Showcase · Data Science

Transit Delay Explorer

A two-year window of public transit telemetry, reduced to an interactive report that explains where delay accumulates and why.

Discipline
Data Science
Role
Data engineering & analysis
Duration
5 weeks
Year
2026

Problem

Delay statistics were published as monthly averages per line, which hid the fact that most lost time accumulated at four interchanges during a ninety-minute window.

Approach

  1. 01Ingested 41M stop events into DuckDB with a reproducible, idempotent loader.
  2. 02Reconstructed trip-level delay propagation rather than treating each stop independently.
  3. 03Validated against manually timed control journeys before publishing any figure.
  4. 04Published as a static interactive report with the query layer embedded in the page.