Analyze the Accident-Rate Spike
Quick Overview
This question evaluates statistical analysis, KPI and data-pipeline validation, causal inference, and diagnostic reasoning for interpreting temporal trends in transportation safety metrics, testing the ability to distinguish true safety deterioration from denominator effects, reporting artifacts, seasonality, or product-mix changes.
A monthly line chart shows the accident rate for Uber trips in one city. The accident rate increases sharply from June through November, then drops quickly after November. You are asked to investigate what might explain this pattern.
Assume the current KPI is defined as reported accidents per 100,000 completed trips in local time, but you should question whether that is the right exposure metric. Describe how you would analyze the trend.
In particular, discuss:
- how you would validate the metric definition and data pipeline
- what additional data you would request
- plausible business, operational, seasonal, and measurement-related hypotheses
- how you would separate a true safety deterioration from denominator effects, reporting artifacts, or product-mix changes
- what statistical or causal methods you would use
- what actions you would recommend under different findings