A Data-Driven Approach to Prioritizing Bus Schedule Revisions at New York City Transit

Abstract

Over two million trips are taken every weekday across the New York City Transit (NYCT) bus network. Revising the schedules for each of these routes is a labor-intensive process, and because of limited resources, fewer than half of all routes are examined each year. Traditionally, schedules have been revised on a first-in, first-out basis, with most schedules rewritten once every 2 years. This approach leaves no room for reviewing routes that need more frequent changes, meaning service may not catch up to changes in demand or traffic patterns for several years. It also requires staff to spend valuable staff time analyzing routes that may be inactive. To better address rapidly evolving bus corridors, NYCT developed a methodology to pinpoint the routes most in need of schedule revisions. This data-driven approach uses automatic vehicle location data and a ridership algorithm that combines automated fare collection data with other sources to infer stop-by-stop boardings and alightings for individual trips. Each route’s schedule is evaluated on service capacity versus actual ridership, and scheduled versus actual running times. Routes that show the greatest discrepancies are designated for later in-depth review. This methodology was applied to develop the 2018 list of schedule revisions. As this process identifies routes with too much capacity or running time, as well as those with too little, resource-costly schedule adjustments can be offset with resource-saving ones. Using this approach allows scheduling staff to react more quickly to changes in customer demand and new development, thereby providing better service to passengers.