Updated May 26, 2011:
- The three sets of charts for each day and location previously published here have been replaced with a single set to simplify navigation.
- Calculations of the average and standard deviation was previously based on headways that had been rounded to the nearest minute. This has been changed to use the unrounded values causing minor changes from the original version of the charts.
- Data for December 29, 2006 was incomplete causing some headway information to be incorrect. This day has been dropped from the “basic statistics” charts.
Effective with the April schedules, the TTC changed the 7 Bathurst to formally be an articulated bus route on weekdays. Headways were changed to reflect the larger capacity of the vehicles, and the TTC trotted out a commonly-cited story that fewer, larger vehicles are easier to manage and can provide better service than more, smaller ones.
Did this actually happen? What does the implementation on Bathurst bode for other major bus routes, not to mention the streetcar system which will start moving to larger vehicles in fall 2014?
In these articles, I will look first at the headways actually operated on the route for March and April 2014 (the before/after pair), and later at the time required for vehicles to make their trips, congestion and layover times that affect the service.
As it happens, I also have data from December 2006 for this route, one of the first sets of data I attempted to analyze when I started on this effort back in 2007. How has the service changed between late 2006 and early 2014?
Note: I presume that readers have seen this type of article on my site before, and I will not go into the gory details of how the TTC vehicle tracking data are digested and presented. For those who want to know what’s “under the covers”, please refer to a separate article on the methodology.
What is the Scheduled Service?
Before we can talk about service as it is, we need to know what it is supposed to be according to the schedules. This is a starting point for comparing the reality of buses on the street to the creative fiction of the advertised service.
Charts of the Headway Data
Charts are linked below for four locations in each of three time periods.
- Northbound service is shown at Barton Street (just north of Bathurst Station) and north of Wilson.
- Southbound service is shown at the bottom end of the on-street loop at Steeles Avenue, (specifically at Carpenter and Bathurst where the loop begins), and just north of Eglinton.
- In both cases, the locations were chosen to show the service as it left the terminal and as it arrived midway along the route.
- Service is shown for the months of March 2014 (before artics), April 2014 (after artics) and December 2006 (for historical comparison).
7_201403_NB_BartonAve_MonthHeadways 7_201403_NB_NofWilsonAve_MonthHeadways 7_201403_SB_NofEglintonAve_MonthHeadways 7_201403_SB_SofSteeles_MonthHeadways
7_201404_NB_BartonAve_MonthHeadways 7_201404_NB_NofWilsonAve_MonthHeadways 7_201404_SB_NofEglintonAve_MonthHeadways 7_201404_SB_SofSteeles_MonthHeadways
7_200612_NB_BartonAve_MonthHeadways 7_200612_NB_NofWilsonAve_MonthHeadways 7_200612_SB_NofEglintonAve_MonthHeadways 7_200612_SB_SofSteeles_MonthHeadways
These charts contain 13 (or 14 depending on the calendar) pages of charts:
- The first 7 (or 8) show detailed headway information (described below).
- The next 3 contain basic statistics: averages and standard deviations by hour for each type of day in the month (weekday, Saturday, Sunday).
- The last 3 contain distribution charts for the data showing how closely (or nor) the values are clustered.
Basic Statistics
The basic statistics charts show information at the most consolidated level. For each type of day (weekday, Saturday, Sunday), the average headway and the standard deviation of the headway are plotted. The values included for each time period combine data for every day of the specific type through the month.
- The average is the arithmetic mean of all observed headways within an hour for the type of day charted.
- The standard deviation is a measure of the dispersion of values with about 2/3 of the data points falling in a span of ±1 standard deviation.
Where a month’s operation is “well behaved” with all of the days operating more or less the same way, then these values will be representative for every day. If there are many days with disrupted service (for example due to weather), the values will be affected because they are based on a mixture of “good” and “bad” days, but it will not be possible to determine when and how severe the problems were on a specific day.
The standard deviation values are also related to the problem of bunching. For example, if the SD is 4, then 2/3 of the headways lie in an 8-minute band around the average. During many periods, the SD is close to the average value, and this indicates that most vehicles are travelling in packs of at least two. This is particularly troubling when it appears shortly after vehicles leave their terminals where, in theory, some headway reliability could be more easily imposed.
Headway Distributions
These charts are, like the basic statistics charts above, based on data consolidated for all days of each type through the month and segmented by each hour of service. Within each hour’s data, the count of vehicles on a specific headway is plotted so that there is one vertical bar for each value from 0 to 30 minutes. The intent is not to show the individual values, but the shape of the data distribution.
If most headways lie in a narrow range, then the cluster for one hour will have a tall spike with much lower outlying values. If the headways are spread over a wider range, the cluster will be flatter. In the extreme case where headways are completely random, the cluster would have no peak at all.
Note that because there are fewer days and less frequent service included in the weekend data, the charts are more sparse and have lower peak values. However, the principle is identical.
Detailed Headway Charts
Each linked file contains 7 or 8 pages depending on the calendar:
- The first 4 or 5 pages show data for weeks 1-4 (or 1-5) of the month.
- One page shows all weekday data as a “cloud” of data points. The purpose is to illustrate both the general variation in values over the course of weekdays, and the very wide spread in actual values.
- One page shows all Saturdays (Boxing day is included here).
- One page shows all Sundays (Christmas and Good Friday are included here).
Each point represents the actual headway when a bus passed at the time in question. Trend lines between the points show the overall behaviour of the values through the day. The spread of points around the trend line shows the band within which the actual headways lie.
These charts show data for each day of the month independently, and allow identification of specific times where a major event such as a snow storm might have disrupted service (as opposed to monthly averages presented in the earlier charts).
Northbound at Barton Avenue
These charts show headways just after the service leaves Bathurst Station. The March trend lines show the shorter peak headways on each weekday except for the March 31 data. This was the day the new, artic-based schedules actually began to operate and so the headway trend for that day does not match the pattern from the preceding week.
One thing that is quite evident throughout March is the consistency of the trend lines from day to day, and on each type of weekend day. The only exception is that midday headways on March 12 are unusually long due to a heavy snowfall.
The other quite evident point is the width of the “cloud” of data points around the trend lines. The TTC’s target is ±3 minutes to the scheduled headway, but the data lie in a band almost 20 minutes wide. Very short headways correspond to vehicles running in packs, and this is troublesome to see so close to a terminal where it would be easy to manage regularly-spaced departures. This problem exists even on weekends, and some headways are twice the scheduled values.
In April 2014, there is no AM peak dip in headways because none is scheduled (see table above). Headways continue to have values well beyond the three-minute target range, although the problem becomes noticeably worse after about 3:00pm on weekdays with considerably more vehicles on very short (i.e. bunched) headways right through the PM peak into the evening. Weekend data are quite similar for March and April which operated on almost identical schedules.
Comparing the two months’ weekday operations, both have a lot of bunching, especially in the afternoon and evenings, and the primary difference is that April headways are longer. There is no indication that the larger, less frequent vehicles in April are managed any better for headway reliability than their shorter cousins were in March.
Data for December 2006 show values that are less spread out than in 2014, although headways well below and above the target are still evident.
Northbound North of Wilson
As with the service at Barton, the March 2014 headways at Wilson are spread out over a 20-minute wide band with many quite low values for all days of the week. The situation continues into April with the artic-based schedules. Indeed, in April there does not even appear to be a concentration of values near the target, but rather a fairly even range of values. This suggests that there is no attempt at all to manage the service or impose any headway discipline.
By contrast, the 2006 data generally lie within a narrower band, although there are still cases of very short headways and some in excess of two times the scheduled value.
Southbound South of Steeles
Service southbound from the northern terminal is even more irregular than northbound from Bathurst Station with the band of values extending into the 20-30 minute range. One reason for this is unscheduled short turns at points south of Steeles.
It is worth noting that service in 2006 was much better behaved except during the PM peaks and on Saturdays.
Southbound North of Eglinton
At Eglinton southbound, the data values are back to the 20-minute wide band we saw in the northbound charts. The contrast with the values at Steeles are due to the fact that northbound Short Turns are included and some of the very wide gaps were filled. However, this is little comfort to someone forced to wait two-to-three times the scheduled headway on the outer part of the route for a bus to arrive.
Service in April shows the same pattern we saw elsewhere with little evidence of clustering around the target headway on weekdays.
The 2006 data are much better behaved with a tighter range of values except on a few occusions, notably just before Christmas.
General Observations
The Bathurst Bus was notorious for irregular service before it was converted to articulated buses, and the pattern has not changed since the cutover. Headways range well beyond the TTC’s target for acceptable service, and passengers can just as easily be treated to a double headway as to a short headway thanks to bunching and the resultant uneven loading conditions.
There are noticeable differences between 2006 and 2014 data in that headway distributions (and standard deviation values) in 2014 are wider/larger than in 2006. This means that service experienced by riders will be more irregular today than in was 8 years ago.
Another point notable in all data is that the standard deviation does not often fall below 4 suggesting that there is a lower bound to the reliability of headways. Whether this is a question of the accepted practice for operations or some “transit fact of life” is worth investigation. However, a striking difference between 2006 and 2014 is that the standard deviations tend to stay at the same level most of the time in 2006, while they grow considerably in off-peak periods in 2014.
Any claim the TTC might have that the larger buses would provide more reliable service are demonstrably false. There may be a theory, somewhere, that might “prove” this assertion, but the service actually recorded on the street reveals a complete abdication of headway management, the alleged goal of TTC service.
In the next article, I will review the relationship between actual and scheduled running times, the amount of time TTC Service Planning allocates for buses to make their trips, as well as other factors that can affect service.
Steve Munro 
10 minute bus frequencies on all routes during off peak is necessary.
Steve: Ah yes, but does that mean 10 minutes over the entire line, or just in the central section thanks to lots of short turns? Does it mean pairs of buses every 20 minutes? Imagine if the subway ran like the Bathurst bus.
LikeLike
Looks just like the Dufferin line. The TTC would be better off replacing one of the drivers with a “headway manager” that has the authority to prevent bunching. Hate to say it, but this all the result of drivers wanting to socialize together at the end of the lines during breaks. I often see drivers hanging out at a station together, then all leaving as a caravan so they can meet up again at the other end of the line.
While the TTC needs more funding, this problem is just the result of mismanagement and could be fixed without using additional resources. It makes the TTC look bad when they are given more buses, only to waste these resources by letting their operators leave/arrive any time they want.
Steve: I will turn my attention to 29 Dufferin in a later series of articles. The behaviour you cite is particularly noticeable on that line where vehicles congregate at Dufferin Loop and Wilson Station.
LikeLike
We probably need a new word to describe ‘actual frequencies’ (i.e. a bus appearing at a stop) and clearly differentiate this with the “theoretical frequencies” (i.e. a bus appearing on the timetable). Ideally these two times/frequencies would be the same but casual observation and these complex route analyses show they differ wildly. I await with some interest the promised analysis of the 504 streetcar where the customer is also faced with a large number of planned and unplanned short-turns.
Steve: Please be patient. I do not want to give suburban readers the impression that I only care about downtown routes (even if the King car is one of my local services). It’s actually important to look at the bus routes to prove that bad service is not just a function of streetcars running on narrow, congested downtown streets with spillover from every imaginable type of construction and special event, but a pervasive problem on the transit system.
LikeLike
Somebody really should ask the mayoral candidates what they think of all this. I know we’ll just get stock answers, but we might finally get some discussion beyond “Andy knows best”.
LikeLike
I was on Bathurst briefly when both vehicle lengths where in use and happened upon a supervisor who told me some operators were driving their artics more carefully – and slowly – than regular buses as they got used to the new dimensions. Bathurst has rather tight lanes in many sections.
I can’t naysay, haven’t driven a TTC bus but once for a media day, hitting a good number of orange plastic barrels …
Steve: In the second part of this article, I will look at running times, and this will reveal whether there is a noticeable change between March and April. I will break out the data by vehicle type to isolate the effect of artics running in March and, yes, 40-footers running on artic headways in April.
LikeLike
Steve near the terminus/stations going away from it, 3 minutes should be enough to be 2 standard deviations. The only situation that should be allowed to occur and that would be an indication of insufficient recovery time would be a bus being released late, as it arrived too late to make its departure headway. Otherwise, shorter than scheduled headway should represent a bus sent out to relieve crowding only.
Drivers need to know the gap in front of them, and need to care that it be maintained. I am not under the impression that either of these is currently true let alone both.
Without effective line management, one would intuitively expect going to larger vehicles to make the situation worse not better. It makes one really wonder what notion the TTC has found that would make it think the contrary.
Steve: As part of the second installment in this series, I will be looking at terminal operations including layover times.
LikeLike
In any industry, I’ve rarely seen such damning performance data. Never mind that the std. dev. is almost always more than 50% of the average, the number of charts on which the two are even closer speaks of serious deficiencies in designing/publishing route schedules and/or in managing problems when they arise on the line. The std. devs. should be WAY smaller if there were serious attempts to maintain schedule and/or headway.
The evenness of the dotted plots clearly shows that the problems occur almost all day, with very little change at times when there are few reasons to think that the headway could not be consistently close to the published value. There’s just about no sense at any time that even some form of ‘narrow-banding’ of the dot positions (i.e. closer adherence to schedule or headway) is occurring. One might have hoped that at least some part of the day would demonstrate better performance. Perhaps the best I can say is that performance seems ‘consistent,’ if bad.
Someone(s) at the TTC has(ve) some ‘splainin’ to do.
Just about says it all, really. 😦
On a separate note, had I been a regular user of a bus route and heard that the published headway was being increased by about 50%, especially at peak times, I’d be looking for another way to travel to work. Taking The Car, perhaps.
I would like to hope that these data are a wake-up call to upper management of the TTC, but, given history, I suspect that the snooze button will be pressed.
LikeLike
Surprise, surprise, surprise! I am not. The 7 Bathurst was a bad route to begin with. Simply switching vehicles won’t solve a problem. Well, the TTC can safely claim there is less short turns cause there is less vehicles for them to short turn. Mission succeeded!
Honestly I think they should get rid of artics and bring the 40ft. buses back on Bathurst. It would be interesting to see if the dwelling time increased. Less buses lead to wider headways. That leads to stopping more often. On top of that, NovaBuses have slower door operations. It takes a second or two from touching the push bar till the door opens. The Orions aren’t like this. Artics are probably better on routes like Steeles East Express.
LikeLike
Is BRT acceptable for Bathurst?
Steve: No, the street is too narrow to support express lanes for buses.
LikeLike
Have they gotten any approval to go ahead with an automated in-vehicle headway management system? One would think that the new artics and the new streetcars will need this system if they are decreasing headways – otherwise there will be on-street issues … the belief that having less vehicles will inherently improve the headway situation without any actual changes is only true in that now the guy with the clipboard only has to keep track of 1790 buses instead of 1800.
Steve: The capital project for a new vehicle monitoring system has been approved, but at this point all that is happening is a requirements analysis. It will be telling what is and is not included in the specification because this will indicate what the TTC thinks is important to have.
LikeLike
The data doesn’t surprise me. As long as there are only 3 minutes of scheduled recovery time on #7 in the PM peak (say 1 minute northbound and 2 minutes southbound or vice versa) buses are going to leave the terminus late. There would probably be better service if the 16 scheduled PM peak buses (124 min cycle time / 7.75 headway = 16 buses) had 9 more minutes of recovery time added so that the headway would be (133 min cycle time / 16 buses = approximately an 8 minute 20 seconds average headway. There really is too much volatility in traffic conditions to expect good headway adherence on a surface route without segregated right of way with less than 10% scheduled recovery time.
Also, you would normally expect that articulated buses would be up to 10% slower than regular buses due to decreased acceleration, increased stopping time, and increased boarding time due to more passengers boarding at each stop. According to this article,it seems as though no adjustment was made to the running time.
Steve: Actually, that “recovery time” is not systematically included in the schedules. It is simply a rounding value to get the headway to come out to a fairly even value based on the round trip time. When I look at running times, I will see the degree to which these match scheduled times, and also what the typical behaviour of vehicles at terminals might be.
I also plan to look at comparative running times by vehicle type.
LikeLike
All those new layovers … isn’t that adding a challenge? Without effective line management how do you stop drivers from padding their layovers (driving too fast or leaving too late/leaving in bunches).
Cheers, Moaz
LikeLike
The main problem with the Bathurst bus (and many other routes) is the woefully inadequate running time at most times of the day. The ops need to be at their relief point at the correct, scheduled times and the best way to do that is by short-turning, which results in erratic service, from a passenger perspective. ‘Across-the-street’ change-overs, similar to what is being done every day on the subway, have been tried, but they are not as effective, and they can also be quite hazardous, depending on the location.
As far as running times are concerned, if you do have the time, I would suggest a side-to-side comparison of the data for the February 2014 board (solo vehicles), March 2014 board (artics) and May 2014 board (new schedule with more buses and additional running time at most times of the week).
I would also humbly suggest an analysis of the 196 York U Rocket (preferably for the Fall-Winter months when demand is highest) just so that everyone can see how reliable (or not) quasi-BRT can be in a Toronto context.
Steve: I have already been looking at running times in preparation for the second section of this article, and yes there are times when they are inadequate. However, this is not true all of the time and the service is still erratic suggesting a lack of attention to regular spacing of vehicles. When the target is plus or minus 3 minutes, this means that for scheduled headways up to 6 minutes it is perfectly acceptable to have buses running in pairs. There is something fundamentally wrong with that, and it has been the case for many, many years. There is an important but subtle difference between running up to three minutes off schedule, and running three minutes off headway. In the former case, gaps of 12 minutes on a 6 minute headway are ok (one bus 3 minutes late, one 3 minute early), and the change in target to a headway basis came only recently and is not, I believe, observed by anyone.
LikeLike
I know from experience that route supervisors on the line don’t always work. St Clair has had them for a while now, and bunching is still the same, short-turn one and wait 10-15 minutes for the next three.
I have said it before, the ENTIRE planning department needs to get out of the office and see how the real world works and not sit at a computer and hope for the best.
LikeLike
Service has to include adherence to headway. I disagree that run times should be that material to headway management, or at least when it is they should all be a little long or a little short. Traffic flow does not explain 2 buses following each other out of the station.
Traffic does build making for longer running times, however, all buses at this juncture should be running late, and the gap should both build and shrink slightly and predictably. The observed large standard deviation all day, is more of a ‘dispatch’ issue. If a vehicle is behind in its route, but less so than the vehicle in front it should be waiting in station until the gap is close to stated headway. I really do hope that the vehicle tracking system will include both headway tracking and reporting of such back to the driver. Heck a red light green light system for drivers in station would be huge.
LikeLike
Of course that is false Steve. Large buses are great for routes with frequent service where it is impractical to add more buses so that there is more capacity. However, unless the larger buses are replacing the shorter buses on a 1:1 ratio, service level is going to decrease. And like I said, you need frequent service first – which I consider to be better than 10 minute service. Otherwise, the bus is simply replacing another bus and not actually doing something to fix the problem.
And route management is something the TTC seems to have an issue with anyway.
LikeLike
Except for late evenings, I notice a 10% increase in round trip times since 2006. Any explanation for this?
Steve: Probably greater traffic congestion, but until I delve into that in the next article and look at comparative speeds over parts of the route, it’s too soon to say.
LikeLike
Slower than the deathly slow Orion 7s (especially the first gen)? I didn’t think it would be possible to take longer than them. (I have not had opportunity to ride the artics yet to see.) It seems that every new generation of vehicle results in longer and longer dwell times (which doesn’t bode well for the new streetcars!).
I was surprised to see that the artics were not pressed into service on the east Danforth subway shuttle (Woodbine to Kennedy) a couple of weeks back, or along Woodbine on Victoria Day — both cases where the higher passenger demand and lower turnover would have been well suited to artics. I assume this is because artics have only been cleared for use on certain routes where they have been tested out of service?
Steve: Also, I suspect it’s a case of which garage they are assigned to and where there are operators trained to drive them.
LikeLike
Actually there was an articulated bus on the B-D shuttle east of Woodbine. Bus #9049 if I remember correctly.
LikeLike
Darren, if route supervisors on the line are still making a mess of the St. Clair car, what will be the advantage of putting the planners out on the route? Unless the planners are making things so tough for the supervisors that the only option the supervisors have is to run streetcars in bunches of three to cope?
LikeLike
The problem with making this a planning issue, is that the simplest of measures would resolve something like this. On St Clair, you could pick a corner, step out and stop the driver until it had been 3 minutes since the last driver went by, and you would have a 3 minute headway.
LikeLike
Steve, I look forward to reading about how your analysis of the Bathurst bendy-bus might apply to Dufferin. My personal feeling is that bunching on Dufferin is no better now than it was before the introduction of articulated buses, though the construction around Bloor and south of Dundas would have affected this as well.
LikeLike
Of course there’s the human element to this; new operators vs experienced operators, as well as operators that come to work vs operators that come to play games.
LikeLike
I remember when I was a child and often saw on street inspectors controlling the headways. Now I go years without seeing any body in charge on the ground and note that the TTC can’t stop their vehicles from bunching up. There is a very serious and fundamental lack of understanding at the TTC as to what transit is.
LikeLike
It has been previously suggested that somebody pull a driver, and have him act as a timer. I think this should be tried, to see how much impact it had.
If it works, then just for fun, I think that the TTC should try and add a basic light at the station stop. Add RFID tags to the buses that identify route and specific bus, readers and a PLC controlled light just for the bus. Light in station posted for the bus route would stay red for say 8 minutes, and then remain green, until the next bus pulled out of station.
LikeLike
I think that this would run afoul of the division between management and union. No union member, with a lot of justification, would report another union member for leaving early or late. It does not take a lot of of brain power to install a “Leave Now Light” with a camera to see that it is done at each loop. I saw a lot of this in Europe.
Steve: I have seen dispatch signals in North America in the 60s, and the equipment looked as if it dated from the 20s. This is not rocket science, and does not require billions in new technology, only the will to do something about service quality.
LikeLike
I think RFID tags and scanners would be easier. It act to both detect when a car had left, and which car it was as well as permit automatic logging. Done now for instance to track returnable packaging in the auto plants. Cheap, quick and easy, as well as creating trackable log. I was thinking however just having a person do it for a couple of days would “prove” its usefulness. Basic tracking hardware is a couple of thousand RFID tags are literally a dollar each.
Basic plc and PC for light would provide tracking of headway a running time. Frankly even if the system did not already exist it would be dirt cheap to do, and provide car specific data. Look to see if one exists if not its a good college project.
Steve: Tracking data are already available for every vehicle in the system — that’s what I use to provide my analyses, and that’s what drives NextBus and other apps. We don’t need yet another overlay of technology, merely the will to use what we have to manage the system properly.
LikeLike
Couple of questions
1. Is the data currently available vehicle specific tracking data (as opposed to route only)?
Steve: Every vehicle on every route.
2. How can this not be a mainstream media issue? Can you imagine the outcry if Wednesday garbage pick-up was anytime from Tuesday am to Thursday PM?
Steve: Quality of service has been an issue for transit riders for as long as I can remember. It gains the media’s attention only fitfully, in part because the standard replies from the TTC have to do with issues allegedly beyond their control. The technical stuff is a big yawn.
This is a basic issue that is at the core of the service that a larger municipality provides. Improvements in this basic service is much more important to congestion than any multi-billion dollar project.
The TTC should not be seen as something beyond the city’s control, and as a consequence this should be the hottest of hot political issues. This is where basic accountability to the public is seen. The issue of transit locally for mayor and council should be about getting the ‘buses to run on time’.
Steve: But most of the candidates only want to promise a subway that will arrive in ten years, not a bus or streetcar that will arrive in ten minutes.
LikeLike
If they can traceably identify individual vehicles, then this should also be able to be used as a tool of performance evaluation of the drivers themselves, with any timing control in place. This makes it doubly painful that they have not addressed this most basic of issues.
Unless the balance of the drive or parking is brutal, I will not ride the bus if I know I will likely face a 15 minute wait, whereas I will knowing it will be 7 or 8. To me I chalk this up to voters not informing themselves. This is exactly the type of thing that Rob Ford was screaming about in terms of city government (and totally failed to address), and in one of the areas of city services with the largest direct impact on users. Councillors on these routes could place considerable pressure on the TTC and then trumpet your numbers both to the TTC and their voters as either the issue, or the fact that they fixed it.
LikeLike
Steve – I mention this for what it’s worth, but the new artics have made their way to Scarborough and a North York, and have been spotted on the 53E.
LikeLike
At least there’s an elevator at Bathurst station.
LikeLike