Analysis of Route 29 Dufferin — Part II: Headways

In the previous post, I began my analysis of the 29 Dufferin route with a look at service on Christmas Day 2006.  Before turning to other specific days and their events, let’s look at the month overall as seen by the reliability of headways and link times at and between various points on the route.  This post presents the headway data, and in the next installment, you will see the link times.

The picture revealed by these data is not a happy one, although it will not surprise any regular user of the route.  Headways are a mess, especially in the evening.  The oft-cited “flexibility” of buses does not appear to yield service any more reliable than on the King car, and in some cases, the service is worse.  The fundamental problem is that very frequent services are left more or less to their own devices, and less frequent periods on such routes suffer from the effects of laissez-faire management.

Of particular note is the service on Sunday evenings, a period when classic TTC excuses about “traffic congestion” simply are not credible.  Headways are scattered over a range up to 20 minutes even though the schedule says 10. 

A brief tutorial for those who are coming to this topic fresh:

Headways are the intervals between vehicles and, in an ideal world, they are actually almost equal to the advertised scheduled frequency of service.  In practice, there is quite a bit of variation.  The TTC considers anything within three minutes of schedule as being “on time”, but there are two basic problems. 

First, this seemingly small variation is huge compared to the scheduled service on many routes.  The gaps it allows make for disjointed service and uneven loading on buses.  Second, the actual service falls outside of this band quite regularly making a mockery of “on time” performance.

If headways are scattered over a wide band of values, the service seen by riders is erratic and unreliable.

Time Points, in this discussion, are locations where a vehicle’s passage is recorded.  These have nothing to do with the TTC’s own schedule time points, nor with “signposts” on the vehicle monitoring system (CIS).  They are chosen to give a representative view of a line’s operation in my analysis, and because the CIS data at these locations is fairly reliable.

Link Times are the travel times between two Time Points.  If these are consistent and clustered around one value, this indicates that travel over that portion of the route for the day(s) and time(s) in question is predictable.  If these stay well-clustered, but vary on a larger scale up and down through the day, this shows the effect of traffic congestion rising and falling.  If the values are scattered, this shows that travel through this area is fraught with unpredictability, and any operational or traffic control countermeasure to restore regularity would benefit the route.  I will cover this topic in the next post in the series.

Northbound from Dufferin Loop to Wilson Station

29 Dufferin Northbound Headways at King
29 Dufferin Northbound Headways at Dundas
29 Dufferin Northbound Headways at Dupont
29 Dufferin Northbound Headways at Eglinton
29 Dufferin Northbound Headways at Lawrence

Each set of headway charts contains five pages.  The first four show weekdays for the four weeks through the month and include trend lines for each day.  The fifth shows all weekday data on one page.  The sixth and seventh show Saturdays and Sundays.

From the first four pages of the data for King Street northbound, we can see that the trend line of the headway is fairly consistent through the day, but that there is a fair amount of scatter in the data.  Some days are worse than others as we have seen on other routes.  The spread in values gets worse the closer we come to Christmas, and things settle down again for Christmas week.

Page 5, with a month of weekdays, shows that the headways lie in a thick cloud between zero and eight minutes, with a considerable number of points above ten minutes.

Page 6, with the Saturday data, shows a more compact set of headways although things get more scattered as the day goes on.

Page 7, with the Sunday data, is fascinating because it shows that evening headways are all over the map from to 20 minutes at a time when there is no reasonable explanation for this behaviour.

At both Dundas and Dupont Streets, things have not changed much, although service at Dupont is not quite as spread out on occasion possibly due to short turns southbound near Bloor.  Sunday evenings continue to have erratic service.

By the time we reach Eglinton, the headways are starting to spread out showing that little is being done to maintain a reliable frequency.  Between Eglinton and Lawrence, there is a scheduled short turn in the peak period, but even in the off-peak, service at Lawrence shows the cumulative effect of bunching and gapping as vehicles travel north.  Week three shows the extremely striking effect of pre-Christmas shopping congestion and the early rush hour on December 22 with streets attempting to handle peak traffic flows while off-peak parking restrictions limited their capacity.

Southbound from Wilson Station to Dufferin Loop

29 Dufferin Southbound Headways at Wilson and Dufferin
29 Dufferin Southbound Headways South of Lawrence
29 Dufferin Southbound Headways at Eglinton
29 Dufferin Southbound Headways at Dupont

The view southbound at Dufferin and Wilson is not unlike what we saw northbound, except that the afternoon peak suffers badly from gaps.  Between 1600 and 1900, gaps of over 20 minutes are seen on some days, and occasionally gaps of 30 minutes especially on December 22.

Saturday service shows the impact of congestion around Yorkdale Plaza especially on Boxing Day (included with Saturdays to which it is most similar) in the extended headways.

Sunday headways are better clustered around the trend lines, but there are still outliers above 10 minutes, and evening service, as with northbound operations, shows headways varying from 5 to 20 minutes.

At Lawrence, the weekday data are badly scattered with gaps between 10 and 20 minutes quite common, and some gaps above 20 minutes especially on pre-Christmas shopping days. 

Weekend data from here south mirror what we saw at Wilson.

By the time we reach Eglinton, any short turns at Tycos Road (29A) have blended back into the route, and the gaps are not quite as atrocious as we saw at Lawrence and Wilson.  Like the northbound service near the south end of the route, headways lie in a band from 0 to 10 minutes that gradually spreads out as the day goes on.

At Dupont, we see the same pattern that appeared with the northbound trips.  The further south we go (that is, the closer we get to the terminus), the more scattered the headways become.  The band holding most headway values on weekdays at Dupont has spread over the 10 minute line with several values ranging up to 15 minutes.

The patterns continue south to Dufferin Loop.

In the next post, I will look at the Link Times to see where the buses are spending their time.

5 thoughts on “Analysis of Route 29 Dufferin — Part II: Headways

  1. I wonder if there’s a statistical measure that can show how erratic the service really is. Yes, it’s possible to eyeball the cloud (although with multiple days it gets pretty confusing), and you have done the jaggedy headway graphs bouncing all over the place. But neither of those gives a statistical measure.

    What I’m thinking is something like a hourly moving headway standard deviation plus and minus. So for every time, show the standard deviation of headways in the prevoius and following thirty minutes.

    What may be easier to implement (and possibly more useful) would be to calculate the standard deviation of the headways of the previous two and following two runs (four headway intervals in total: -2 to -1, -1 to run, run to +1, +1 to +2). It may also be useful to normalize the standard deviation by either the observed average headway, or by the scheduled headway.

    Steve: This would be possible, but there’s a limit to how esoteric I want to make the displays. So far, I have been trying to present information in a way that is fairly easy to understand. Bouncing headways, clouds of data points on headway and link time charts, zig-zag plots of operations — all of these relate directly to someone’s actual experience of the line. Standard deviations are one step away, and I think that the dispersal of the clouds gives a pretty good idea of what is going on.

    Hmm, a graph of actual headways versus scheduled headways (or some other comparison between the theoretical TTC world and the actual results) might be interesting.

    Steve: One problem here is that the headways change on different routes and locations within routes at different times. A tedious job to pull together the underlying data. Also, routes with “frequent service” don’t publish detailed schedules.

    Of course, the TTC should be doing this themselves! It’s not like a good statistical analysis requires anyting beyond intro college-level stats.

    I hope that the reason they’re not doing this (or at least not letting anyone know they’re doing it) is NOT because the results are too embarrasing and must be buried.

    Steve: The problem as I understand it is purely bureaucratic. The people who would love to have the analysis (Planning) are not in the department that would produce it (IT), and the department actually running the service (Ops) has until recently claimed they were doing a fine job, thank you.


  2. Steve:

    The volume and depth of your analysis continues to astound me! Your initiatives and interests far exceed the commitments of most of the over 10,000 TTC employees — from yardmen to senior staff.

    I must also compliment your literacy skills. You write with clarity and brevity — the hallmarks of successful journalists.

    There are lots of “hot button” topics coming forward in 2008. Have a Happy and Successful New Year.


  3. The headways are a HUGE mess. I use that route somewhat regularly, and there are instance where the bus I am on will layover, and allow 2-3 buses to overtake it. It’s really frustrating. I noticed that buses do come to Dufferin fairly frequently, but the buses seem to leapfrog each other a lot on the trip north.


  4. Barry McDermott wrote: “Your initiatives and interests far exceed the commitments of most of the over 10,000 TTC employees — from yardmen to senior staff.”.

    As a TTC bus operator, I find such a rash generalization of TTC employees to be extremely insulting and totally inaccurate. Is Barry stating that he actually believes that anywhere from 5000 (50%) to 10,000 (100%) of TTC employees are not committed to providing the best service possible? I would invite Barry to apply for an operating position with the TTC and personally experience our job first hand.

    I personally (as do 99% of the operators that I know) take pride in my work. I have received numerous commendations for excellent customer service. I have shown up for work on time and operated my bus on both of Toronto’s recent snow storm days.

    On the first storm day (a Sunday), I ended up being the only bus in service on my route due to the other buses becoming stranded at the bottom of an icey hill that had not beeen salted. I suggested a diversion that would minimize missing stops to CIS and would allow me to continue in service.

    I am aware that there are some operators who engage in “soaking” and playing for short turns (running so far behind schedule to force a short turn), but in my experience these operators are in the minority. The biggest problem, in my opinion, is route mismanagement. This is caused by CIS ‘s focus on schedule adherance rather than headway adherance.

    The unfortunate root cause of this is that senior management wishes to see vehicles run to schedule rather than have a vehicle show up on a headway based performance. If vehicles enter service late due to lack of operators or lack of vehicles, it tends to cause headway problems.

    As well, I have experienced vehicles parked at stations due to lack of relief operators to take over. If vehicles are missing from a route, this causes tremendous gaps in service that compound as time goes on.

    Steve, you wrote: “The problem as I understand it is purely bureaucratic. The people who would love to have the analysis (Planning) are not in the department that would produce it (IT), and the department actually running the service (Ops) has until recently claimed they were doing a fine job, thank you.”.

    I think that you have hit the nail directly on the head. The TTC is very bureaucratic, with each and every department jealously “guarding its turf”. I can give many examples of this within my own operating division with the conflicts that develop between transportation and equipment over allocations of vehicles to routes as an example.


  5. Gord:

    My apologies to you and the many men and women who take pride in their work with the TTC — and I agree with you when you suggest that they represent perhaps “99%” of the employee base. I think that is a considerable over-statement on your part! You then go on to point out that the rotten apple in the basket disproportionately spoils the rest of the fruit.

    My post to Mr.Munro was prompted by his untiring efforts to improve the transit that most of commuters want to appreciate and respect — from the yardmen to fronl line faithful drivers such as Gord.

    It is all about team work. Kudos to Steve for pushing this initiative in such an analytical and persuasive manner.


Comments are closed.