Bus Scheduling, Part I: Run Times, Layovers, and Propagating Delays

Introduction

Reliability is a major concern for both OC Transpo and its passengers. According to OC Transpo's own surveys, reliability of service is the primary concern of its riders, while Transit Committee is dominated with discussion of reliability: fleet, scheduling, customer satisfaction, plans, and more. It's clear that this is a major issue for Ottawa's transit sphere. 

This is the first of three articles examining OC Transpo's scheduling practices. In this piece, I wish to examine how poor scheduling can propagate delays across runs and routes, using a number of case studies. 

Disclaimer: I am not a bus scheduler, or any other kind of expert in this field. This is purely a casual glance at OC Transpo's practices, and there are always hidden "backroom" factors that are difficult to account for. 

In posting this analysis, I simply hope to shed some light into OC Transpo's scheduling practices, in order to demonstrate best practices in other cities and how OC Transpo is, or is not, utilizing them. 

OC Transpo has produced a report of its own on scheduling practices, which you can read here. 

Run Time

This is a rather esoteric subject, and I feel that an explanation of the terms in the headline is warranted. 

The runtime of a route is how long the a route takes to run from point A to point B. The scheduled runtime depends on many factors - congestion, the busyness of the route, its length, and corridor. The scheduled runtime is compared to the real world runtime, which is the actual amount of time the bus takes on its routes. This varies across different runs and days, so the goal of the schedule is to write a runtime that reflects the real runtime as best as possible - a difficult task. 

I have complained about inadequate runtimes in the past - on many routes, trips are consistently arriving late at their terminus, despite leaving their origin on time. However, taken to its logical extreme, the opposite can be just as damaging. 

When a route has excessive runtime, the padding can cause buses to arrive early, frustrating customers and causing wastage. For example, the new Route 40 from Greenboro to St. Laurent has excessive padding throughout the day, as seen below. Especially northbound, there is more scheduled time than needed, which reduces on-time performance and causes more buses to be used than necessary. 

A sheer waste of time and money. 

Layovers

Layovers, or recovery time, is the amount of time a bus has scheduled between runs. A higher recovery time reduces the amount that delays bleed from one run into another, but recovery time costs money and reduces the amount of service that can be made available to customers on the network. 

Ideally, a well-designed schedule would provide enough time to recover from the vast majority of delays without introducing excessive waiting at terminals for operators. 

Of note here, through the collective bargaining agreement with the ATU Local 279, OC Transpo is required to provide 5% of runtime in layovers, which began after the 2009 strike. 

There are a few ways to measure layovers: in minutes (usually with the trip time included), as a percentage, which is the method used by OC Transpo, and in minutes-per-hour, which is the method I will use in this piece. The amount of layover varies with schedule fragility; the Barrhaven local route running at 2300 on a Sunday doesn't need much more than five or six minutes of layover an hour, whereas the 6 at 1600 on a Thursday might require ten, eleven, twelve or more minutes an hour. It's context dependent. 

Propagating Delays

Delays don't just disappear at the end of a run. If a route has no layover time going into the next run, the delay accumulated at the end of one run becomes the delay at the origin of the next trip. If the layover time is inadequate, that time can "eat" some of the delay, but often, it will not eat all of the delay. 

Additionally, a route that falls behind risks bunching, as an above-normal amount of people are waiting at the stops, and a full bus accelerates slower; this pushes the front bus further behind, while the next bus carries fewer passengers and pulls ahead, often resulting in bunching. 

Good schedule design (and active management, more on that in the interlining article), can help avoid these issues. 

Delays beget delays, and when your schedule has a small margin of error, this holds even more true. 

A Few Case Studies

In order to illustrate my point, I'd like to highlight a couple of blocks, or the schedules for the buses. I want to state that while I am presenting only the delayed buses as a demonstration, most routes do not act this way. Because of the optics of bus bunching, and because of the principle that delays beget delays, this does, however, reflect the experience of many riders across OC Transpo. 

All of these examples come from 8 May, which was a Thursday. The bus blocks and schedule do not change across a schedule period, so every weekday from 28 April until the end of June, has these schedules, even if they play out with different fleet assignments (and actual runtimes) everyday, their general shape remains the same. 

Our first case study is 4849, a regular LFS (the newer small buses). One may note that the interlining caused this bus to be used on peak direction runs in the peak period, on some of the busiest routes in the network. I will have more on this in the interlining article. 

The first two runs already demonstrate both how delays from one run can bleed into a different route altogether, making active management difficult. The 80 and 53 run both show inadequate runtime, and the 80 run has a scheduled layover of just six minutes on an hour long run, which is inadequate for a route which deals with such variable traffic on Merivale. 

The next layover and deadhead is fifty minutes, which puts the route back on schedule. But if you look at the runtime deficiency, these continue to accumulate across the entire midday with some Baseline-bound runs being over fifteen minutes deficient, so that by the afternoon peak, the bus is forty minutes behind schedule - the 82 runs every thirty minutes, mind you - and one trip has to be cancelled (a good example of active management that seems destructive but is actually necessary to keep the service on schedule). This puts the bus back on schedule, and the long break at Bayshore means the delays on the 11 are not too destructive to schedule adherence. 

This block is a great learning example. First, schedules need to reflect on real-world conditions. This means that buses should not be consistently fifteen minutes slower than scheduled, because that will surely put them behind their schedule. Second, layover time isn't a magic bullet. Eight minutes an hour is more than enough for a low-traffic route like the 82, but in practice, the totally inadequate runtimes are impossible to make up for with layover. That is the first priority for fixing. Finally, the active management at the end of the midday, in which a Lincoln Fields-bound trip is cancelled, is important to keep the later runs on schedule - at the cost of spreading delays to the next run. However, OC Transpo is not proactive in communicating cancellations, which results in justified frustration among transit riders. 

Our next example is 8133, a double decker bus assigned to the 6 and 98. 

Again, this demonstrates how good scheduling is being violated on the urban network. The 6's scheduled runtime, which is, I'm told, influenced by the bus type (double deckers are slower than the articulated buses), is half an hour shorter than it should be. Why we're assigning double-deckers to the 6 at all is a question both for OC Transpo management, who insist on an automated system that nobody else uses in this way, and City Council, who have failed to procure the buses we need. 

What's more, the layover time is five minutes an hour for an urban route that is notorious for its unreliability, running on a congested Bank Street (and other congested streets). Total fail. 

This one is 4694, another LFS. 

Here we have another example of runs which create delays. Notably, most of the delay is accrued on the second run; and the third/fourth runs would be an example of "delays beget delays" if every other bus on the 14 was not also similarly delayed. The 14 is a mess in general, but this illustrates that it only takes one bad run to mess with an entire block of schedules, and even though the other runs "only" accumulate a delay of nine minutes, with an inadequate layover and an original sin of the twenty minute delay, the block falls behind schedule and is never able to recover. 

This run also has only about six minutes of recovery per hour, even though the 14 is extremely unreliable. This is a case where adding one bus (and fifteen minutes of layover) to the schedule would result in reliability improvements for riders. 

The next example is 6611, an articulated bus. This block is interesting, because it illustrates both good and bad sides of OC Transpo scheduling. 

During the AM, the interlining is rather crazy, but the routes run on time. This continues through the midday - the layover time is about eight minutes an hour, not bad and adequate if the runtime is also acceptable, which is mostly is. However, on the final midday 12 run, the runtime goes behind by twenty minutes, and even with a ten minute layover at Tunney's Pasture, this knocks the 57 trip behind by twenty minutes also, forcing the cancellation of an inbound trip. 

The block returns to on-time performance in the evening, though the layover at Bayshore is excessive - remember that layover is context dependent and late evening trips do not need excessive layover - and the block ends just twenty-two seconds behind schedule at Hurdman. 

We can reinforce lessons with this block. First, the proper layover time is context-dependent and a better written route schedule (or a route with less schedule uncertainty - this involves concrete, or transit priority measures) needs less layover to stay on schedule. Next, one badly written trip schedule can throw everything behind it behind schedule, which is why "most trips stay on schedule" is not an excuse for any poorly written schedule. Active management can put trips back on schedule, at the cost of a scheduled trip. Focusing on KPIs ("number bad!"), rather than a holistic view of service management as a whole, can lead to scenarios like Toronto's where a fear of introducing cancellations, or short-turns (depending on the city) to pad numbers can lead to worse service for riders actually on the route. Finally, the delay on the 57 shows that even if one route's schedule is well-written, interlining can wreck its on-time performance anyways. 

Our final example is 8105, another double decker. 

This demonstrates how "delays beget delays." The first 41 trip, in which the bus fell twenty minutes behind schedule, followed a cancelled trip, for unclear reasons. While active management can be a blessing, it can also help cause delays like this one (though the next bus on the 41 fell ten minutes behind also ...). Thus, a poorly written schedule on one block can bleed delays onto different bus blocks - delays beget delays. 

The PM period had nearly fifteen minutes of layover an hour, which meant that active management was not necessary (or utilized) on this trip. That layover time is excessive for any well-written schedule though - and it might have been an artefact of the frequencies of this area, as well as interlining. In a normal system, that would be a waste of resources. But because the schedules were so inadequate, it was neccessary just to stay on schedule - you can see the fifteen minute layovers become one minute, which is basically the time it takes to turn around a bus. 

Lessons Not Learned at OC Transpo

Reliability is a hot topic at OC Transpo. And every time it came up at Transit Committee/Commission, Renee Amilcar, the general manager, said that New Ways to Bus would improve reliability. However, she kept adding one caveat: the traffic times were based on 2023 traffic levels. 

Yeah, we can tell, Mme. Amilcar. 

According to the memo distributed by Pat Scrimgeour to council, OC Transpo's automated scheduling system (we were using AI before it was "cool," with predictable results) needs to be activated months ahead of time to adjust travel times. 2023 is two years ago, which is not months, as far as I am aware. Time seems to work differently at OC Transpo. 

After September 2024, there should have been adequate notice to adjust runtimes. That they weren't done so is an indictment of the processes at OC Transpo, and of management who have failed to properly manage their transit system. 

What can we do? 

Flood OC Transpo and your councilor's office with complaints. If your bus seems to come on time, but arrive late, that means it's not getting enough runtime. Complain loudly and often, and perhaps the people in charge will notice that something is wrong. 

If you feel devious, you can tell your councilor that you plan to vote for the candidate who promises reform at OC Transpo. 

Whatever you do, please don't yell at your bus driver. They don't have the power to change scheduling, and even if they sit in the layover area, they don't deserve to be shouted at for using the bathroom after two hours on the bus (as seen in some of the charts above). 

Also, please don't harass OC Transpo's schedulers and planners. They work under a large set of constraints, inflicted by both council and management, which are impossible to work out. Employment by public opinion is not a good outcome. 

When it comes to scheduling, there's more where that came from. Next up: a look at interlining and how this wastes resources while delivering a worse service. 

Until next time.