Working Out the Kinks in the MBTA Dashboard

Since the initial launch of the MBTA Performance Dashboard, we’ve noticed a few small discrepancies in some of our data, including bus ridership, Commuter Rail reliability, and Green Line reliability. We’re correcting these issues, so if you’ve been following the dashboard closely you may notice some changes to a few numbers.

First, there were some small problems in the way bus ridership numbers were being reported to the dashboard. These were primarily due to how holidays were counted when calculating the average weekday ridership. We’ve corrected these problems, which means the ridership numbers are slightly higher than what we had shown in prior weeks (about 5% in some months).

Second, on Friday, April 15th we made a small change to our performance reporting system in the way Green Line was handled at the stops from Lechmere through Boylston Stations westbound. This change was intended to better capture performance at these stations, but in fact resulted in almost all passengers there being counted as having good wait time, biasing the reliability numbers upward. This was particularly pronounced on the C and E branches since they serve more of these stations than the B and D branches do. We’ve reverted back to the version of the software that was running before April 15th and have recalculated performance for the 15th-19th, where the data were incorrect. We are working on a better fix that will solve the problem we had originally targeted without causing other issues.

Finally, Commuter Rail data also contained some discrepancies between the dashboard and the official numbers Keolis is judged on under their contract (Keolis is a private company that operates the Commuter Rail on behalf of the MBTA). First, some trains that don’t actually take passengers (non-revenue trains) were being included in the dashboard’s performance metric. Second, the Commuter Rail GPS data is less robust than bus or Green Line GPS data, so some corrections may be made to it in the day or two after a train actually runs. This means that the reliability for a line as reported the next day may change over the coming days if the data is corrected. Please keep this in mind when viewing Commuter Rail reliability for recent days.