Headway Quality Measurement: A Proposal

For regular readers of this site, it will be no surprise that my opinion of the TTC’s reporting on service quality is that it is deeply flawed and bears little relationship to rider experiences. It is impossible to measure service quality, let alone to track management’s delivery of good service, with only rudimentary metrics.


  • The TTC reports “on time performance” measured only at terminals. This is calculated as departing no more than one minute early and up to five minutes late.
  • Data are averaged on an all-day, all-month basis by mode. We know, for example, that in February 2020, about 85 per cent of all bus trips left their terminals within that six minute target. That is all trips on all routes at all times of the day.
  • No information is published on mid-route points where most riders actually board the service.

Management’s attitude is that if service is on time at terminals, the rest of the line will look after itself. This is utter nonsense, but it provides a simplistic metric that is easy to understand, if meaningless.

Source: March 2021 CEO’s Report

There are basic problems with this approach including:

  • The six minute window is wide enough that all vehicles on many routes can run as pairs with wide gaps and still be “on time” because the allowed variation is comparable to or greater than the scheduled frequency.
  • Vehicles operate at different speeds due to operator skill, moment-to-moment demand and traffic conditions. Inevitably, some vehicles which drop behind or pull ahead making stats based on terminal departures meaningless.
  • Some drivers wish to reach the end of their trips early to ensure a long break, and will drive as fast as possible to achieve this.
  • Over recent years, schedules have been padded with extra time to ensure that short turns are rarely required. This creates a problem that if a vehicle were to stay strictly on its scheduled time it would have to dawdle along a route to burn up the excess. Alternately, vehicles accumulate at terminals because they arrive early.

Management might “look good” because the service is performing to “standard” overall, but the statistics mask wide variations in service quality. It is little wonder that rider complaints to not align with management claims.

In the pandemic era, concerns about crowding compound the long-standing issue of having service arrive reliably rather than in packs separated by wide gaps. The TTC rather arrogantly suggests that riders just wait for the next bus, a tactic that will make their wait even longer, rather than addressing problems with uneven service.

What alternative might be used to measure service quality? Tactics on other transit systems vary, and it is not unusual to find “on time performance” including an accepted deviation elsewhere. However, this is accompanied by a sense that “on time” matters at more than the terminal, and that data should be split up to reveal effects by route, by location and time of day.

Some systems, particularly those with frequent service, recognize that riders do not care about the timetable. After all, “frequent service” should mean that the timetable does not matter, only that the next bus or streetcar will be reliably along in a few minutes.

Given that much of the TTC system, certainly its major routes, operate as “frequent service” and most are part of the “10 minute network”, the scheme proposed here is based on headways (the intervals between vehicles), not on scheduled times.

In this article, I propose a scheme for reporting on headway reliability, and try it out on the 29 Dufferin, 35 Jane and 501 Queen routes to see how the results behave. The two bus routes use data from March 2021, while the Queen car uses data from December 2020 before the upheaval of the construction at King-Queen-Roncesvalles began.

This is presented as a “first cut” for comment by interested readers, and is open to suggestions for improvement. As time goes on, it would be useful for the TTC itself to adopt a more fine-grained method of reporting, but even without that, I hope to create a framework for consistent reporting on service quality in my analyses that is meaningful to riders.

Continue reading