Source
Daily MBTA LAMP subway performance files. Each row describes a trip-stop observation with actual and scheduled travel or headway fields.
OPEN METHOD · VERSION 1.0
Every score should
show its work.
TransitMeter is an interpretation layer, not a black box. Here is exactly how the Boston rail score is produced.
65%
+Headway balance
The share of comparable arrivals whose observed headway is between 0.5× and 1.5× the scheduled headway.
35%
=Travel adherence
The share of measured station-to-station segments completed no more than 120 seconds above scheduled travel time.
0–100
Meter score
Rounded to the nearest whole point.
Comparable observations
Headways must have observed and scheduled values, a scheduled interval from 2 to 30 minutes, and an observed interval under two hours. Travel segments need positive actual and scheduled times, with observed travel under one hour.
Line roll-up
Green Line branches are combined under their shared trunk. Each line score is calculated from all valid observations in the window. The system score is weighted by the number of valid headway and travel observations.
Trend
The seven-day delta compares the mean score of the latest seven service days with the first seven service days in the 30-day window.
LIMITS
What we do not claim
- The score is not an official MBTA metric.
- It does not measure crowding, accessibility, safety or the total door-to-door trip.
- It does not identify why service was delayed.
- Missing or revised source records can change results.
- Comparisons are strongest within the same mode and measurement window.
CITE THIS RELEASE
TransitMeter. “Boston MBTA Rail Reliability Index.” Data window 2026-08-17 to 2026-09-15. Generated from MBTA LAMP subway performance exports.
STATION-PAIR TRAVEL TIMES
Start at actual departure
For each origin, we use the movement timestamp for the segment leaving that station—or the observed stop timestamp plus dwell time when available. This avoids counting terminal layovers as rider travel time.
Keep direct completed trips
A station pair is included only when both stops occur in order on the same trip. Waiting and transfer time are excluded because the source records measure vehicle movement, not a passenger's full door-to-door journey.
Require repeated evidence
No journey page is published with fewer than 100 valid observations in the rolling window. Invalid, negative and longer-than-two-hour intervals are excluded before aggregation.
Publish the distribution
The typical time is the 50th percentile. The slower-trip allowance is the 90th percentile. “On pace” means the observed ride finished within two minutes of the schedule active for that trip.