For many years, the TTC has had a fetish for reducing short turns, or more accurately, for reducing short turn counts. Together with “on time” stats, this is a primary metric for TTC management, the one they get gold stars for.

When former CEO Andy Byford hired the current CEO Rick Leary, one of his first tasks was to reduce short turns. The result is that the CEO’s report concentrates on two factors to the exclusion of all other measures of service quality.

Here are two pages from the June CEO’s Report for streetcars. There are similar pages for the bus network.

In a recent newsletter to constituents, Councillor Brad Bradford, also a member of the TTC Board, included the following chart to show how the TTC is attacking the short turn problem. Short turns on 501 Queen were the lowest in number among all streetcar routes in early April, and fell to zero on Monday April 15 (along with a big drop on other routes too). This looks good, but as we will see later, has unintended side effects.

Bradford’s newsletter included this text:

We all know the frustration of too many short-turning streetcars, especially during rush hour.  As a member of the TTC Board I’ve been working hard with my colleagues to improve the streetcar service and reduce the number of short-turns.  I’m happy to report that April 2019 had the lowest number of short-turns since 2014, including the 501 streetcar which had ZERO short turns for several days.

Short turns disproportionately affect Bradford’s constituents as they live at the eastern end of the 501 Queen and 506 Carlton routes. However, irregular service which I have documented in numerous articles affects riders along the entire route. Gaps of 15 minutes or more in what is advertised as “frequent service” do not encourage ridership, and the unpredictability of service leaves many people walking or taking alternate modes to the TTC.

Measures of Service Quality

The count of short turns only tells us how many cars did not reach a terminal over the course of the day. It does not tell us:

• What proportion of the service this count represents. The scheduled service to Neville over the course of a day (6 am to midnight) is about 200 trips with a similar count at Humber. 40 short turns represent about 10% of service assuming they are equally divided between the two terminals.
• What time(s) of day were most affected. Certain times of the day are disproportionately affected by short turns, notably the hours immediately after the am and pm peaks, and through the evening especially on busy nights in the club district downtown.
• Whether the short turn was successful in restoring more regular service, otherwise known as “filling the gap” on its next trip. There appears to be little or no management of cars re-entering service from short turns, and they may well reappear immediately ahead of or, even worse, behind a “through” car without reducing the headway. The average headway looks better, but it’s a bunch of two cars with a wide gap, not evenly-spaced service.

Unless one sees a breakdown such as the one published by Councillor Bradford, the numbers in the CEO report are accumulated for all routes over all periods of operation including weekends. This is a very generic average value, and it gives only the most general idea of short turning as a trend, rather than pinpointing problem routes and times of the day.

The other published metric is the “on time departure”. This is something of a misnomer because “on time” is defined as a window from one minute early to five minutes late, in other words six minutes’ grace relative to the schedule. For a route with a five minute scheduled service, three cars could depart close together and be “on time” for the stats.

The TTC does not report on headway reliability and bunching, issues which are at least as important as short turns. A rider on a Queen car bound for, say, Dufferin Street from Yonge does not care if their car gets short turned at Sunnyside, but they do care if no car shows up for 15 minutes or more, and they cannot board the first one in the parade.

If the scheduled times were 12:00, 12:05 and 12:10, the actual departures could be 12:05, 12:07 and 12:09 and fit within their allowed windows. The service is supposed to look like this:

X————X————X————X————X————X————X————X————X————X

But it can look like this and still be “on time”:

—————X—X—X——————————X—X—X——————————X—X—X——————

It would not take long for cars to bunch together as triplets running across the route. However, the on time measurement only applies at the terminals, and what happens after vehicles leave is not reported.

The TTC produced route-by-route statictics for five months, April to August 2018, but they have been missing-in-action ever since.

An important metric is the distribution of headways by time of day. Charts showing this information are published regularly on this site. (See this article about headways at Neville Loop and scroll down to charts showing the range of headways actually operated.)

The TTC would do well to report on the proportion of headways that are less than 50% of the scheduled value, or more than 50% over that value, broken out by by time of day and by location. Situations where vehicles are running very close together (except for the few routes and periods where the scheduled service really is that frequent) should be flagged. This is not a particularly challenging exercise, but potentially quite embarrassing.

As a quick check, I looked at 501 Queen westbound at Yonge on Wednesday May 1. Of the 215 trips crossing Yonge Street, 54 were on a headway of two minutes or less, or one quarter of the service. There were 25 trips on headways ranging from 10 up to 18 minutes, over 10% of the service. Those are all-day values, and the proportions vary by time period. (The eastbound stats are comparable.) Here is a chart of the day’s data. Note that the trend line sits at about 5 minutes for much of the day only slightly above the scheduled headway, but what riders experience is wide gaps followed by at least two cars.

(For those who might argue that this is the fault of a mix of new and old vehicles, 501 Queen and just about every other route in the system has looked like this for years.)

Although there will be more data to digest, the TTC would have a better idea of just what riders face. Exception reporting would quickly flag areas of concern, although from my own experience looking at these data, there would be a lot of “exceptions” until the TTC addresses the problem of service reliability, not just of “on time performance”.

Crowding statistics are produced from time to time, but these rarely are broken down in public reports. This should be a standard part of the CEO’s report. If the TTC does not identify when its service is overcrowded, how can the public or the politicians with responsibility for transit funding and oversight reconcile claimed service provision with rider experiences?

Crowding and headway reliability are linked in that uneven headways lead to crowded vehicles (the “gap” car) and underused capacity. The average load might look good, but the average experience is not. Most riders are on the crowded vehicles and the “average rider” will report an overcrowded trip not to mention the possibility that they boarded the first vehicle to show up with difficulty, if at all.

Vehicle reliability is reported on an overall basis, not for its effect on service. Once a failure causes a delay of five minutes or more, it “counts”, but there is no distinction about the severity of the delay or the actions taken to restore normal service. Delays caused by infrastructure issues or by external events, and again the actions taken to counter their effects, are not reported. There is no metric for how well or poorly service was managed when things went wrong.

A TTC Strategy For Combating Short Turns

At its June 12 meeting, the TTC Board will consider a proposal from management that will officially recognize a practice that has been in place for some time – giving vehicles sufficient running time so that only the most extreme of disruptions would make it impossible for them to cross a route and be on time for their return trip.

The Corporate Plan outlines a vision and mission for the TTC “to provide a reliable, efficient, accessible and integrated bus, streetcar network.” It also identifies five critical paths to achieve this, including Critical Path 3 – Move more customers, more reliably. It commits the TTC to improve schedules and on-time performance so that routes reflect actual operating conditions, are better tuned to demand, and are more reliable for customers. [p 2]

…

The TTC carries out research to understand the needs of customers and non-customers alike. It has identified five shared needs that retain and attract customers: reliability, crowding, wait time, trip duration and affordability.

All five shared needs are closely related, but reliability and crowding are inseparable. Effective route management techniques and a resilient and accurate transit schedule ensures service is reliable and accommodates customer demand. [pp 2-3]

There is a fundamental problem with the concept of an “accurate” schedule, one that reflects conditions on the route, because these conditions can vary substantially by hour, day and season. In a previous article, the travel times on Queen are shown at various levels of detail. Looked at on average, these times change through the day, often hour by hour, and do not fit into the TTC’s breakdown of five scheduling periods. Moreover, average trip times can be ten minutes different on Queen from day to day, and individual values may lie over an even greater range.

The TTC is aware of this problem:

Despite past approaches to preparing reliable schedules many routes are not delivering the service to customers as planned or advertised. Starting in 2015, the TTC has made a coordinated effort to update schedules to reflect actual driving conditions. Numerous approaches to determining the appropriate amount of cycle time (the time for a vehicle to complete a round trip) have been tested. A particular challenge has been maintaining a consistent service level over an operating period. This involves selecting a representative value for cycle time throughout the period that, in most cases, spans at least three hours despite changes in traffic congestion.[p 3]

On 512 St. Clair, the mismatch between scheduled and actual travel times caused average headways to be wider than advertised, and this in turn meant that vehicle loads were larger than the service design should have produced.

Following a schedule change, these lines lay closer together. Note, however, that this was achieved by making the scheduled headways even longer than the “actual” trip times shown in the March 2018 chart above. A change from a planned headway of 3.5 minutes to 5.0 minutes is a 1/3 cut in service capacity. This is similar to what is about to happen on 501 Queen.

The problem here is that the TTC did not just adjust travel times and average headways to match what cars could achieve in service, but they have overshot the mark to guarantee success. This is like correcting one’s archery scores by painting a larger bull’s eye on the target.

The TTC plans to set scheduled times based on the 95th percentile of observed times, and this value will prove generous for many trips leading to some combination of slower operation along the route and backlogs of vehicles at terminals. In theory, however, it should be possible to dispatch vehicles on time, and to a more strict standard, than now happens, because most will arrive at the terminal with time to spare. This is unlikely to happen in practice given the TTC’s laissez-faire attitude to line management which shows up today during periods when there are few traffic disruptions and existing schedules are well-padded.

Problems will remain with bunches of cars outbound to a terminal. If there are no short turns, there will still be a large gap in front of this bunch as it returns inbound.

A further issue is the flexible nature of operator breaks or “recovery time” at terminals and their departure times. TTC schedules include recovery time on most routes during most periods, but this has nothing to do with operators catching their breath after a long trip. In most cases, the recovery time is allocated to make the headway come out to an even value for the number of vehicles on a route, and especially so that branching services will blend (at least on paper) properly. These recovery times are routinely converted to and from scheduled “driving time” for construction projects, and on some routes the longest “recovery times” are during off peak periods. This encourages a culture of expecting long end-of-line breaks and of running off schedule because it is always possible to get back on time for the next trip.

The TTC’s strategy for improving service does not address basic problems with schedule design and line management, let alone provide meaningful targets and reporting on service quality.

Monday, April 15, 2019

The service actually operated on Monday, April 15 is worth looking at in detail to see just what a “no short turns” approach can produce.

The charts in this section are in time-distance format with time across the horizontal axis, and position vertically. Westbound trips slant up to the right, while eastbound trips slant downward. Cars that are stationary appear as horizontal lines, and the speed of travel is reflected in the slope – the more “vertical” a line is, the faster a car is moving.

The day begins uneventfully with service coming out of the east end carhouses. (When Roncesvalles fully reopens, the early west end trips will originate there.)

The period from 7 to 10 am covers the morning peak period. Among the effects evident here are:

• Congestion eastbound to Lansdowne which grows progressively worse until after 9 am.
• Congestion eastbound between Spadina and Yonge between 9 and 10 am.
• Despite the claim of no short turns on this date, the eagle-eyed will spot a few both at Roncesvalles and at Connaught or Woodbine.
• A problem arises with data completeness in The Beach due to a change in line management so that operators are “short turned” by swapping vehicles rather than turning the vehicles themselves. Because drivers sign off the vehicle tracking system when, or even shortly before, they leave, and somewhat after they take over another vehicle, there are gaps in the tracking. (A signed off vehicle does not appear in the tracking data.)

By mid-morning, some problems of bunching are starting to appear.

• Despite the claim of no short turns, a few are evident at both ends of the line.
• Just before 10:30 am a car sits for an extended period east of Greenwood holding vehicles behind it. This is almost certainly a crew change that did not go as planned. The parade this creates echoes back westbound across the route.

In the early afternoon:

• A bunch of cars arrives at Neville Loop and some have substantial layovers before departing. A gap precedes this group of cars westbound until at Jarvis, and again at Spadina, a few cars preceding the gap appear to be held in an attempt to space out the service.
• Starting at about 2:45 pm, there was an eastbound diversion via Shaw, King and Spadina for what was described in the eAlert as “a mechanical problem”. This did not clear until nearly 10 pm. A side effect was that eastbound service became bunched.

By the afternoon peak, service bunching was severe in both directions, but there were few short turns. Note the gaps westbound from Neville at 17:25 and 17:50 where no eastbound cars were turned to provide westbound service despite an abundant set of candidates.

The travelling gap continues to worsen through the evening widening to over half an hour with no attempt to remedy the problem. This is a profound abdication of the duty to manage service, and an indictment of the empty “value” of having no short turns.

Although regular service resumes before 10 pm, the aftermath of the giant gap continues to be felt across the route. Finally, some runs are short turned in an attempt to restore a reasonable headway, but the effect is not visible until after 11 pm westbound at Yonge (thanks to some short turns at Church Street).

This is probably a “worst case” for the days from mid-April onward, but it is an example of what happens when the goal of regular service runs headlong into a management metric that rewards avoidance of the very technique that is used to unscramble the effect of major disruptions.

Below are full chart sets for the service on April 15, and for headways westbound at Yonge for the month of April. The latter, in particular, shows the huge variation in headways during all operating periods. This cannot be explained away by that standard TTC shibboleth “traffic congestion”.

• 501_20190415_ServiceChart
• 501_201904_YongeSt_MonthHeadways