Bus Scheduling, Part II: Interlining

Introduction

This is the second part of a three-part series. To read the first part on runtimes and layovers, click here. 

In this article, I will focus on the practice of interlining. 

According to OC Transpo, whose memo you can read here, interlining is 

essential to delivering the most efficient scheduling solution ... There are two main goals of interlining: to permit more flexibility from a vehicle type and capacity perspective. For example, high-capacity buses can be moved throughout the network to deliver service on specific trips that require those buses; and to increase efficiency by reducing the overall number of buses and bus operators required to provide the service, which can help to reduce operating costs.

 

 As we shall see, neither of these things are true. 

Currently, 27 per cent of all in-service trips are followed by a trip on a different route, and 40 per cent of all existing blocks include four or more different routes. Most (72 per cent) deadheads are short and have a duration of five minutes or less. A quarter of all deadheads are between five and 30 minutes in duration, and about three per cent of all deadheads are 30 minutes or more in duration, to or from the most remote parts of the system. 

 

Interlining

Again, I believe that I should explain the practice of interlining. 

To quote OC Transpo's explanation, interlining is the practice of running a bus and/or operators on multiple routes between the time it leaves the garage and returns. 

I believe the situation will become clear once I show a few blocks. 

OC Transpo claims that interlining can "improve efficiency" and enable "flexible" operations. Both of these things are true, but OC Transpo does not follow industry best practices when it comes to scheduling, and so the actual system of interlining is one that wastes resources, increases bus wear-and-tear, decreases reliability, and reduces the flexibility of active route management to make real-time adjustments. 

OC Transpo uses an automated system, HASTUS, to generate bus blocks after timetables have been finalized, which often produces interlining. While this, in theory, reduces costs, in reality it makes calculating "cheap" service changes (the third part of this series) impossible, and reduces the flexibility to adjust runtimes, layovers, and other service adjustments. 

It is important to note that automated crew scheduling is common, but vehicle assignments are not usually automated. 

On most transit systems, buses are assigned to one route and it runs that route all day, or at least for a substantial period of time. In these cases, interlining can genuinely reduce costs on routes with awkward runtimes, or provide cheap service improvements, but OC Transpo's overuse of interlining is extremely rare. 

One note - I have picked out some of the more egregious examples as a demonstration of bad practices. Unlike the last piece though, interlining is statistically common on OC Transpo, and this is backed up by their own claims: 27% of deadheads are over five minutes, which in practice, means that OC Transpo needs a layover for the deadheads themselves, spiraling bus use out of control. 40% of blocks include four routes or more, meaning that there are more opportunities to propagate delays. 

Some Examples

As with the first piece, I will provide some examples of interlining to show how it harms the system. All examples are taken from 8 May, but apply to every weekday until the end of this service period. 

The first example is 4738, an LFS (small bus). 

Between 0830 and 1000, we can see that the bus deadheads back to Laurier, on a route with equal service levels in both directions. This is a waste of ninety minutes of operating time - at the average cost of $130/hour, this has an annual wastage of around $50 thousand - as well as the wear and tear of taking a bus fifteen kilometres everyday. 

In a normal transit system, the bus would simply run back and forth between Laurier and Bayshore, providing more service on its way and freeing some bus to go run elsewhere. 

In most transit systems, there are more buses running in the peak direction than otherwise. This means that most buses should be deadheaded in the counterpeak direction. This bus is deadheading in the peak direction. 

Our next example is 6684, an articulated bus. 

This block illustrates how delays can appear and spread to different routes with adequate runtimes. In this case, there are two examples of wastage: the half hour the bus took deadheading from Hurdman to Barrhaven Centre, and the twenty minutes (more on the schedule) from Terry Fox to Baseline. Additionally, the delay that appears on the inadequately scheduled 61 reappears as a ten minute delay on both the 56 and 53, contributing to bunching on both and reducing the ability of the system to run a schedule even further. 

The practice of pulling buses from frequent routes to cover infrequent ones also appears, with a (very late! Late afternoon 12 runs also appear in the first article) 12 run pulled halfway to cover a cancelled bus on the 13, which inevitably leads to further delays on the next WB trip. 

The third example is 6543, another articulated bus. 

The bus begins with a 94 run in Findlay Creek, moves to Barrhaven for a single express run, then deadheads back to the east end to run a school route from Hurdman. In three hours of peak period running, this bus only actually ran a route for two hours, costing us thousands; and this isn't even mentioning wear and tear on the (old) bus. This pattern is repeated in the afternoon, with a two-hour deadhead to run a school route whose travel time is three minutes. We also see another PM deadhead with a duration of one hour. Someone has to drive the bus during that time. 

Additionally, the pattern of late routes delaying other, well-written routes, appears in the PM, with the bus departing Baseline late towards Tunney's Pasture, resulting in the 75 trip also departing Tunney's Pasture late. This problem is "solved" by wasting an hour of operator time deadheading from Barrhaven to South Keys - but in a well-scheduled system, this wouldn't be necessary. 

The next example is 6573, yet another articulated bus. 

Again, we see how interlining wastes resources. This bus spends nearly two hours to run a single school route (admittedly, this does not calculate the effect of the thirty minute mandated break), which could easily be run by some other counterpeak bus coming back from Orleans for cheaper. 

Interlining has the effect of obscuring cheap service improvements, or even service improvements which would not cost money; adding an extra trip to the 30's frequent peak period service, in effect lengthening its period by 30 minutes, would not require another bus make the trek out to Orleans, likely cancelling out the extra cost. The forced interlining and automated block generation hides cheap and easy improvements like this one, an idea which I will explore more in the next and final article in this series. 

The final example is 4727, an LFS - both of its blocks. 

This one is a little ridiculous. This bus has three runs, which is enough to take it from Findlay Creek to Kanata. Additionally, the 163 run is delayed, because the driver seems to have gone down the pre-New Ways to Bus route, which is why it departs twenty minutes late. This one is incredible because the bus spends two hours on the road, not including the time it takes to get to the garage, and half of that is deadheading. 

But wait, there's more! 

The afternoon 4727 block is a failure of both management and of interlining. The bus spends about ninety minutes between termini. However, this hasn't seemed to affect the on time performance, even if it does cost lots of money. 

However, one can see that on the 80 run, the driver left its origin early, and then got to its destination early. Then they sat there for 4x the actual layover time, spiraling the Citigate-bound trip into a run that is seven minutes late. This is a case where a manager needed to tell the driver to get going, and to sit at timepoints unless otherwise informed. This pattern plays out again on the 12, where the driver sits at the layover for an excessive amount of time, causing delays, and highlights how many facets of active management - cancellations, short-turns, rerouting, and layovers - there are, and how skilled these people need to be. 

Efficiencies and Savings

In its quest to increase efficiencies, OC Transpo has done the opposite, turning the scheduling system into one that is not flexible, not reliable, and not efficient. I am told that this has occurred under current management, and is a stellar example of why we need radical new thinking - not just the old guard - to improve operations and service for our transit. 

The current system of interlining increases costs, increases wear-and-tear, and decreases the ability of managers to actively manage buses. It propagates delays, as seen both here and in the first article of the series, by spreading delays from one route to the next, and doesn't provide the benefits that OC Transpo claims (or thinks?) it does. 

There will be more to come on the downsides of interlining in the third article of this series. For now, know that while its ability to cause delays seem mostly limited to its ability to spread delays like a virus, it essentially costs more money to run the same service by handing scheduling blocks over to a computer, rather than using human supervision in a manner inline with industry practices, while preventing easy service gains. 

OC Transpo's practices are far out of the norm for transit agencies, and for good reason. Don't let them fool you into thinking that this is an improvement. 

Until next time.