09 April 2011

Mind The Service Gap

Having cobbled together additional funding for the next fiscal year, Caltrain has published a revised 76-train timetable that is now proposed to go into effect this summer. This timetable is an improvement over the previous proposal, a controversial 48-train peak-only timetable (previously analyzed here) that caused an outcry up and down the peninsula. To save $3.3 million in annual operating costs, the 76-train timetable cuts ten trains from today's peak service by eliminating the Baby Bullet express trains and substituting fewer limited-stop trains.

How does this new 76-train timetable stack up? We put it through the Metricator to extract key rush-hour trip time statistics, using the same methodology as before... with the expectation that the results might not be so great.

First, as a reminder, we consider the current timetable. The basis of comparison is today's 86-train-per-day, 5-train-per-hour timetable, to which we assign a score of 100.
And now, the proposed 76-train timetable:
Wait, is that right? One-hundred-and-four?

Did You Say Better?

Four trains per hour with no Baby Bullets is better than five trains per hour with Baby Bullets? Surely this shocking result must be wrong?!? The short answer is no, the numbers don't lie.

The long answer requires a little bit of background discussion. You see, the Baby Bullet has a dirty little secret. In order to achieve such stellar trip times, other trains must clear the tracks ahead of it, because if the express ever caught up to a local, then it would no longer be an express. In practice, that translates to very long service gaps just before a bullet comes through. Service gaps (i.e. how many minutes pass between two successive departure times for a given origin & destination pair) are an important component of the convenience of taking the train. In these calculations, an effective trip time is computed based on the following sum:
  • 70% of the average trip time
  • 30% of the best trip time (to favor express service)
  • 20% of the mean wait between trains (far less than the random arrival figure of 50%)
  • 15% of the maximum service gap (to penalize very large gaps between trains)
What happened in the new 76-train schedule can be understood by taking a closer look at the numbers. Some best trip times got a bit worse, due to the lack of Baby Bullets; some mean wait times between trains got a bit longer, due to 4 rather than 5 trains per hour; but the maximum service gaps dropped precipitously, enough to outweigh the other effects when considered for all origin and destination pairs, even after weighting the results by ridership.

Various other observations on this new timetable:
  • Rush hour service is significantly improved for stops such as Burlingame, Lawrence, Sunnyvale, Cal Ave, etc. as can be readily observed in the effective trip time savings. Note these are all the stops where service was degraded when the Baby Bullet service started.

  • Closing Hayward Park makes sense. Despite all the talk of transit-oriented development at this location, the fact remains that ridership is so low that wasting three minutes of everybody else's time to stop there isn't worth it. Besides, the station will be less than a mile away from Hillsdale when that station is moved to the north, as long planned.

  • There is a long and awkward service gap departing San Francisco between 6:45 and 7:30 PM.

  • San Bruno, where Caltrain is investing a nine-figure amount to rebuild the station, is left with dismal rush hour service.

  • Express service isn't inherently bad. It adds the most value when trains ahead don't have to get out of the way, i.e. there is a way to overtake trains without penalizing local service by imposing large service gaps ahead of the express. There exists a six-train-per-hour timetable that scores 145... but it requires a mid-line overtake. That's why a phased implementation of future peninsula corridor improvements should include a mid-line overtake facility as extensively discussed here and here.
So should the Baby Bullet be scrapped? That depends on one's beliefs... namely:
  • Do service gaps matter? The opinion presented here is obviously that they do. Caltrain seems to consider trip time only in terms of how long a passenger spends on the train (see page 16 in their presentation), without much regard to how long they might spend on the platform waiting for the next train. The underlying assumption is that every passenger builds a routine around the same train every day, and shows up just in time for that train.

  • How valuable is the Baby Bullet brand? There is reality, and then there is perception. People buy a product based on their perception, and the Baby Bullet undeniably has a certain cachet. The Baby Bullet has been marketed very effectively, to the point that most people believe that Baby Bullet trains are faster than they actually are. That belief doesn't show up in the raw metrics, but it does enter into people's choice of transportation.
When all is said and done, the choice between the proposed 76-train timetable and preserving today's timetable comes down to $3.3 million. That amounts to a mere 3 percent of Caltrain's annual operating budget, and makes this particular decision pale in comparison to far more pressing issues such as providing Caltrain with a dedicated funding source and advancing the electrification project.


  1. I think that reducing the service gap is important at stations where you:
    * Transfer between caltrain and muni/vta/bart/samtrans bus/light rail
    * Can easily bike/walk to your destination/origin

    On the other hand, the service gap is not important to travelers who take an employer shuttle from their destination station (usually on peninsula), as the those shuttles show up only a few times and are timed to meet the baby bullets.

    Therefore, riders of employer shuttles find baby bullets more important than lower service gap since their commute is fairly optimized.

    Riders who don't use a shuttle would want a minimal service gap since they're not tied to any schedule.

    Based on your statements and Caltrain's marketing message, Caltrain caters more to "employer shuttle" passengers which tend to be their higher end clientele. Then again, with the proposed schedule, are they having a change of heart? Does Caltrain provide statistics of how many passengers take employer shuttles? (They only seem to list the Caltrain sponsored shuttle ridership)

    Disclaimer: I'm an "employer shuttle" rider, so I'm in favor of preserving baby bullets and minimizing my commute time.

  2. "Does Caltrain provide statistics of how many passengers take employer shuttles?"
    Only every month.

    5468 shuttle "riders" (presumably meaning "trips" -- half that number of humans aboard) of 36682 train "riders" in Feb 2011.

    So a bit under 15%. This doesn't include Marguerite, which is the 400lb gorilla, or the far greater hordes who are forced to endure the Muni Catastrophe at the northern end of the line. (FYI circa 4000/day use PA station, if not the Stanford shuttles, and 8600/day use 4th&T, if not Muni.)

    Many of those have alternatives to the shuttles, and very few (almost none) are taking shuttles and then riding a train the fulll length of the line, so not only as Clem states the comparatively small train run time end to end is a fairly small deal compared to fixing absolutely insanely bad headways in the existing Doty-esque "Baby Bullet" confection, but for real riders (nearly all of whom aren't going SF-SJ or anywhere-SJ for that matter) the on-train penalty is half or less than the end to end slowdown.

    Now to flame tangentially ... The shuttle program is total complete mess ... and I speak as somebody who commuted on Caltrain and worked in south valley sprawl hell nowhere near any station for nearly two decades.

    If anybody really cared about increasing transit ridership by giving away free bus trips they'd run free and NON-MUNI buses from 4th&Townsend to Market Street in SF, which, despite it all, is where the real market is. Instead we throw tax money at a handful of empty buses in the south bay that, to a large extent, simply duplicate empty and comparably speedy, but un-coordinated VTA and SamTrans routes. (Not just to "to a large extent" to my two longest-term employments sites, but literally the same route, sucking passengers from the county buses. Crazy!)

    What the peninsula needs is predictable (meaning every half hour, or every 15 minutes) train service, that can be timed and coordinated with connecting useful bus service, with coordinated and integrated (which doesn't mean the defense contractor welfare program scam that is "TransLink"/"Clipper", but time- and cost-free intermodal transfers). Running a shuttle bus from an office park to meet a train that only comes once an hour is a dead end.

    Anyway, the OBVIOUS problems with Caltrain's timetable are:
    * No predictable "Takt", so no way to make real connecting transit really work. In effect they have three completely different schedules: peak (different NB and SB), off-peak weekday, and off-peak weekend. Nuts.
    * Hourly off-peak service is almost at the "why bother" point, let alone 45 minute service gaps right at the edge of peak hour from the highest use station on the system.
    * Terrible, wretched, abysmal, appalling operating inefficiency due to FRA regulation, 19th century trains, lack of level boarding, abysmal equipment design, and unconscionably high (2-3 too high!) crewing levels. Electrification as proposed by Doty and Caltrain does nothing to address any of this ... it would just be $700m down the drain!
    * Tamien-Gilroy is a basket case and in any even remotely rational world would have been killed a decade or more ago. It just shows that neither Caltrain staff nor the Caltrain "saviour" agencies give a damn about practical and effective service.
    * They've completely and indefensibly abandoned San Bruno.
    * College Park station exists. Good God.

  3. Richard said:
    "Does Caltrain provide statistics of how many passengers take employer shuttles?"
    Only every month.

    5468 shuttle "riders" (presumably meaning "trips" -- half that number of humans aboard) of 36682 train "riders" in Feb 2011.
    Actually, are you sure that this includes the private Yahoo, TiVo and Apple shuttles that I see in MV? I have a feeling that Caltrain only counts the official "Caltrain logo shuttle" riders in that statistic. I can probably figure that out quickly by asking our HR if they report the shuttle counts to Caltrain.

  4. Richard said:
    "* No predictable "Takt", so no way to make real connecting transit really work. In effect they have three completely different schedules: peak (different NB and SB), off-peak weekday, and off-peak weekend. Nuts."

    Electrification can help with this, of course, but it does the most good with a mid-Peninsula overtake.

    "* Hourly off-peak service is almost at the "why bother" point, let alone 45 minute service gaps right at the edge of peak hour from the highest use station on the system."

    Electrification can help with this as well.

  5. This makes much sense. At peak times journey times are less critical, with congestion on roadways negating some of their benefits. Also peak trips are inherently well suited to transit, with no luggage and not tying up a car for a whole day. It is the off-peak where the Bullet trains could make a difference. There is surely a market for fast all day trips between SJ and SF? Most trips, as a society, we make are off peak- even on rail services (where an off-peak service is provided!). Despite problems, the Caltrain route is a continuous urban corridor with a major urban center at each end. Peak fares, better off-peak services would go someway to righting the current crisis, and vindicate the investment we tax payers are making.

  6. That's an idea. Have local trains during peak hours, and baby bullets (only?) during off-peak hours.

  7. "At peak times journey times are less critical, with congestion on roadways negating some of their benefits"

    Ah, no. Memo from the real world to blog commenters: Transit+Caltrain+transit is not time-competitive for nearly any journey at any time of the day.

    Pretty much any carpool+highway will beat Caltrain every single time, and drive alone will do so in the vast majority of cases.

    The reason for not running express service off-peak is because the train (with worse than wretched connections on BOTH ends) is so non-competitive that there is no demand to run those extra empty seats. It's crazy to talk about "saving" time when the frequency of service is so miserable: nobody is going to wait an hour for an express train instead of getting in a car NOW to make any trip except, perhaps, at peak congestion hours.

    The reason for running express trains during peaks is because there is demand for extra seats (as an marginally less unpredictable alternative to road congestion) at that time, along with the fact that running only slow all-stops trains both discourages extra revenue riders while increasing operating costs.

    A solid ridership base requires predictability (Caltrain doesn't have), excellent access/transfers at both ends of the train (Caltrain's genius engineering department spends hundreds of millions preventing this!), good and predictable and memorable (ie 15/30 minute Takt) headways, good and predictable and memorable service patterns (ie no random mish-mash of many different semi-express stopping patterns), comfortable and quiet and safe vehicles (hah!), and fast trip times.

    It's a system. Look at the systems that work.

  8. "No predictable "Takt" ... Electrification can help with this, of course ...

    "Hourly off-peak service is almost at the "why bother" point ... Electrification can help with this as well."

    Those are nice theories, but have no grouding in Peninsula Corridor Joint Powers Board reality.

    Sadly, due to the unilateral and bat shit insane decision by Caltrain's former "Chief Transformation Officer" Bob Doty that running under FRA rules along with freight would be a "fun project" (his literal words to me to explain his decision!), Caltrain plus overhead wires will be almost exactly as bad as today (see their Draft EIR!) while costing more to operate!

    Anywhere outside North America one might very well expect a 19th-to-21st century transition in operating efficiency and patronage to accompany a billion dollar spend on a brand new rolling stock fleet and a brand new power system, but that's not what Caltrain wants or what Caltrain's "engineering" department is "designing".

    Sad, but true.

    As usual the problem is that agency staffers seek solely to spend money, while having less than zero interest in providing service. (Just look at the electrification DEIR! Not a single timetable. Just look at Doty's "Caltrain 2025" plan. No a single timetable!)

    I don't see any way out except for the agency and its staff to be erased from the face of the earth, and I've spent decades looking for alternatives.

    Strategic rail planning = Timetable + Rolling Stock + Infrastructure.

  9. Richard: Unfortunately, in many cases, there is no VTA service that duplicates what is provided by the Caltrain shuttles. For example, there's nothing that runs from downtown Mountain View to Google. There's nothing at all that runs to the Lawrence station, so you'd have to either go to Sunnyvale and take the 32 followed by a long walk, or to Santa Clara and double back on the 60, possibly also followed by a long walk. And all a Takt schedule will accomplish is making it so that there's a 27 minute wait between the train and the bus every time, rather than just most of the time as it is right now.

  10. Adirondacker1280011 April, 2011 13:38

    Anywhere outside North America one might very well expect a 19th-to-21st century transition in operating efficiency and patronage to accompany a billion dollar spend on a brand new rolling stock fleet and a brand new power system, but that's not what Caltrain wants or what Caltrain's "engineering" department is "designing".

    You have to get out more. Anywhere outside of the Bay Area.

  11. "You have to get out more. Anywhere outside of the Bay Area."

    Oh dear. I had it all wrong, I see now. Metro North crewing levels, equipment costs, fleet maintenance costs, track maintenance costs, fleet availability, fleet utilization, track availability, power system reliability, ride quality, and energy efficiency: who could possibly want anything more?

    But wait, there is in fact more to want! Forget NYMTA, forget LIRR/MNCRR and even the futuristic MBTA commuter trains ... it turns out SEPTA is clearly Caltrain's staff and consultants' role model, and by God SEPTA is the best we deserve. Good enough? I'll say!

    I must have been hiding under a rock during the 15 years I lived in the northeast of the USA not to have understood that I was living at the epicentre of rail transportation excellence, and I only wish I'd been a more humble student when I visited the holy of holies of New Haven and Philadephia, but, well, now, so belatedly, the scales have been lifted from my eyes! Thank you, sir, thank you.

  12. Adirondacker1280011 April, 2011 15:39

    Electric trains running on track that has PTC to suburban stations with level boarding. It's gawd awful,

  13. The proposed schedule has pluses and minuses. The shorter trip times to Mountain View currently arrive exactly at the time the Winchester trains are supposed to depart. Net result: approximately 14 minute waits for the next train most days. The longer trip times actually connect better. The total trip time to my destination is about the same no matter which train I take.

    Sunnyvale service will be better, for sure, and the 32 bus goes closer to my office than light rail does.

    I was amazed there was no passing zone around Redwood City when the baby bullet service began operation.

  14. Trains can't really pass at Redwood City without one of them actually slowing down and stopping outside of the station. Ideally, passing involves at least one station so that no one looses any time.

  15. This new schedule is essentially skip-stop service. Also, if you're not going to serve a station in rush hour, you might as well shut it down completely.

    Personally, I think the schedule would be better with local trains every 30 minutes and express trains every 30 minutes (stops @ Sunnyvale peak direction, Mtn View, Palo Alto, Redwood City, Hillsdale, and Millbrae). I also still suggest cutting Gilroy service if funds can't be found.

  16. Actually, make one of the Express trains a limited that makes express stops + California Ave, San Carlos, San Mateo, Burlingame, and S. San Francisco.

  17. Well, SEPTA did plan on modernizing, and had artifacts of an S-Bahn-style service plan until recently. The managers were used to urban transit and figured that running commuter rail the same way would produce efficiency. They even had plans to sever the Reading side from the FRA mainline network.

    Then the railroad old-timers rebelled, and everyone with any experience running trains left for Amtrak and Conrail. Eventually, management caved, and SEPTA remained a marginally better version of your steam-era American commuter network.

    The less said about the hourly trains and 75 mph intercity speed limit of Metro-North, the better. Though, looking at the five-train-per-day schedules and mixed high-plus-low-platform stations on MBTA has made me appreciate Metro-North just a little bit more.

  18. @Caelestor, if you think you can do better, please submit a tab-delimited text file of your schedule (using similar station-to-station timings). Running it through the Metricator takes all of ten minutes and we'll see what score you get. That's the beauty of metrics, there is no room left for opinion.

  19. @Clem there's still plenty of room for opinion, it's just a different sort of opinion. We can debate whether your metric is good enough, whether it includes enough variables, and whether they're the right ones, as well as whether they're weighted appropriately. We can also ask what empirical basis your metric has. Still, basing the discussion around metrics makes the true nature of the problem clearer: we're dealing with constrained optimization, we can't give everyone everything, and so we need to balance priorities to maximize some sort of objective.

  20. @Arcady, well stated. Poor choice of words on my part. The subjective discussion is removed from issues such as how many express trains should stop in Sunnyvale to the much more important framing questions regarding what makes good service. That is a discussion we need to have, and Caltrain isn't conducting it in public.

  21. Argh, my computer restarted on me, and I have no time to create the massive text file, but here's my proposal for NB morning peak (SB evening peak would be similar).

    BB train every hour, on the hour. This brand is arguably the reason for Caltrain's resurgence and can't be stripped away completely. Takes 1 hour from SJ to SF, stopping @ Sunnyvale, Mtn View, Palo Alto, Hillsdale, and Millbrae.

    Local train every hour, 20 minutes past the hour (I have to manipulate the schedule because of the lack of a bypass at RWC and prevent congestion at SF). 86 minutes, 19 stops.

    Limited train every hour, 10 minutes past the hour. Stops @ the BB stops + Santa Clara, Lawrence, Cal Ave, RWC, and San Mateo. 10 stops, 70 minutes. Keeps 20-30 minute interval at these stations.

    Semi-limited train every hour, 40 minutes past the hour. Makes all stops to San Carlos except San Antonio, then stops @ Hillsdale, Burlingame, and Millbrae (sorry, S. SF has too low ridership). 12 stops, 72 minutes. Keeps 20-30 minute interval at these stations or helps out growth at stations with high ridership in the past (Cal Ave, Santa Clara, Lawrence).

    I'll come up with the reverse peak schedule on the plane tomorrow.

  22. Keep in mind that the Baby Bullet brand applies to more than just the service. It also applies to the physical trains themselves, and that makes a huge difference. Finally, passengers get modern trains with decent suspension. They get tables and power outlets so they can work on their laptops, and they get a train that doesn't look (and bounce) like some relic of the 1950s. When I first rode Caltrain, I used to think the tracks must have been in awful condition, until I rode a Bombardier train and realized that no, it was just that the gallery cars had really awful suspension.

  23. Adirondacker1280013 April, 2011 14:32

    here's my proposal for NB morning peak

    Sounds great for the first hour. What happens in the second hour when the second express of the day has to pass trains that left in the first hour?

  24. "Keep in mind that the Baby Bullet brand applies to more than just the service."

    The assignment of gallery (1940s design) vs bi-level (1970s design) cars to "bullet" vs other trains is somewhat random from the point of view of passengers.

    In fact Caltrain (sometimes) assigns more gallery sets to the fewer-stop trains in "reverse" commute direction, because of their higher bike capacity. (Sometimes. Not enough to rely upon for a reliable trip to and from work, mind you!)

    As for Caltrain "brands", they have three campaigns going:
    * this perennial never-goes-out-of fashion classic (the Little Black Dress of Caltrain style);
    * "No can do!";
    and the very most up to date makeover,
    * "The sky is falling!".

  25. Sounds great for the first hour. What happens in the second hour when the second express of the day has to pass trains that left in the first hour?

    Caelestor's proposal requires no overtakes.

  26. @anonymous yes, I'm aware that the the assignment of equipment to runs is somewhat random, and sometimes varies from day to day (though generally not that much). But still, in my experience, many pasengers consider the "bullet" to be not the train that goes fast but the one with the pointy looking nose.

  27. By all accounts, my commute (22nd-Menlo, 'reverse') would be same/improved, with better frequency/predictability. However, I worry about crowding- some of these peak hour trains are pretty full now, and it seems like the pseudo-bullets lumping together the most popular stations (MP+PA, for example) sets up the potential for some capacity issues. Has anyone gone to the next level of looking at current occupancy rates and predicting how full these would be?

    Don't even get me started on Giants game days...

  28. Do they still have to run the same number of trainsets for this schedule? If not, they could take lengthen peak trains by one or two cars to ease the capacity issue, could they not?

  29. My back of the envelope calculations show that the new timetable reduces the required number of consists from 16 to 13, which means that they can lengthen trains from 5 cars to 6 with no increase in required cars. But the peak hour loads are not evenly distributed: the express trains are the most heavily loaded, and there are some other trains that are half empty, so while going from 25 cars per hour to 24 will be a slight reduction in capacity, the new timetable should even out demand and ensure that the best use is made of the capacity. One can also cynically conjecture that the 20% decrease in trains won't be a problem given the likely accompanying decrease in ridership.