Analysis of 29 Dufferin for March 2012 — Part I: Headway Reliability

[Apologies to those of you who are pining for even more articles on the Queen car.  They will show up in due course.  I have been diverted onto Dufferin by recent events.]

Service on the 29 Dufferin bus has been a burning issue for decades.  Buses run in packs, they arrive full of passengers, and the advertised service bears little resemblance to what riders see  on the street.

Recently, this was the subject of an article in The Grid by David Topping.  In it, the TTC’s Brad Ross trots out many of the usual explanations of why service is unreliable.  In deference to Brad (who is really a nice guy), I don’t want to spend an article eviscerating his comments line by line.  I will leave readers to contemplate information in this and following articles and make up their minds.

What riders see is “headways”, the time between vehicles, as well as the degree of variation in that time.  If the TTC says a bus will appear every 5 minutes, and the service manages to achieve this, more or less, most of the time, then a rider will consider this “reliable”.  Even on a wider advertised headway, if buses appear at roughly the expected interval, riders know what to expect.

However, if the headways vary widely from the scheduled value, this makes a service unreliable and riders must, at a minimum, build in additional travel time to account for the possibility of a long wait.  Moreover, at the end of the wait, they may be faced with a jammed bus they cannot board.  There might be another one (or two) right behind, but that makes no difference to the length of the wait, and those buses might not be going to the rider’s destination.  Providing frequent service “on average” is not what riders want to see.

In this article, I will review the actual headways provided by the Dufferin bus at various locations during the month of March 2012.  Although this is technically “winter” (most of it), 2012 was a balmy year and the route operated without the kind of severe weather delays we have seen in 2013.

In future articles I will turn to running times and the effects of congestion on the route’s ability to maintain regular service.  I will also look at a few days’ operation in detail to see exactly what was going on.

The information used for this analysis comes from the TTC’s GPS-based vehicle tracking system which reports the position of every vehicle every 20 seconds allowing fine-grained resolution of movements at any location.

Service Levels

When looking at the actual service, it is important to know what the schedule claimed the service should be.  Route 29 Dufferin was among the routes whose service was cut in February 2012 as a budgetary measure.  The loading standards were changed to allow more riders per bus, on average, during various periods of operation.

The following table shows the pre- and post-cut headways, and also includes notes about service improvements since March 2012.


Observed Headways

In the charts linked below, each set contains eight pages with a similar layout.  In all cases, time reads across the X-axis and length of headway reads up the Y-axis.

The first five contain data for the weekdays subdivided by weeks within the month.  March 2012 had no statutory holidays (Easter fell in April in 2012) and so the only variation comes in the first short week which has only two days.  Weeks 2-to-5 have five days each.

The sixth chart contains all of the weekday data on one page to show a “cloud” of points and the general distribution of data for all of the weekdays in the month.

The seventh and eighth charts contain data for Saturdays and Sundays respectively.  These days are broken out because of their different traffic and service patterns.

Trend lines on all charts (except the sixth) show the best fit of each day’s data.  These generally lie at about the value of the scheduled headway except when service is erratic and infrequent.

Summaries of Observed Headways

The first set of charts show all of the data with one point plotted for each bus passing a location.  Another way of looking at the same information is to summarize it by time period and by the value of the headway.

In the summary charts, the number of buses at each headway (1 minute, 2 minutes, etc.) for a period such as the AM peak are summed, and these values are displayed as percentages.  The TTC aims to have 65% of its bus service within 3 minutes of the scheduled headway.  If this actually happens, a chart will have a peak at the scheduled value and will quickly drop off on either side.  However, in some cases, the values are spread out over a wider range.  In others, the most common values seen are 1-2 minutes (bunched service) with a wide range of other values some of which are far from the scheduled headway.

Northbound Service


Although the Dufferin bus has a branch that runs through Exhibition Place during certain hours from Dufferin east to the Princes Gates, the main service originates at Dufferin Loop at the Exhibition’s western gate.  King Street is a few blocks north and is the first major cross-street.  Service at this point should be fairly well-spaced because most buses have just left the terminal.

Many of the data points are clustered around the trend lines, at least within a value of ±3 minutes which is the TTC’s goal for service punctuality.  Because of the low scheduled headways, this means that many “punctual” vehicles are running very close to the one in front.  In other words, they left the terminal in a bunch.

By the time we get to Bloor and the connection to the subway (2.4 km north of King), the weekday data points are roughly in the same pattern as at King.

At Lawrence (a further 6km north), part of the peak service has short turned at Tycos (half-way between Eglinton and Lawrence).  The data points are more widely spread especially in the PM peak.  The wider scheduled peak headway on the northern section of the route simply compounds the general unevenness of peak period values, well beyond the TTC’s 6-minute target band.  This is an example of how a metric that looked only at certain parts of the route (say, at Bloor), or which averaged various points together, may mask problems on selected parts of a route.

On weekends, especially in the evenings, the scheduled headways are longer, but the distribution of actual headways is very wide, much beyond the six-minute range.  The situation is bad at both King and Bloor, but by Lawrence headways bear little resemblance to the advertised service.

The headway summaries tell the same story but present the data in a different manner.


Northbound at King, the weekday service headways lie mainly in the ±3 minute band, but the data spread out as the day wears on, especially in the evening.

Saturday mornings fit the target pattern, but by Saturday afternoon service is running with headways below the scheduled value (1 or 2 minutes) being the most common, and the rest of the service spread out over a range up to 10 minutes.  Saturday early evenings continue this problem.  By late evening, the headways range fairly evenly over a range from 0 to 15 minutes when the schedule calls for 11’00”.

Sunday early mornings should be a period when it is easy to maintain a scheduled headway, but even then, there is a wide range of headway values.  The range tightens up a bit in the late morning, but there are still many buses running close together on a scheduled 5’00” headway.  Sunday evenings show headways over a wide range of values.

The common factor here is that even near the route’s southern terminus, the service is not operating on a reliable headway much of the time.  The only possible explanation is that buses are not leaving on time because they have only travelled a few uncongested blocks when they reach King Street.


By the time the service reaches Bloor, some of the unevenness seen at King has worsened.  One minute headways are the most common value seen for most of the time on weekdays, and this shows that service has become even more bunched as it travelled north from King.  In the late evening, service is spread out more than the six-minute range around the scheduled 7’15” headway.

On Saturdays, as at King, the service moves further away from the target headway as the day goes on.  By the afternoon, over 25% of the buses arriving northbound are on headways of 1 minute (this bar actually represents values from 0’30” to 1’30”), and service is badly bunched.  This continues into the early evening.  By late evening, as at King, the headways range widely with much of the service outside of the target six-minute band.

Sundays continue the pattern seen at King Street with a wide variation in headways especially in the evening.  It is very difficult to believe that this is the result of traffic congestion, but rather the complete absence of any attempt to dispatch service at headways resembling those on the timetable.

Southbound Service


Southbound headways are shown first at Transit Road and Wilson, just outside of the line’s northern terminal.  The next point is Eglinton, 4km to the south, and then Bloor, another 4km down the line.

At Transit Road, weekday service shows a similar pattern to that at King Street northbound with most vehicles clustering around the trend line.  As at King, the scheduled headway is short enough that “punctual” service can actually be operating closely behind its leader.  A greater scatter of headway values shows up here in the PM peak just as it did northbound at King.

By the time we reach Eglinton southbound, the service is a bit more spread out (compare the width of the weekday “clouds”) than it was at Transit Road, but some of the longer headways seen in the PM peak at the north end of the line have vanished.

At Bloor, the values have spread out further, and headways above 10 minutes are common throughout the day with no real indication of a peak period effect at all.

Weekend values show a lower service quality with a greater spread of values and bunching is still common.


As with the northbound service, what we see at Transit Road southbound is fairly well-behaved headways until the PM peak, but then the values spread out.  However, there is not the same degree of bunching as the service leaves its northern terminus.

Weekends show the same pattern of spread out headway values in the evening as we saw northbound.


At Bloor southbound, buses are near the end of their inbound trips.  During the AM peak, very short headways are most common showing the bunching that has occurred enroute from Wilson Station.  This continues through the day and even into late evening when a considerable amount of service is running close together on a nominally 7’15” headway.

Saturday service is fairly well-behaved until the afternoon when short headways dominate showing badly bunched service.  This continues into the evening when headways well beyond the six minute band are common.

Sunday service starts off with widely scattered headways and never reaches a point where values lie mainly in the target band.  By Sunday afternoon, bunching is common, and in the evening headways are over a wide range of values as on Saturday.


Headway reliability on the Dufferin bus is very poor and this occurs even at times when there should be little interference in service from traffic conditions.  On weekends evenings, service is unpredictable because the headways lie over so wide a range and many buses are running in groups of two or three.

The observed data suggest that for some operating periods, nobody is “minding the store” and no attempt is made to hold vehicles to the target headways.

38 thoughts on “Analysis of 29 Dufferin for March 2012 — Part I: Headway Reliability

  1. I may understand why people may exit buses (or streetcars) if the center doors are blocked by a snow (or ice) windrow. However, if the stop is clear, I would always use the rear doors to exit the bus. Having able-bodied continue to use the front doors to exit greatly slows down the service, as the people wanting to get on have to wait longer. In turn, it creates even more bunching of buses.

    The other problem is people who see a couple of buses trying to board the front bus, while the following bus remains empty. I remember even on the Bloor or Queen MU streetcar trains, people trying to board the front streetcar while the following streetcar (coupled) having few passengers on it. This is a worry of mine that when the new light rail trains come into service, I might see crowding on the front vehicle. Hopefully, having far-side stops may negate that.


  2. Steve —

    Thanks for the detailed analysis. I’m a regular user of the Dufferin 29, so this hits close to home for me. A few thoughts come to mind:

    1. The Dufferin 29D route extension: this should be considered an added bonus or convenience (when it is running), rather than an integral part of the route, because the TTC discontinues service on this extension whenever anything is happening at the CNE (including set-up and take-down activities for events). The 29D doesn’t run for most of the summer (save for a few weeks here and there) or during a long period in November (Royal Winter Fair), but Heaven help you if you try to find out about these disruptions from the TTC — there is rarely any notice when they occur.

    Steve: That’s why I omitted the 29D branch from discussion. Looking at the day-by-day data, it can be unpredictable when a bus will actually show up.

    2. One interesting way to look at headways would be to “bin” the headway data into 30 or 60 min chunks, then calculate the mean headway and standard deviation for each of these chunks. I can imagine this would be a lot of work, but it would show how accurate and how precise the service is over the course of the day.

    Steve: The data is already in that format in the headway summary charts, although it is consolidated across several days to get a good representation of data. I could include means and standard deviations, but most readers don’t understand what these numbers represent. The pictures, however, tell the story. Anyone can tell the difference between closely grouped and widespread data values just by looking at them.

    3. In fairness to the TTC, Dufferin is generally a crowded street, with lots of on-street parking and no left-turn lanes south of about Castlefield. On weekends, it is not unusual for the line-up of southbound traffic waiting for a green light at St Clair to be backed up almost all the way to Rogers Road. That said….

    4. Unless I am reading too much into the data, your stats on northbound headways at King seem to be a “smoking gun” that can only be explained by driver misbehaviour or a total lack of oversight of the route by the TTC (probably both). The fact that the headways are so widely distributed at the third stop on the route should be a wake-up call to the TTC, and I hope someone in a position of power there takes notice and does something to correct the situation.

    Steve: One important point I am making here without being too noisy about it is that this is a busy bus route on a congested street that is 4 lanes wide. Anyone who thinks that replacing streetcars with buses will improve service reliability doesn’t know what they are talking about. The underlying problem is line management and the adequacy of capacity to handle riders. These are common to both modes.

    5. As a side note, I once contacted my City Councillor’s office to ask whether there were any plans to tackle congestion and transit problems on Dufferin — and was asked to explain what those problems were. It’s as if they had no idea that there was even an issue. Not very encouraging overall.


  3. I looked at the graphs and it was not pretty. There is no doubt that a lot of effort has been put in to try to understand route 29 riders face on an average day. Transit is not only a question about schedule adherence on a line. Without O/D information, we just cannot see the true impact or other ways to mitigate this problem.

    Not all riders originate on streets close to Dufferin. For example, someone may choose to use route 29 because it is more frequent than the Ossington or Landsdowne bus a block away. Someone may use this route to avoid having to transfer at Bloor as in the case of route 7 and 126. If parallel service can be increase, then the impact of his awful schedule adherence would affect fewer passengers. With fewer connecting passengers, crowding issue would also be eased.

    In addition, the impact of waiting for a wolf pack of buses does not end after boarding the 29. What happens if that person needs to connect to Mississauga Transit at Islington Station? Arriving 10 minutes late could mean another 20 minutes wait. It does not matter much on most of the TTC network as it runs very frequently. If one needs to connect to a not so frequent service like MT, it will sour satisfaction ratings very quickly.

    My advice to the TTC is this, split the 29 service to allow for an express service. This will get the connecting passengers off the local service. These are the people who take up the seat for the trip. With a bus skipping every two to three stops (on average), at least a full bus does not need to stop at every stop trying to fit just one more passenger on board.

    Steve: The Bathurst bus is no picnic either and is not necessarily an alternative to the Dufferin bus. You mention O/D information: an important point here is the degree to which traffic originates at local stops rather than at major transfer points that would be express stops. That’s always a balancing act.

    Proposals for express services on many bus routes were killed off thanks to budget problems even before the Ford crew came into office.


  4. Interestingly, most of the extreme headway gaps northbound at King have closed up a bit by Bloor. This can be seen in the daily clouds, as well as the aggregated bar charts.

    Which suggests to me that some operators leave late for reasons unknown, and then flog their buses and have made up a few minutes of the gap by Bloor.

    I wonder if there are any systematic running time differences between the Vs and the VII hybrids. Does your data include bus number? Obviously the NextBus realtime feed does.

    Steve: Yes, I do have the bus numbers. When I turn to running times in the next article, I will have a look at whether the vehicle type affects the time over various sections of the route.


  5. The way the data is presented doesn’t really give a clear picture of how well the route is performing. Partially because it’s so crowded, but also because they give the impression that extremely short head times balance out long head times and as you mentioned both of those indicate problems.

    Is the data used to produce the graphs publicly available? I’d like to play around with it to see if there’s a better way of visualizing it. Thanks.

    Steve: The raw data are quasi-publicly available in the sense that I have a standing arrangement with the TTC to receive extracts. I ask for specific routes and months from time to time because they may show something of interest, or be part of a series I’m looking at. Most of the stuff I look at never shows up as articles because people would tire of an unending stream. (They would also notice how similar things look in various examples showing that issues with service quality and management are fairly common across the system on major routes.)

    A few years ago, the “old” Commission passed a motion that this data should be posted publicly on a standing basis, but that was never implemented. I can send you my copy via email in zip format.

    There is a big jump from the raw data to what you see in the charts because the data starts out as GPS co-ordinates that have to be (a) cleaned and (b) mapped to whatever route one is analyzing. Over the years, I developed a number of programs to automate this process and transform the data. I have an article in preparation that explains how this is done, but it’s not ready for publication yet.

    If you have specific examples of how you would like to portray the data in a different manner, please leave a comment here, or we can start a private thread via email.


  6. Saturday late service and Sunday after about 7 there needs to be more buses on this route.

    Steve: As we will see in future articles, a big problem on weekend evenings is that the operators appear to be ignoring the schedule and running on whatever headways they feel like, complete with various short turns. Service would be greatly improved simply by ensuring that the vehicles assigned to the route actually ran on a dependable headway.


  7. “Anyone who thinks that replacing streetcars with buses will improve service reliability doesn’t know what they are talking about.”

    Steve: It seems appropriate that you might mention this today. Even on “Anyone’s” personal press release newspaper it was noted that the plan is somewhat impractical given the capacity of streetcars vs. buses and the congestion that twice as many buses as streetcars might cause. Reference was also given to the large investment in new streetcars and the current new barns under construction.

    Also, in the newspaper article, based on an interview given on the strangely appropriate Buffoon TV, it was stated that the reason “Anyone” might oppose streetcars is because they block the progress of an SUV making its way downtown. Even on this retrogressive bottom of the barrel journal, internet commenters mentioned that the SUV was being blocked by a streetcar with many passengers, compared to the one in the (self driven) SUV. It seems to me that “Anyone” who wants to replace streetcars with buses is not only uninformed, but actually a little selfish.


  8. Steve, are you still generating those time vs. distance charts that show the progress of individual runs throughout the route? Those are quite effective in showing patterns of gaps and bunches, as well as showing where delays are along the route (especially with the increased accuracy now that positions are reported frequently via GPS).

    When looking at the cloud of data points, and the trend line, it struck me that the trend line doesn’t pass through the centre (average) of the cloud. Rather, it’s below the centre (the weighted average), with most of the data points below the trendline offset by a few data points far above the trendline. In other words, we don’t just have an average 5-minute headway consisting of two buses every 10 minutes — we often have an average 5-minute headway consisting of three buses every 15 (or 4 every 20). This looks to be consistent with the complaints in the Torontoist article. It doesn’t help that the spreading occurs in the afternoon rush when ridership (and visibility) is highest.

    Steve: Yes, I still produce those time/distance charts, and you will see some in a future article when I look at the details of specific days where service was notably different from the overall pattern. Recently, I decided to publish the overall headway and running time charts first because these relate fairly directly to what people see on the routes and show the dispersion of values. What is interesting is that this dispersion is fairly well-behaved and this is evident from the similarity of trend lines and data point clouds from day to day.

    The time/distance charts require a bit of study to see overall patterns, and it’s better to come to them in search of the details for something odd that shows up in the monthly charts. Believe me, you don’t want to look through 31 days, 8 pages each, of time/distance charts.

    The behaviour of the trend lines is not surprising with so many values clustered near zero. This shows up in the headway distribution charts where instead of a nice symmetrical curve on either side of the target headway we see a large percentage of very low values with the rest of the service smeared out above it. Where the “cloud” actually does have a uniform distribution on either side of the target value, the trend line is in the “right” place. There is no perfect way to deal with this because different time periods and locations have different distribution patterns in their data.

    My hope is that the average interested reader can see the general patterns and understand how they demonstrate with TTC data what they see every day on the street. With the new focus on customer service including better reporting and an effort to provide more reliability, this sort of thing will make interesting comparisons in a year or two when the effects of any new protocols should be evident.


  9. It would be interesting to ask Brad about what they do differently at different times of the day…

    I think probably rush hour the only challenge is when there is a major delay (in which case they just need to organize short turns and potentially alternate service), most of the headways really are kept low just by the sheer amount of service…and traffic would only cause problems during these times…

    But more critical would be managing when there aren’t a lot of vehicles on the route, because the potential for headways to grow is high, and there are less options for short-turns (which could end up creating bigger gaps elsewhere), it seems on weekends etc. it really comes down to managing when driver’s leave stations, whether they have built in delays in their schedule or not, and making sure there aren’t “forced gaps” due to staffing or shifts…

    I just wonder what the staff ratio is for route management is for rush-hour vs weekends, evenings etc.

    Also for most people – being late to work due to TTC is part of life…but it’s way more frustrating to be late for a party, or kids soccer, or personal activities – the TTC loses a lot more friends during these periods than they do during rush-hour…especially because buses already come less often.

    Steve: This is a common pattern on routes I have analyzed. For all the problems of peak period bunching and crowding, the issues are much worse during off peak periods when there is less service on the street. This is particularly true evenings and weekends. In some cases, such as streetcars fighting their way through the club district, there is a clear external reason, but provision for this should be built into schedules so that short-turns are not required from a predictable delay. In other cases, there is no indication that traffic or passenger congestion is the problem, and I have to put unpredictable headways down to poor/absent management of service and operators who make up their own schedules.


  10. There was an interesting comment thread in the Grid TO article that said (based on observations at the south terminus) that much of the problem in the south end is because of drivers taking breaks together and leaving together.

    Your data seem to suggest abnormal problems with the south end of the line, meaning that be the case…meaning the TTC might want to consider reviewing their line management.

    One question I’ve had is about making changes to the route. I feel like the route could be divided in 2 routes ending at Eglinton West (shifted closer to Dufferin and Eglinton when The Crosstown opens) with one limited stop “express” that stays on Dufferin.

    Cheers, Moaz

    Steve: The problem of buses leaving in packs was identified in an earlier analysis I did of the Dufferin route many years ago (Parts I, II and III). Nothing has changed, and it is clear that the TTC either does not understand what is going on, or chooses to ignore the situation. This type of behaviour (which can be seen at other locations) is a major challenge for the TTC, and it will be interesting to see how the new focus on “customer service” addresses this.


  11. Could the city not simply eliminate the street parking on Dufferin? That would help the buses not having to move in and out of parked cars. Also can the new streetcars make it up the Dufferin hill? I agree with you that streetcars themselves would not solve the problem. However streetcars combined with no on street parking might do the trick.

    Steve: We are unlikely to see parking removed from Dufferin, but in any event a lot of the problems in the bus service are outside of the peak period, the most likely time any restrictions will be in place. As for streetcars, they could handle the hills (and the new underpass at Queen was designed to ensure that they would be able to get through there. However, articulated trolley-buses are more appropriate for this route. The TTC looked at this years ago and one big reason the proposal was blocked was the still-active runway at Downsview. Overhead wires cannot be strung along Wilson south of that runway.


  12. Any attempt to limit on-street parking on the southern stretches of Dufferin would be vehemently opposed by city councillors.


  13. What if the trolleybuses turned on Billy Bishop Way instead of Wilson? Could that not be an easy fix. Suggesting Streetcars is a admittedly selfish request as I avoid the bus like the plague because I find the ride rough and inconsistent. That being said I was under the impression that the new Streetcars could handle more people and Dufferin was one of the city’s busiest routes.

    Steve: Yes, it is possible that Billy Bishop Way would provide an alternative route from Dufferin to Wilson Station that would not violate clearance rules.

    Streetcars can handle more people, but unless we are going to plan for substantial expansion of the streetcar (as opposed to LRT) network in Toronto, I am not sure Dufferin is a good idea as a one-of. Conversely, a network of trolley coach routes could operate some of the major north-south lines that have to deal with hills.


  14. Every time I read one of these posts, it seems so obvious that making sure vehicles left terminals on an evenly-spaced headway could yield a huge improvement in service quality. And it’s “low-hanging fruit” — could be as simple as stationing a supervisor there; at most it’d take an extra operator for step-back crewing.

    But given that all these years later the TTC seems unlikely to make a concerted effort to try this and see what happens, I wonder if there’s a way to find the answer in the data. Is the data organized in a way that would make it possible to produce a formal correlation between departure spacing and headway reliability further down the route?

    Steve: I suppose that could be done, but I have seen a simpler example of this from data for individual days organized by location. Starting at one end of the line, the headway graph follows a familiar up and down pattern of wide and narrow values. With successive graphs showing locations along the route, one can see almost an “animation” of the service pattern as the peaks and valleys move slightly on each image due to the progression of time. The peaks and valleys generally become more pronounced, not less, except at locations where short turn vehicles merge into the stream. I will include a few of these in later posts when I turn to the daily detail of the route.

    What this shows quite simply is that the service starts off badly spaced and only gets worse along the line, and there is no sign of an intervention to re-assert proper spacing part way through.


  15. Considering the height of the Best Buy and Home Depot, it would be unlikely that Billy Bishop Way would pose a clearance issue. That said, just how much (say per km) would it cost to implement trolleybus artics/40 footers? How much (more) would each vehicle cost? It even possible to have wires alone both sides of Dufferin/Bathurst (or whatever route) to allow buses to pull in and out of curbside stops/on street parking?

    Steve: I will have to dig out some current construction costs for new TB lines. As for the overhead, generally one set of wires allows a vehicle to reach three lanes (two travelling lanes plus the curb to serve stops). Bypass wires can be added to allow TBs to skip around each other, and of course off wire operation is available for emergency diversions.


  16. Could they just put gates on the exits to the terminals? Only let one vehicle through and only do it when the “computer” says that it is ok to go? They could add a sensor to check that the vehicle actually leaves, and then close the gate….seems like this would work at a lot of places…such a system would essentially be fairly straight forward (the gate technology exists at all parking lots) – just hook it up to a program that analyze’s the GPS data instead of a ticket reader.

    Obviously this only works at terminals where the geometry allows it, but for the locations where it does it might be very useful…at other locations a stop light or a signal of some type (maybe a sign that says which busses and streetcars can leave, and which can’t) would be a good middle ground.

    Steve: The operators already have a display that tells them where they are relative to schedule, but it is driven off of the old “signpost” system, not GPS. Also, once service is disrupted in any way, a computer system is not going to be able to figure out the appropriate times to open the gate. The basic issue is that cars should be dispatched, spaced and monitored, and operators should be able to understand where they are expected to be relative to preceding and following vehicles. Yes, that’s not easy, but the first requirement is the willingness to do it.

    By the way, how many parking lots have you seen with the exit arms broken off because they don’t work properly?


  17. How much of this is due to the TTC service cuts and loading standards currently in effect? Would a return to pre-2011 service levels be of some help?

    It seems to me that because of the very low operating subsidizes gets from the city, province, and federal governments (insert sinister laughter here), that the TTC will remain behind city service levels outside of North America. Personally, I think the service by the TTC is still better than most other North American cities, but less than those outside North America.

    Steve: I included the table showing the before and after services in the article to show that, for the most part, the Dufferin bus was not affected by the cutbacks in all periods. Some that have the most off-target service were not affected at all.

    However, ridership growth would have built into the additional “capacity” created by the changes in standards, and it would have taken longer for the TTC to react. As also noted in the table, some service periods have been improved, but not necessarily the ones that were cut in Feb. 2012. The data shown are from March 2012, and so any such growth will not have had time to affect loadings.

    Two questions do apply here. First, how much capacity is wasted with unevenly spaced service — overfull buses at the head of the pack followed by one or more that are only partly loaded. Second, what is the latent, unmet demand from people who would use the route, or who abandoned it because of unreliability?


  18. The main problem with line management in most cases is that a supervisor is running usually 4 lines at once, in some cases 10. They also handle emergency calls, transit control, etc. and use an outdated CIS. Schedules in most cases do not provide enough running time (especially off peak and weekends). Until staffing and technology gets changed it’s most likely to stay the same.


  19. I would argue that the service cuts (and increases) shown in the table are essentially meaningless, given the consistently high fluctuation in headways throughout the day, every day. For example, during the midday period Monday to Friday, an extra three buses were added to the route (an 11% increase in vehicle allocation and operating cost), which had the effect of decreasing the average headway from 3’40” to 3’20”. It would seem to me that this minor decrease is lost within the wide variability in headways. You could probably achieve the same increase in capacity by reducing the bunching along the route, especially if this is within the TTC’s control by more proactively scheduling departures from the terminals.

    Steve: I definitely agree. For years, the TTC has talked privately of how the cheapest capacity is there for the taking with better headway regularity, but then they don’t step up to the job of providing it.

    With the introduction of the customer charter, TTC brass have been talking about “quick wins” and easily implemented measures to make the TTC more accountable and to make riders’ trips more pleasant. But the key is to challenge the TTC’s culture. There are a number of things that make service frustrating for riders and inefficient for the TTC that, to a lay person, seem easily addressed if not for internal culture, workplace politics etc.:

    Surface vehicles leaving the terminal in bunches.
    Short turns that affect a vehicle more than xx% (half? one-third?) full. (Someone suggested that short turns are more effective when there is a more active partnership between supervisors and operators, so that the most appropriate vehicle(s) of a bunch is turned and the rest are unaffected.)
    Operating practices that encourage riders to try and jam on the first vehicle of a bunch, when one does occur, and that prevent trailing vehicles from sharing a greater load (e.g., encourage bunched buses to leapfrog to share the load and travel faster; permit/encourage operators to bypass stops if their vehicle is full and if another vehicle is immediately following).
    Transit priority detector loops that don’t work. (I think someone said that these are under the control of Transportation Services, not the TTC. Move the responsibility and funding to TTC, who have a direct interest in ensuring appropriate maintenance, as well as expansion of the network. Actively solicit operators for suggestions for improvements — they see the conditions every day and know where congestion is a problem that might be addressed through signal priority, lane reallocation, modest construction etc. Provide incentive for suggestions re: priority measures that can be implemented with a demonstrable improvement on service.)
    Automated switches that don’t work.
    Orders to stop before proceeding through all switches, especially aggravating on Spadina where it slows down service, bunches up cars, and increases the number of cars needed on the route.

    Address these things and you speed up service (happier passengers; maybe enabling more frequent service), and make more efficient use of the capacity that’s currently unusable because of bunching. It’s like a free service increase across the system.


  20. Dufferin is unlike almost any other TTC bus route. The amount of time it takes people to appear at the stop is unparalleled, the bus ahead can be less than a block away and there will already be people at the stop before the next bus arrives. The sheer speed at which people accumulate at stops drastically increases the likelyhood of “the snowball effect” where a bus that misses a few too many lights or is delayed for any number of reasons becomes late and is quickly overwhelmed. Of coarse the bus behind it will catch up and possibly become ahead of schedule due the smaller headway it is now in. This route is so sensitive to this sort of thing that I’m not sure that anything short of true signal priority along the entire line will make much of a difference.

    Saturday and Sunday evening service: there simply aren’t enough vehicles to accommodate the amount of people wishing to use the service. The drivers aren’t “screwing around” unless you consider using the washroom or eating (god forbid) at the end of the line “screwing around”. When there are mobs of people at almost every stop there just isn’t enough time to load them all in the allotted time so buses become late and things start to fall apart quickly.

    Steve: To be clear, there are two types of activity that show up in the detailed data. First, the service on weekend evenings is demonstrably much more disorganized than at other times, and short turns with the relatively lower level of service create even bigger gaps. Second, there are some operators who love to travel in packs and do so from one end of the line to the other. This has nothing to do with the cumulative demand at stops which is definitely a “problem” in the sense that there is strong demand on the route. I will be looking at running times in the next article in the series.

    Road surface and signal priority: many of the prioritized intersections are no longer working and do not hold the light or reduce the cycle time to favor Dufferin St, correcting this would do wonders for service reliability. The road surface of Dufferin St is so poor that it actually affects the operation of the line, buses on Dufferin generally travel well below the posted limits in the really poor spots because its hazardous and uncomfortable for the passengers to travel at speed, smooth braking is also difficult due to these severe undulations. Right turns on to Bloor from southbound Dufferin need to be banned at all times, the heavy pedestrian traffic and sheer amount of buses trying to access the subway stop is a terrible mix leading to many dangerous situations and missed lights.

    Supervision: Anyone who knows the 29 will know when it’s being ignored. The much maligned 29S will be a rare bird and flocks of buses followed by large gaps will be the consequence. If you think the drivers are having a good time why don’t you ask them how they feel about it? Surely they’ll be overjoyed to be late, packed to the doors and dealing with hordes of unhappy passengers while they fight other buses for space at every stop along the route.

    Steve: This is an issue that the TTC completely ignored in its “Customer Charter” announcement — the degree to which inadequate service creates a poor working environment leading to unhappy staff and passengers.


  21. A few thoughts/observations/questions.

    1) Steve, have you ever formally invited anyone at the TTC to publicly respond, on your blog to analyses such as this? I think it would serve the TTC and public well, both in practice and as a public relations exercise if Mitch S. and/or Chris U. or Brad R. (or Andy B) went point for point w/you (or even posters/contributors).

    Steve: Anyone who wants to respond is free to do so. However, I get far more frank comments off the record when people don’t have to worry about stepping on toes of people in other departments. There have been many discussions both about how the TTC should measure its service quality and what might be done to improve things for riders. The problem, particularly in the past few years, has been that everyone is just trying to “keep the lights on” in the face of budget cuts. I do know that some of the techniques I have demonstrated here are being adopted internally, but it’s an uphill battle for an organization whose “culture” consisted of being “the best” for so many years.

    2) I think we have by now clearly covered that +3/-3 is not an acceptable way to measure on-time performance. Andy Byford has even publicly stated that he thinks vehicles leaving early is entirely unacceptable, particularly from terminals or for last service of the night. So can we get the TTC to agree to +1/-3 as a reasonable way to evaluate service? That would create a 4-minute band, instead of 6, and would not penalize drivers for real congestion (-3) but would for leaving a stop more than 1 minute early. Surely we could all agree that would be fair. It also wouldn’t solve the whole problem, but [it’s] not a bad place to start!

    Steve: The TTC has committed in its new Customer Service Charter to introduce a new way of measuring service quality this year, likely for the fall, and I know they have been working on it for some time. However, I don’t know what the targets will look like. There is a problem with the concept of “on time” because if everything is off-schedule, then any measurement relative to that schedule is meaningless. The question, then, is how to move to headway based management even if everything is late, and spacing out service could mean that the bus at the back of the pack will be even later.

    One possible way is to use express operation to let some of the buses pull forward and spread out the service. This can look bad to people who have been waiting in the gap only to see an express roar by, but it could be better overall. From a measurement perspective, there is a challenge of knowing that a bus is not really stopping everywhere and so should not be counted in the headway at that stop.

    3) Leaving early at terminals really does need to [be] tightly controlled. In the absence of subway style signal controls; how about making this a point where leaving more than 1 minute early w/o supervisor or transit control approval results in penalty, as-of-right (like running a red light). My object here is not to beat up on staff; but you can’t put an inspector at every terminal of every route, cost effectively. There needs to be some sort of ‘default’ on more egregious abuses to the schedule that a computer can monitor for; with room for approvals for exception by TC or by a line inspector/supervisor.

    Steve: As I said above, this protocol fails if all of the service is running late because vehicles would be dispatched as soon as they arrive (the same thing happens in the subway which has this sort of scheme already). What is needed is a pair of eyes looking at all of the service near the terminal and spacing service based on what’s in the pipeline.

    4) I’m being a bit fanciful here … and bus specs are far from my field; but in hearing discussion of issues w/the Toronto Rockets about the level they were at w/platforms, there was some discussion that this was affected by the way the train adjusted to more or less weight from passengers. Suggesting that the train actually responds in someway to varying weight levels.

    If this is so; then total passenger weight, is in some way, measurable by the equipment. That number could be understood as a rough approximation of volume of passengers on board. That being the case, could buses not be similarly outfitted, and could that data not inform decisions about route management.

    (There are, I am aware many other ways to measure ridership, but I was thinking if line managers and transit control could access real-time volume numbers, it might lead to more informed decisions on how to manage service). (ie. short turns, gap vehicles etc.)

    Steve: The train is not actually “weighing” the passengers, but is (or should be) attempting to keep pressure in the air suspension such that the distance from the car body (which is sprung) above the truck (which is not) stays constant. Also, I think that there is some latitude in the suspension system so that it is considered to be “on target” within a range of values rather than spending all of its time making fine adjustments in reaction to small changes in load. You have probably heard subway trains “sigh” when a large volume of passengers get off, and that’s from the suspension system letting air out because the car has become lighter.

    I suppose that this could be converted into some sort of measurement, but there would have to be some allowance for “bounce” to smooth out the values. For buses, there is the additional problem that lowering to meet the curb involves deliberately changing the suspension pressure. Possible? Maybe. Likely? Not much especially until the vehicle monitoring system is upgraded to have more data bandwidth so that it can report extra info like this.


  22. I have said this before, I think a lot of this could be dealt with by operators. If two vehicles are supposed to be 3 minutes apart, and the operator comes to a stop while the bus ahead is still there, they should be required to contact control, control can then get a supervisor to restore vehicle spacing. If you have a herd of two buses, and you don’t do anything about it, then you will eventually have a herd of 3,4,5,6,7 or 8 buses.

    Some routes are also broken, the real key when a route is not working, is send the Engineer around with the driver, during the busiest times, so that they can see why it’s not working, and then they can take their maps, and conversations with the driver to find a way to fix it.


  23. Hi Steve,

    Do articulated trolleys have advantages over articulated buses?

    We should not forget that re-introducing trolleys will required a new maintenance yard. Furthermore, a connected network of trolley routes will be desirable if they are used at all. That means both N-S and E-W lines.

    Steve: I am not actually advocating that we launch into a trolleybus network build, only saying that if Dufferin is going to be converted to something other than diesel bus, trolleybuses make more sense as this could be the spine of a northwest network.


  24. How far up would a hypothetical Dufferin Streetcar run? Dupont (in order to have room for a loop at the mall)? St. Clair? Even further?

    Cheers, Moaz

    Steve: There is no point in having a car stop at Dupont or St. Clair when there is so much through demand from further north. In any event, please see my reply to a previous comment in this thread.


  25. I used to live right at Springhurst and Dufferin (the end of the line) and saw buses leaving in bunches regularly. I’d wait for the bus right at the end of the loop, on a weekday morning during rush hour, and see bus after bus pull into the station, until no more could fit in the lot just south of Springhurst. The drivers would get out of their buses and chat, then suddenly they would all get into their buses and leave all at once. The end result is I’d have to wait 10-15 minutes to get on a bus. It’s amazing to see that the statistics show what’s happening.

    Considering how much money is spent on a busy line like Dufferin, you’d hope the TTC would start paying more attention to this kind of thing. If I’m correct, all the buses now have GPSs – why not simply have someone look at busy routes like Dufferin and make sure that operators are not deliberately leaving in packs? Anyone can already look at Nextbus and see where the buses are bunching up, it amazes me that after all the publicity the TTC still allows this behaviour to take place.

    Steve: I have changed the URL supplied by Joe because it did not point to a working Nextbus display (it actually brought up a map of Yonge 97). The link I have supplied points to the 29 Dufferin route. Note that Java is required to run this app.

    And, by the way, the TTC already has displays based on the GPS data.


  26. Yes, that’s not easy, but the first requirement is the willingness to do it.

    At the 2007 ‘fix the 501’ meeting addressing similar issues (bunching, cuts) a veteran TTC representative admitted that TTC culture was slow to change.

    I also think that with the subway system being such a priority to the TTC, bus & streetcar routes are overlooked. Traffic interference is a handy excuse for doing nothing.

    Steve: That’s about the time when “TTC culture” started to replace “congestion” as the stock answer to service problems when my analyses showed that most congestion was predictable.

    Another factor is that drivers (including many citizens, politicians, TTC & transportation officials and even a significant number of riders) look to existing & planned underground operation as a way to remove transit off the surface and out of the way of cars.

    I usually walk or roller blade after dropping off or picking up my bike from my mechanic a couple of kilometres away, but for some reason last week I decided to take the 29 from Queen & Dufferin around 11pm. I waited about 23 minutes with a crowd of people, a few of whom escaped into the two cabs that passed by. 6 buses passed south in a convoy. About 15 minutes later 7 buses came north (I waited to count when I got off near Bloor). No one complained verbally as this is what regularly passes for customer surface.

    Steve: If the TTC is serious about “customer service” and of getting “good value” for taxpayer dollars, they could do worse than examining this sort of problem and restoring evenly space service with the vehicles already on the street. If as some ops claim this is simply a problem of bad scheduling, then fix the schedule. However, I can’t help thinking this is more a question of long-standing tolerance of bad operation practices. At the very least, if there is a parade it should be broken up by short turns, not travel back and forth on the route.


  27. It’s currently 11:30pm and I just watched three buses leave the Dufferin Gate of the Ex grounds almost together. Two were coming southbound with the trailing one catching up at a great pace. They met somewhere within the Ex grounds and then came back and paused for some time at the Dufferin Gate (probably the loop). A third bus turned at the loop and it appears the first two left immediately after, having intentionally waited to follow it. A fourth bus turned at the loop and left without waiting before the 3-pack had even reached Bloor even though the spacing on the rest of the route would suggest this is not the scheduled headway. There was no northbound bus above Dupont at this time. The pack remained quite tightly together all the way to Wilson where all three went out of service.

    At midnight a southbound bus obviously driving hard and fast caught up to its leader by Bloor Street. It probably would have sooner were it not for stop lights. They met for a break at the Dufferin Gate. 42 minute service gap southbound between Wilson and College at this point. Trailer kept pace 3-5 minutes behind leader all the way back northbound. They left as a pair from Wilson Station.

    Also watching an eastbound and a westbound pair on 504 at 12:45am. There seem to be a large number of errors in the tracking downtown with a lot of cars disappearing and reappearing with alarming frequency. It was showing two cars westbound between University and Shaw which then jumped up to six and then down to four with three more approaching University. Have you noticed this type of error in the raw data stream or is this just NextBus screwing up?

    Steve: Re the Dufferin bus: This is what I mean when I complain about operator behaviour, and it is odd that this route has a particularly long and strong history of it. Other busy routes don’t behave this way. The TTC shows no sign of trying to do anything about it.

    Re disappearing cars on King: This looks like a Nextbus problem, although it may be related to GPS problems on a route with many tall buildings. I am not sure what Nextbus does when it gets either a rogue GPS reading (in the middle of the lake, say) or when a vehicle doesn’t report in for a few intervals. This really screws up arrival projections, needless to say.

    In my analyses, the process of data cleaning has three components: discard off-route data, smooth on-route data where small-scale GPS errors produce nonsensical readings (such as oscillations between two nearby points), and interpolate missing data between reported points. Only if a vehicle “disappears” for an extended period do I consider this to be an actual break (typically triggered by an off-route short turn or a diversion) and skip the interpolation. I am using the TTC’s raw data, the same info that Nextbus gets, but have the advantage that I don’t have to guess what is going on in real time and can look at nearby data points to sort out the strangeness.


  28. “Streetcars are better at carrying more people than buses”

    I don’t dispute this claim if the demand is predominantly local, but is this still true if a significant portion of the demand is more regional (long-haul)?

    Steve: This is one of those “it depends” questions. If we are discussing on street operation, it does not matter what the vehicle is, it will have to deal with traffic or will need provision for an exclusive, express lane to serve long-haul traffic. When this is on a regular street, the question then becomes whether the “express” component of demand warrants giving over that much space while leaving the “local” demand with whatever remains.

    If we are talking about private right-of-way operation, there are two possible configurations. In one, the service provides end-to-end capacity much like GO (although not necessarily as commuter rail), while in another, a shared right-of-way is used for a line-haul function by many routes with a common destination (e.g. the York U busway).

    There is also the question of what you mean by “regional”. Metrolinx has dubbed the former Transit City lines as “regional” even though they don’t come close to crossing municipal boundaries, especially after the scope cuts. They are “regional” only because this fig leaf allows the province to seize control of them. If the Eglinton line ever reaches the airport, it will have a regional component, but its demand will remain overwhelmingly local. Conversely the ARL is designed to serve only traffic to the airport, and even at that, only a small subset of total demand. Its ability to revert to a local service is dubious.

    There is also the question of demand level. Based on trip length, there is probably substantial demand on the subway where the trips could be called “regional”, but would you really want to try putting them on buses?

    In summary, a streetcar almost by definition is a local vehicle, but so would a bus be running on a comparable route.


  29. What is the benefit of streetcar vs trolleybuses? I’m sure I have never been on a trolleybus but I cant see how it would be much different then a regular bus.

    Steve: The question is one of acceleration under load on hills, something that challenges diesel buses on busy city streets, but particularly on a route like Dufferin.

    What happens in the future when Dufferin is redeveloped from Eglinton to Yorkdale? That sounds like its far far off in the future but starting at Lawrence now we are seeing a number of proposals. The Eglinton LRT will speed up the development on Dufferin there.

    Is it possible to run transit in mixed traffic part of the route, but in ROW another part? For instance south of Eglinton it could be run within mixed traffic and north of Eglinton it could be run in a ROW.

    Steve: The problem is that where it runs in mixed traffic, a route, whatever the technology, will have a challenge. Yes, demand may rise on this route, but it will also be intercepted by a new station at Eglinton and Dufferin. Before we commit to a technology change, especially to a jump to streetcars along the entire route, we should learn how the route will behave once there is a rapid transit connection at Eglinton.


  30. Ideally, there should be a separate service supervisor in charge of each individual route. If there is only one bus on that route, then that supervisor will be the backup driver, but only for that one route. A novice supervisor could be given the smaller route, to get needed experience. But the supervisors should concentrate on the best service possible on their route at any or all times of the day.

    In turn, the division manger should be checking on the service supervisors to make sure there are no such problems as reported here, and/or correct them.

    Steve: Whether we need 1:1 supervision or not, we do need an equivalent of a “station manager” with the responsibility for tracking groups of routes, identifying problems, and reporting on tactics and requirements to make improvements.


  31. Steve wrote:

    “They are “regional” only because this fig leaf allows the province to seize control of them.”

    One could also argue they are “regional” because it allows the province to fund them!

    Steve: The province has funded lots of transit, including 1/3 of the Spadina subway extension, without grabbing control of the projects. It’s done for accounting, but brings a lot of baggage as well.


  32. Michael Forest says:

    March 2, 2013 at 1:50 pm

    Hi Steve,

    “Do articulated trolleys have advantages over articulated buses?”

    They have several major advantages:

    1) They have much higher initial acceleration rates.
    2) They can go up hills faster than diesel buses.
    3) They can use regenerative braking which reduces brake wear and save energy.
    4) By using 6 small AC motors they can have one for each set of wheels which gives much better traction in snow and of slippery roads. This would stop the problem of pusher buses jack knifing. They may not need all wheels powered but it would be much easier to power the middle axle than on diesels.

    Most European articulated buses had the engine mounted under the floor and powered the axle in front of the articulation. This resulted in fewer incidents of jack knifing. GM introduced the rear drive articulated because it was their practice to mount the engine at the rear of the bus.

    It might be possible on a pusher articulated to run a long drive shaft to also power the middle axle but I think this would create a lot of maintenance problems. I do not know if any of the current North American trolley buses power the middle axle but it would be a lot easier to do than on a pusher diesel.


  33. @Robert Wightman: thank you for the response and information.

    So, it might be a good idea to covert Dufferin to trolleys.


  34. Michael Forest says:

    March 3, 2013 at 4:05 pm

    “@Robert Wightman: thank you for the response and information.

    “So, it might be a good idea to covert Dufferin to trolleys.”

    There are probably a number or hilly routes that would benefit from being trolley bus. As Steve said it would be nice to have group in one geographical area to allow for a north south and east west network that could be run from one garage.

    Steve: Of course, no sooner would we propose this, than the battery/hydrogen/wireless bus crowd would launch into an explanation of why their technology was better than trolley buses. Sometimes the alternate energy crowd spends more time shilling for their pet products than concentrating on what works and what the job at hand might be.

    Remember how we lost the TBs to natural gas buses in the first place? A confluence of interests at TTC (who wanted rid of them), the Ministry of Transportation (whose alternate technology boffins had nothing to show for what we paid them), the natural gas industry (who had a product bubble they wanted a market for), the manufacturers of the gas tanks, and finally Ontario Bus Industries (who got an untendered contract).

    Somewhere in there, a bus actually carried some riders, but it was a classic case of BS Ontario “industrial development” run amuk.


  35. @Steve

    I remember that the NG buses had such wonderful range that they could not stay in service all day but had to return for refuelling. This may not have been a problem for companies that changed buses when they changed drivers but that would have added a lot of dead head mileage and cost for the TTC.

    They also had the disadvantage that they could not operate into underground stations, required special fuelling apparatus, took longer to fuel, needed modifications to the garage to remove and leakage etc. All in all a wonderful design.

    We have so much to thank the Ontario Government for; let’s count the ways:

    1) SRT instead of LRT
    2) NG buses instead of trolley buses
    3) overweight CLRVs instead of modern updated lighter PCCs.
    4) UP express instead of all day service to Brampton with the help of the feds and SNC Lavalin

    To quote Monty Python, “Now for something completely different.” I was part of the phone poll that asked people about what they were willing to pay, if anything, for better transit. Given the way some of their questions they asked that the results were as good as they turned out. I don’t know squat about designing polling question so these may have been good but the seemed strange. The ones about cost gave a real low ball figure followed by the highest figure such as; “What would you be willing to pay in extra taxes to improve your Commute? Less than $5.00 per year. more than $100 dollars per year.”

    Later that night I was on a stake holders panel to help redesign Brampton’s strategic plan. I don’t know why I was on but it was very representative of Brampton, 4 white guys over 60, 1 white woman in her 50s, 1 south Asian man in his 50s 1 south Asian woman in her 40s and 1 male in his 30s who worked for Brampton. We all agreed that transit needed improving, both within the city and between area of the GTA, the city could not let industrial land be rezoned for housing, density needed to be increased but not with super sized condos and the OMB should be scrapped. All in all it was an interesting experience.


  36. The schedules for the 29 bus on the TTC web site naively show buses leaving a stop every 3, 4, 5 or 10 minutes like clockwork depending on the time of day or night. This seems rather quaint and disingenuous when NextBus shows gaps of 15 minutes or more somewhere along the line almost every time you look at NextBus for that route.

    Perhaps a disclaimer on accuracy is called for.

    Perhaps the page showing the “next 3 scheduled buses” should be replaced by NextBus data.


  37. Actually it is inexcusable that the pages on the. TTC website hasn’t been updated with Nextbus data, rather then still showing Nextbus predictions. It should be only a few hours work.

    On the other hand, the website was contracted out, so they probably quoted a ridiculous price to make that change.

    Steve: I think you mean “rather than still showing scheduled times”.

    The TTC has a big problem that they implement things piecemeal, and then find that integration is almost impossible. It’s a combination of bad specs and interdepartmental wrangling over what is important, I suspect.


  38. The bunched *departures* of buses and streetcars from base are astounding. I’ve never heard of this in any system other than Toronto. A cultural problem? Certainly. “Must be stopped” is a phrase which comes to mind. What is going on in the TTC? On frequent routes, headway maintenance is pretty much the first thing riders care about, and it’s being flagrantly ignored. This is just *odd*.


Comments are closed.