Design an Uber feature and analyze safety

You are interviewing for a summer Data Scientist internship at a ride-sharing marketplace.

Part A: Product case

Uber is looking for proposals for either a new rider-facing or driver-facing feature, or a substantial improvement to an existing workflow, that would materially improve the app. Choose one idea and address the following:

• the intended user and the pain point it solves;
• how the feature would be expected to shift user behavior;
• the north-star metric, the primary success metric, and the guardrail metrics;
• the tradeoffs among rider experience, driver experience, safety, reliability, and revenue;
• how you would test the feature, covering the randomization unit, experiment length, power or minimum detectable effect considerations, and how you would account for marketplace interference or spillovers.

Part B: Safety trend analysis

Suppose the monthly accident rate in a single city is defined as the number of accidents per 100,000 completed trips. A line chart indicates that the accident rate rises sharply from June through November and then falls quickly after November. Explain how you would investigate this pattern, being specific about:

• checking the metric definition along with its numerator and denominator;
• plausible explanations for the rise and later decline;
• which internal data slices and external data sources you would examine;
• how to tell a genuine safety change apart from a reporting or measurement artifact;
• which statistical or causal approaches you would apply and what follow-up actions you would suggest.

Overview:

The question tests a data scientist's product sense and analytical skills, including feature ideation, metric definition and validation, randomized experiment design that accounts for marketplace interference, trade-off analysis across rider, driver, safety, and revenue dimensions, and time-series and causal inference techniques for safety trends in a ride-sharing marketplace. It is common in analytics and experimentation interviews and assesses both conceptual understanding and practical application through metric literacy, hypothesis generation, experiment power and randomization decisions, and the ability to separate true safety changes from reporting or measurement artifacts.