the geometric shapes of transit’s success

In my work for transit agencies, I'm always insisting that reports should not just explain how routes perform (typically in ridership per unit of cost) but also why.

Here's one partial example from an infographic developed by TransLink, the transit agency in Vancouver, Canada.   




All other things being equal, long, straight routes perform better than short, squiggly and looping ones.  The reasons are obvious to most transit riders (and are laid out in detail in Chapters 4 and 14 of my book) but you'd be amazed how many well-intentioned people  haven't figured this out, and continue to advocate land use patterns that make effective transit impossible.  (Mantra:  It's not Transit-Oriented Development unless it's oriented toward transit that can succeed.)

Now, TransLink can use this in their explanation whenever someone demands that a route should squiggle to serve their interests. 

A core of my own practice is in developing ways to build understanding of the causes of transit's success, so if your transit agency is struggling to explain productivity, put them in touch with me!

23 Responses to the geometric shapes of transit’s success

  1. Alon Levy August 26, 2013 at 3:02 pm #

    What anchor is there at the southern end of the 3 or the 20, or at either end of the 16?

  2. Edward August 26, 2013 at 3:23 pm #

    Seems logical, though I’m not sure that the top and bottom graphics are fairly comparable. Most of the routes in the latter carry the “C” prefix, which indicates a community shuttle – often small buses that serve an area in which long buses are impractical or where potential ridership isn’t high enough to support long buses, but distances, hills, or other barriers make accessing larger routes difficult for at least some users (especially seniors).

  3. neil21 August 26, 2013 at 4:08 pm #

    Perfect illustration of the difference between ridership (top) and coverage (bottom).
    In fact – for Edward – the C might as well stand for Coverage! Predictably, and hence intentionally, unprofitable routes for ‘social’ goals.

  4. Jarrett August 26, 2013 at 4:15 pm #

    Edward. The small buses assigned to “C” routes are a result, not a cause, of low ridership potential. I believe TransLink is planning to drop this designation so that they can more freely assign vehicles sizes by trip based on capacity needs.
    Alon. The 3 lines you reference prosper despite relatively weak southern anchors, although 16 ends in a business district (Marpole) and 3 usually ends at Marine Drive Station, where towers are under construction.

  5. Kyle Z. August 26, 2013 at 5:42 pm #

    The lowest preforming routes in Metro Vancouver serve as “coverage” routes as outlined in Human Transit. Four of the lowest preforming routes (C88, 606, C84, C89) lie in Delta and Ladner (two cities south of Vancouver). Ironically, it is the mayor of Delta that is calling for more transit service and a separation from Translink.
    So is it fair to cut service from this part of the region because the routes are low preforming even though the residents that live there equally contribute taxes?
    Regarding the 20 and the 8 (not shown) buses, I think it would make sense to create a southern anchor by creating a transit exchange at Knight and Marine Drive. Maybe Jarrett can comment on that.
    Note that due to the lack of a southern anchor on the 20, Translink has decided to cut every second bus at 54th street.

  6. Christopher MacKechnie August 26, 2013 at 5:48 pm #

    There is kind of a transit exchange on Marine Drive and Knight of sorts, with the 22, 405, and 407 stopping in a dimly lit space under the Knight St bridge. All that would be needed is an extension of trolley wire on Marine Dr.
    It’s worth noting that most, but not all, of the community shuttles used to be routes operated with conventional size vehicles and given conventional route numbers.

  7. Adam August 26, 2013 at 7:07 pm #

    @kyle: they are adding articulated trolleys to another route so I suspect the 20 is being cut back to free up vehicles.

  8. mike0123 August 26, 2013 at 7:35 pm #

    The 3 and 20 are very busy south of Hastings, where they both turn to the west, and not at all busy on Hastings. This is because it is faster to get downtown (and nearly everywhere) by connecting to SkyTrain from nearly everywhere along the street.
    For the 20, the point along Commercial where it becomes faster to take the 20 directly is so far north – close to Venables – that it is loosely within walking distance of Hastings. With so many other routes on Hastings, there’s no reason for the 20 to join them.
    Both the 3 and 20 are slowest on the least used parts of their routes. The 2005 Vancouver-UBC Area Transit Plan recommends this change be made to the 3. It should also be made to the 16 further east and the 20, so that frequency can be concentrated where it will be most used.
    The southeastern corner of Vancouver proper has routes that loop like community shuttles. All of the north-south routes in Vancouver should be extended to cross the city from Hastings to Marine Drive with anchors at:
    Marine Drive Station for the 3, 15, 16-Arbutus and 17.
    Knight Street Bridge for the 8, 20, and 22
    West or East Fraserlands for the 16-Renfrew-Elliot and straightened 27-Kerr and 28-Tyne

  9. Jonathan Hallam August 27, 2013 at 1:26 am #

    Setting the merits of the routes aside, I think this is a bad infographic. Why is 259 bad – it’s straight and direct. 16 is far from a direct link between its ‘anchor’ points, yet its somehow ‘good’. C89 is as ‘direct’ as 319 – any set of vertical and horizontal lines connecting both corners of a square will have the same length [since neither doubles back], yet one is bad and one good. The little loop at the end of 20 is okay but the one at the end of n16 puts it in the dustbin. I can believe there are real reasons for all of these – but as an infographic its just confusing.

  10. Beta Magellan August 27, 2013 at 1:54 am #

    The issue with 259 isn’t so much route geometry as the route itself—it connects two small, seaside communities in West Vancouver and the bulk of its route is along an barely-populated stretch of the Sea-to-Sky Highway. I could tell it was a coastal route and suspected it was up against hills by what I know of Vancouver, which gives it a small catchment area, but I think you’re right in a sense—it’s not necessarily low-performing because of route geometry but because of the geography it serves. Although the diagram makes the point about circuitous routes better by leaving out just where they serve, it’s worth remembering that for at least some of these routes there’s more at play than just poor route design (259 strikes me as about as well-designed as it can be for its given market; whether there should even be a bus there is a different question).

  11. Alon Levy August 27, 2013 at 2:13 am #

    Jarrett: yes, Vancouver is building towers along Marine Drive, but they aren’t there now, so the 3 is perfectly productive without them. Whatever is there now, it’s not UBC, SFU, or Metrotown.
    More broadly, the metrics used in the infographic assume that overcrowding is good – routes get to be in the infographic based on boardings per hour, capacity utilization, and cost per boarding. The first and the third are useful for the passenger; the second is a measure of pain, and is only useful if you believe that raising passenger-km on local urban routes is good in itself.
    Measured by cost per rider and boardings per hour, the unanchored 8 is more productive than the 49, which has anchors but nothing in between. The key here is that the 8 has high turnover, so it achieves more passenger boardings per hour but fewer passenger-km. And if turnover is more important than anchoring, then Vancouver shouldn’t be building mid-rises at Oakridge and Langara and towers at Marine Drive, but the reverse: it should create development patterns, both on the local street level and on the regional level on SkyTrain, that look more like those on the 3 or the 8 or the 20 than like those on the 49.

  12. Greg Ryan August 27, 2013 at 6:42 am #

    Jarrett, interesting post. How long is too long when it comes to ideal route length?

  13. M1EK August 27, 2013 at 8:08 am #

    Isn’t the loopiness/turniness of some of these routes a somewhat inevitable unintended consequence of bragging for so many years about buses supposed flexibility?

  14. Sedela August 27, 2013 at 8:23 am #

    what if the streets of your metro area force you to have a lot of bottom half type routes?

  15. Jarrett August 27, 2013 at 11:35 am #

    Greg. “Too long” is mostly “too long to operate reliably” because reliability is a function of how long the vehicle has gone since it’s last driver break. Usually 60-90 minutes one way.

  16. Mike August 27, 2013 at 3:53 pm #

    I think it is dangerous to start calling routes good or bad, as all routes contribute to making a viable public transit service for a metropolitan region.
    Sure we should promote proper development that allows for efficient bus routes. But we should not be demonizing low performing routes, as not every route is going to be a winner, and public transit is about serving residents.
    It is a well known fact that Translink’s obsession with bus performance is borne from a financial crisis, and nothing more. Translink is trying to justify a bus optimization program, and these graphics are part of that.
    But the truth is Translink can’t really optimize bus service that operates at dismal service frequencies of 30-60 minutes, so they have to demonize the outer suburban bus routes to justify cutting service in some cases entirely to areas.
    You can optimize service on super high frequency service by shifting buses around, and this is what Toronto does. But most of Translink’s optimization is coming at the expense of outlying bus routes which if anything should probably have more service to attract people to transit. Not cuts.

  17. mike0123 August 27, 2013 at 6:14 pm #

    In the parts of Metro Vancouver where most buses have frequencies of every 30 minutes or less, the network is designed around timed transfers. A pulse network minimizes travel times when frequencies are low, but it always comes with a lot of deviation and delay from buses having to go out of their way to then wait at exchanges.
    The inherent waste in a typical timed transfer network might be 5 minutes of deviation to get to each pulse plus 5 minutes of waiting at each pulse. If a pulse happens every 30 minutes on each route, 1/3 of service is wasted. In other words, a timed transfer network with routes running every 30 minutes might be replaced by a random transfer network of routes running every 20 minutes.
    Some parts of Metro Vancouver still have networks that were designed (or more likely just kind of ended up that way) for timed transfers, but the network no longer has any actual pulses. These networks, especially in Burnaby, New Westminster, and the southern half of Coquitlam, are the best candidates for optimization.
    North Vancouver has a remarkably extensive and well-designed timed transfer network, but its busiest routes no longer meet the pulses. The 239, the most frequent and practically only east-west bus route, wastes 12 minutes getting to the pulse at Lonsdale Quay but doesn’t wait for it. This route is duplicated by other routes that wait for the pulse everywhere that the Seabus is the fastest way downtown. There’s no reason for the deviation any more and it needs to be straightened.

  18. theo August 28, 2013 at 12:47 am #

    I really like this graphic. While oversimplification is always a danger with infographics, this one makes a point without being moralistic.
    A note about the 16: @Jonathan Hallam
    The 16 is successful because it acts like two separate lines:
    – a North-South route (Western half) connecting Marpole, Kerrisdale, Arbutus Village, and False Creek / Kits to downtown (all dense residential / commercial areas)
    – a SE-NW dogleg route that connects two skytrain stations with the PNE, Hastings Street, and downtown. Hastings Street is essentially one long commercial strip from renfrew all the way to downtown, so it can sustain a large ridership (both the 16 and the 135 serve Hastings, in addition to the 10 and, I believe, a couple of other routes.)
    The “U” shape only works because (i) it is fed by a variety of smaller anchors throughout, and (ii) the largest anchor is in the center of the route. In other cities the 16 might be broken up into two separate route names served by the same bus, which would alter its sign once it reached the downtown transit mall.

  19. Jeff Becker August 28, 2013 at 12:26 pm #

    Actually, simple graghics aren’t nearly as good as those that pack a lot information into one. Edaward R Tufte wrote a series of 4 books beginning in 1990 explaining and demonstrating this in detail (kind of the guru on the topic). As suggested above, the gragpic should show densities, activity centers, bus connections, etc. to make the point. Also, classification of service to local, express, community, etc. is absolutly essential for policy makers, riders and the public to understand the application of these performance measurement concepts.

  20. Daniel August 29, 2013 at 6:03 am #

    A study of bus routes in Melbourne last year noted that for some routes (from terminus to terminus) the route length was more than double the shortest distance by road.

  21. Mike August 29, 2013 at 4:15 pm #

    One thing I would like to know is how many of these low performing routes are “low performing” because Translink never invested in offering an attractive service level on these routes? Because of course bus routes running every 60 minutes are going to have dismal performance.
    I also think the low density argument is used a little to often in this case. Many of those low performing routes which only operate every 30 or 60 minutes serve areas no different than suburban areas of Toronto where buses come much more often and are full.

  22. Al Dimond August 29, 2013 at 11:34 pm #

    If you try to look at the graphic without assigning a value judgment to high and low performance (just for a second) it tells you something different than, “One route good, another route bad.” It tells you that when you look at the performance report you shouldn’t be surprised to see routes with certain characteristics doing well, and routes with other characteristics doing poorly. If a low-performing routes still performs a vital service, the agency can say, “Look, this route can’t be expected to perform well on these metrics with characteristics like this. It’s still the best way to perform a necessary service.”

  23. Jase December 19, 2013 at 11:29 pm #

    The word “performing” could be unpacked at length, in the context of this infographic.
    Jarrett, you always say coverage can be an acceptable goal for transit agencies to pursue, but the unqualified reference to “performance” suggests this particular graphic is biased to ridership.
    The comments by Jonathan Hallam above are valid. Much of the information in this graphic comes from the dot points. It’d be good to use shading, width of line, or colour to indicate the density of jobs and dwellings along the routes.
    For example, I can infer route 259 is rural because it gently wiggles. But I cant be absolutely sure.

Leave a Reply