Service Analysis

by

South Spadina Headways: What the Riders See

The matter of service on Spadina south of King has been before the TTC on two previous occasions, most recently in March 2011.

In brief, local riders and their Councillor, Adam Vaughan, complain that service is poor and riders are packed into streetcars.  TTC staff reply that the average load on a streetcar south of King is 29, and therefore additional service is not justified.  This matter was held down in March in order to get updated riding counts on the route.

The problem with the TTC’s analysis is that it looks at overall averages, not at the specifics of actual experience on the route.  Most importantly, if a service that is supposed to run every 7’30” is badly disrupted and unreliable, then wide headways will be common.  Because more riders accumulate waiting for long gaps than short ones, the “average” rider’s experience will be a long wait followed by an overcrowded car.  This situation exists on many TTC routes, but it is particularly troubling if that is the case on a short route with an entirely private right-of-way.

First, let’s have a look at how well behaved headways are in this area.  Before you even open the file, I will warn you that it’s a real mess.  The intention is to show the mess before attempting an analysis.

South of King Northbound Headways February 2010

Those of you familiar with my previous analyses of TTC’s vehicle monitoring data will recognize this chart as a scatter diagram of headways by time of day.  Each dot represents one car.  The vertical position on the chart is the headway as seen just south of King Street northbound, and the horizontal position is the time of day.

If the service were well-behaved, these points should be clustered on either side of the scheduled headway through the day.  However, the points appear as a cloud with little discernible concentration.  This means that the actual headways seen by customers are essentially random, and they are spread over a wide range.  Many cars run close together, and gaps over 10 minutes are common.

My next step was to break down this information based on the TTC’s standard that service ±3 minutes of schedule is considered to be “on time”.  Of course riders don’t care about the schedule, only the headway, and it is the headway that determines the wait time and average load on a car.  For the purpose of this analysis, I subdivided the information about each car into three categories:

  • Early (headway less than 4.5 minutes on a 7.5 minute schedule)
  • Late (headway greater than 10.5 minutes on a 7.5 minute schedule)
  • On time (headways between 4.5 and 10.5 minutes)

It’s worth noting that, assuming a uniform arrival rate for passengers, that a car carrying a 10.5 minute headway will have over twice the passengers than one carrying 4.5 minutes.

Headway Distribution ±3 minutes South Of King Northbound February 2010

There are four pages for this chart, one for each half-hour interval from 0700 to 0900.  The vertical bars represent each weekday, and the rightmost bar averages the values over the month.

The dark red portion of each bar shows how many cars were “on time”.  The cream and blue portions are the “early” and “late” cars respectively.

On a 7’30” headway, there should be 4 cars in each half hour, although minor variations may give us 3 or 5 where a car expected in one period actually arrived in an adjacent one.

Between 0700 and 0730, things are fairly well behaved, but the service becomes less reliable as the AM peak continues.  At least half of the service fails the TTC’s own “on time” test, even averaged over the entire month.  For some periods and days, most of the headways lie outside of the acceptable range.  This means that a large proportion of the riders on the south end of Spadina receive far worse service than is advertised, or even what is considered acceptable by the TTC’s own standards.

If we considered a tighter standard allowing only ±2 minutes, the situation is even worse.

Headway Distribution ±2 minutes South of King Northbound February 2010

When the TTC reports its riding counts, it does not include any information about the reliability of service.  Moreover, if short turn trips are mixed in, these may be counted and dilute the average even though they are not of use to all riders.

In the case of northbound service at King, this isn’t an issue.  However, short turns southbound do affect the quality of service on Queen’s Quay eastbound to Union Station, as well as westbound for riders wishing to travel northwest to locations on Spadina.  During the AM peak, almost all of the scheduled service reaches Union Station, albeit on an erratic headway.  However, starting about 9am, short turning at Queen’s Quay Loop becomes fairly common and disrupts service reliability on Queen’s Quay.

York Eastbound Headways February 2010

This chart contains five pages.  The first is the monthly headway scatter diagram similar to the one shown above for service northbound at King.  Once again we see a cloud of data points spread over a wide range of values.  It is worth mentioning what the scheduled headway is supposed to be here:

  • AM peak: 7’30”
  • Midday: 5’40”
  • PM peak:  6’00”
  • Early evening:  6’00”
  • Late evening:  6’00”

The next four pages of the chart break down the cloud of data into individual weeks and add trend lines for each day.  These lines show how the typical headways are, overall, higher than the scheduled value.  This is due to short turns at Queen’s Quay Loop.  The variation in headways with many values well over 9 minutes (the high end of the “on time” standard for a 6 minute headway) shows just how far actual service quality is from what the TTC advertises.

Soon, the bus network will switch over completely to GPS-based vehicle tracking, and the data stream is supposed to be made public.  Whether this actually happens depends in part on the success of TTC managers, who prefer to hide information about service quality, in convincing the new Commission of the need for this secrecy.  In our new era of “Customer Service” and “Transparency”, there will be no excuse, and we may finally start to see just how bad service is on the system as a whole.

Meanwhile on Spadina, TTC staff will again claim that there is lots of service.  The problem is that it does not run reliably.  Capacity could be provided out of thin air simply by spacing cars regularly so that wait times were predictable and loads accumulated more evenly.  However, that would require the TTC actually do something about its service rather than gripe about the cost of more service, the lack of equipment or traffic congestion.

by

Trackwork Delayed at Queen & Connaught?

Both the TTC and the City of Toronto websites include announcements of major watermain and trackwork near Russell Carhouse that is scheduled to begin in late May.

This project involves replacement of streetcar tracks including the special work at Queen and Connaught, the west entrance of the yard at Greenwood, and the track on Connaught south to Eastern Avenue.

However, it is unclear whether this will actually occur in 2011 as there have been requests from local merchants that they be spared a second year of construction in a row (the City had Queen Street torn up for watermain work in 2010).

Needless to say, any track construction affecting access to Russell Carhouse will have a considerable effect on operations there.  No details have been published yet on alternate schemes for providing service during the project, whenever it might occur.

When there is a definitive answer on this issue, I will update this article.

by

Roncesvalles Construction Resumes

Completion of the Roncesvalles reconstruction will proceed over the next four months from mid-March to early July.

The City’s Construction Notice explains the staging of the remaining work between Queen and Dundas Streets.

  • March 21 to 30:  Housekeeping work on sidewalks left over from 2010.
  • March 28 to May 16:  Sidewalk work on the east side of Roncesvalles south from Fern to Queen, and on both sides between Dundas and Howard Park.
  • April 4 to 25:  Enbridge Gas will relocate a main between Geoffrey and Howard Park on the west side of Roncesvalles.
  • May 9 to June 13:  Sidewalk work on the west side of Roncesvalles south from Howard Park to Geoffrey.
  • June 20 to July 4:  Road reconstruction on the west side of Roncesvalles from Geoffrey to Howard Park.

From March 27 onward, the 504 King service will revert to the schedule used in fall 2010 with a bus shuttle running from Dundas West Station to Sunnyside Loop.  Southbound service will run via Lansdowne, and northbound service will run via Roncesvalles.  In early May, construction will reach the point where buses can operate southbound diverting via Howard Park, Parkside Drive and High Park Blvd.  Streetcar service will resume in July.

A related project, the reconstruction of track on King from Roncesvalles to Close, is scheduled to begin in late May and continue until mid August.

by

TTC Service Changes Effective March 27, 2011

The list of changes for the “March” (actually almost April) schedule period is rather short because the TTC deferred implementation of service cuts on lightly used routes to May 8.

One intriguing side-effect of that decision is that the amount of service operated will be “over budget” because the cuts were incorporated in budget plans in November 2010, long before there was any public discussion, and at a time when rumours of cuts were met with denials from the TTC.

In our wonderfully new “transparent” Toronto, how many other changes are lurking, unpublished, in the budget?

Changes in Hours of Service

53F Steeles East Staines Express: Two new morning trips will be added weekdays from Morningside at about 5:21 and 5:48 am.

Diversions

505 Dundas: With the resumption of watermain work on Dundas, streetcars will divert via Spadina and College until late June.  Headways at all times will not be changed, but one car will be added to the route to handle the longer round trip times.  There will not be any replacement bus service on Dundas.

504 King: Reconstruction of Roncesvalles Avenue will resume to allow work by Enbridge Gas as well as completion of the sidewalk reconstruction.  During this period, the same bus shuttle that operated in 2010 will run from Dundas West Station to Sunnyside Loop, and King cars will loop through Roncesvalles Carhouse.  The Jane night bus, which runs on Roncesvalles, will not be affected.

In May, work is expected to begin on track replacement and paving between Roncesvalles and Close on King Street.  This will complete the rebuilding of the King route to the new resilient track standards.  While this is in progress, King and Lake Shore cars will divert via Queen and Shaw.  The work is expected to be completed before the CNE (with associated traffic problems in Parkdale) opens in mid-August.

Unlike the period of watermain work on King West in 2010, there will not be any replacement bus service on King west of Shaw.  I can’t help wondering why the TTC doesn’t simply route the 504 to Dufferin Loop so that service would be maintained for riders from Shaw to Dufferin whose access to Queen is limited by the rail corridor.

by

If One Bus Is Good, Seven Must Be Better

Recently, I received a note from a reader about bunching on the 29 Dufferin bus.  This is one of two routes on which increased supervision has led, or so the TTC claims, to improved service (as discussed in my recent article about the Customer Service plan).

I’m wondering what TTC is doing to prevent bus bunching as this seems to be a big problem on many routes…

I wrote to TTC customer service but of course I didn’t hear back from them. I would be interested to know what measures are in place to space out the buses. It seems to me that buses are just rushing to get to the end of the line; they basically do not care if they have 2-3 buses traveling together or if the last buses have no passengers.

I took this video as an example, there are 7 buses passing Bloor stop in under 3 minutes … such a waste from TTC. Just imagine being the “lucky” one that just missed all 7 buses. How long they will need to wait for another bus?

The TTC talks a good line about service management, but a casual look at any of the routes where real-time monitoring is possible (through the open data interface to GPS vehicle data) routinely shows bunching even at 7am when there cannot possibly be “traffic congestion” effects.  Two basic questions about bunching emerge from all of the service reviews I have done:

  • Why are vehicles allowed to leave termini very close together rather than regularly spaced?
  • Why are vehicles entering service from yards or from short-turns not spaced between through runs so that even headways are provided?

This does not require millions in high technology to implement, only the will to manage the service, something the TTC once did regularly with no more than route Inspectors on the street.  With the tools now available for vehicle tracking, it should be much easier as all vehicle locations are available online.

I plan to review the Dufferin bus operations again in the Spring (my last review was published in January 2008) once the effects of winter weather are not a concern.

Short turns disrupt service near the ends of routes, and on Dufferin this can show up as poor service to Liberty Village thanks to turnbacks southbound at Dundas.  On King, short turns eastbound at Parliament and westbound at Roncesvalles deprive riders on Broadview and on Roncesvalles (which have demands to some degree independent of the route downtown) of service.

If the TTC is going to strive for having 99% of the service advertised actually on the street, it must also strive to have that service on the route, not sitting in a short turn, or running in a pack followed by a large gap.

by

The “Ooops” Factor in Planned Service Cuts (Updated)

Updated January 20, 2011 at 9:30 pm:

The table showing the route cuts linked in this article contained an error, and this has been corrected.  In the original table, I calculated the imputed cost/hour and cost/passenger of the services under review by dividing the annual vehicle hours saved and passengers affected into $7-million.  However, that saving would only have taken place over 9 months, and its annual value would have been about $9.33m.

Therefore, the imputed values, corrected for full-year savings, are $70.74 per vehicle hour and $7.16 per rider.  Some reports have claimed that the current subsidies are $30/rider, and this is not borne out by the TTC’s own numbers.  The actual values will vary from route to route and day to day, but the average is about 1/4 of this claimed value.  Indeed few of the services up for cancellation reach the $30/rider subsidy level as is evident on the updated and expanded spreadsheet.

The link to the table is reproduced here for convenience.

2011.03 Route Cuts Analyzed

In other news, I have now learned that the service improvements implemented in January 2011 are to be funded out of the savings from the late night cuts.  These improvements have an annual cost of about $4m, leaving only $3m for fall 2011.  However, about $1m will be lost to the delayed implementation of the cuts, leaving only about $2m for service additions in late 2011.

Updated January 13, 2011 at 3:05 pm:

Maps added:

Map of routes facing proposed cuts

Map of routes with proposed additional service (Fall 2011)

It is worth noting that some routes appear on both maps indicating that they have a proposed service cut in one period to pay for an improvement in another, although we don’t know exactly what that may be yet.

[The original article follows below.]

Continue reading →

by

Analysis of 512 St. Clair Operations for July 2010 — Part III (Headways)

In two previous articles, I have reviewed the St. Clair car during its first month of operation on the new right-of-way over the complete route from Yonge to Keele.  Running times during busy periods are down compared with April 2007, when the only right-of-way was between Bathurst and Yonge Streets.  However, the situation with headways, an important factor in how riders perceive service quality, is quite another matter.

For the entire period of construction, the idea of regular, scheduled service was something of a fairy tale on St. Clair, and both the streetcars and buses made their way such as they could along the route.  One would commonly see vehicles taking long terminal layovers, and headways were not a big priority.

In analyses of other routes, there is a common factor that is independent of the route’s length, the time of the year, the weather, eclipses or any other phenomena:  vehicles do not leave terminals on a regular spacing.  They leave when they get around to it, a practice abetted by the TTC’s standard that ±3 minutes is considered to be “on time”.  Pairs of vehicles can travel together on routes with short headways and remain within this standard.

Continue reading →

by

Analysis of 512 St. Clair Operations for July 2010 — Part II (Link Times) (Updated)

In the first article of this series, I gave an overview of the data for one day’s operation on the 512 St. Clair route in July 2010.  Here, I will review link times (the time taken to get from one location to another) on the route for the entire month, and compare this with data from April 2007 when the St. Clair car last ran over its full length from Yonge to Keele.

Updated December 29, 2010 at 9:50 am:

A set of charts has been added comparing the running times between Keele and Yonge for April 2007 and July 2010.  See the end of the article for links and commentary.

Continue reading →

by

Analysis of 512 St. Clair Operations for July 2010 — Part I (Introduction)

Updated at 3:20 pm, December 27: The scale on the headway charts has been changed to 30 minutes with 3-minute gridlines, and on the link time charts to 18 minutes with 3-minute gridlines.  The intent is to spread out the data points to give a better view of the fine details.

[My apologies for the appearance of this series many months after the fact.  It took quite a while to get the GPS-f0rmat data from the TTC for reasons unknown.

I look forward to the plans for an Open Data access to data for all routes as a regular online service so that this type of analysis will not require special requests for data extracts.  Whether the “new” TTC goes ahead with providing this data remains to be seen, although “transparency” is supposed to be a watchword in the new administration.]

June 30, 2010, brought streetcar service to the full St. Clair route out to Gunn’s Loop for the first time since 2007.  In an earlier article, I reviewed the line’s operation in April 2007.  This was a “before” snapshot intended as a comparison to the “after” construction line behaviour for which we have waited so long.  Now we can look at the “new” St. Clair to see the benefits, such as they might be.

On two weekends, service did not operate over the entire line due to street festivals.  No sooner was streetcar service restored to Keele, but it vanished again on July 3-4, and again on July 17-18.

This article reviews the basic information available and some of the analysis I have done using a single weekday as an example. Continue reading →

by

Parliament and Roncesvalles 2010 Track Work (Update 11)

Updated December 19, 2010: Streetcar service resumed today on Roncesvalles Avenue to Dundas West Station.  The construction is not yet finished and this has predictably upset the neighbourhood.

The TTC seemed unusually ineptly prepared for this changeover.  Electric switches at many locations had not been reactivated requiring operators to throw points at commonly used junctions by hand.  The Sunday Stops on Roncesvalles which were not supposed to be part of the new design remain in place both at stop poles and in onboard stop announcements.  Indeed, the location of some stops appears to be a leftover from the shuttle bus operation.

At least one errant auto, parked in the wrong direction and foul of the southbound track, was struck by a passing streetcar.  Permanent signs indicating where people can and cannot park don’t exist yet, although a number of temporary “emergency, no parking” signs have appeared.

Anyone interested in watching service reliability can do so via various monitoring sites.

Meanwhile, Parliament Street reopened to regular traffic recently, and this morning, both the King and Dundas cars diverted bothways via Gerrard and Parliament to bypass construction on Broadview.  No pointman was provided for the westbound manual switch at Parliament, although on previous occasions the TTC has spent a small fortune manning this location for diversions.  Why the switch isn’t electrified is a mystery considering how frequently this diversion is used.

Continue reading →