Measure driver experience quantitatively
Overview
This question evaluates a data scientist's competencies in designing composite metrics, event-level aggregation, statistical validation, debiasing for sample mix, and causal experiment design to measure policy impacts.
Driver Experience Index (DEI) — Design, Validation, Debiasing, and Experimentation
Context: You are a Data Scientist at a ride-hailing company. Define a quantitative Driver Experience Index (DEI) that combines four pillars of driver experience: earnings stability, utilization, fairness, and product friction. Be precise about event-level definitions, aggregation, validation, debiasing, and experimentation.
Requirements:
(a) Specify exact event-level metrics for each pillar and the weighting/aggregation scheme into a single DEI score.
(b) Describe how you would validate the reliability and sensitivity of the DEI (e.g., test–retest reliability, correlation with guardrail outcomes, responsiveness to known changes).
(c) Explain how to de-bias for driver mix (tenure, geography, vehicle type, shift) via stratification or reweighting.
(d) Propose an experiment to estimate the causal impact of a new dispatch policy on DEI while avoiding seasonality and Simpson’s paradox. Be precise about the randomization unit, expected effect size, and power.
Solution
Login required