.
   • Specify the randomization unit and explain when exposure should be measured.
   • Explain how you would calculate and interpret a confidence interval for the treatment effect.
   • Mention any power or minimum detectable effect considerations.
4. Why might user-level randomization be a poor choice in this marketplace setting? What alternatives would you consider, and what tradeoffs do they introduce? Your answer should explicitly discuss issues such as confounding, selection bias, marketplace interference, and the difference between correlation and causation. Overview: This question assesses a data scientist's abilities in causal inference, experimental design, and marketplace analytics, including understanding of confounding, selection bias, interference, and the distinction between correlation and causation.