DCX started as a measurement engine for DC Metrobus. The hard part — reconstructing ground truth and benchmarking predictions — is city-agnostic. Here's what's live, what's being built, and what's next. No dates; honest status.
Live now
The measurement engine
Continuous ingestion & map-snapping.Every WMATA bus matched to its route in real time, with detours and ghost buses surfaced, not hidden.
Ground truth, derived.Real arrivals reconstructed from vehicle positions — the basis for every score.
Per-route reliability scorecards.Expected wait, headway regularity, on-time, schedule deviation, detour rate — for every route and direction, daily.
The benchmark.Our predictions logged side-by-side with the agency's own feed from day one, both scored against ground truth. Unreconstructable after the fact — so we started logging immediately.
Public board, live map & methodology.The reliability board, a live fleet map, and a fully transparent methodology.
Calibrated letter grades.Grades anchored to the realistic range of bus performance, so they discriminate across the system instead of failing everything — an honest absolute scale, not a curve.
Building
From measured to proven
A learned ETA model that beats the agency's feed.The launch gate: we don't claim "better predictions" until the benchmark proves it at the horizons riders care about.
Public launch.Open the scorecard site to everyone — the reliability data is meant to be public.
Next
The product
Reliability data API.Keyed, programmatic access to the enriched warehouse — for operators, researchers, and newsrooms.
Ask the data in plain English."Which routes degrade most during Ward 7 rush hour?" — answered conversationally, no SQL required.
Automated reliability narratives.Plain-English summaries and weekly system writeups, generated from the data, ready for press.
More cities.The pipeline is agency-agnostic. Point it at any GTFS feed and the reliability layer follows.