Photo by the Author

Since October 2004 Sounder commuters with full-fare passes have enjoyed free access to Amtrak Cascades between Seattle and Everett through the RailPlus program.  Barbara Gilliland, then Sound Transit’s Deputy Director of Transportation Services, called it, “One of the easiest agreements I’ve ever worked on.”  Yet very few riders utilize the service; in February 2011 only 126 RailPlus tickets were issued for the entire month.  (Equivalent to 2 people making one round-trip per day!)  Ridership for the past year has generally ranged between 80-160 boardings per month.

Cascades times north of Seattle are hardly ideal for commuter use, with two-peak hour trains from Seattle (510, 516), one mid-day train from Everett (513), and one late night train from Everett (517).  Further, only full-fare passes are accepted, with no E-Purse upgrades permitted.  Due to the higher fare on Sounder vs. ST/CT buses, most Northline Sounder riders have employer-subsidized passes, increasing the likelihood that riders are peak commuters into Seattle for whom the schedules would be unworkable (except for Train 516).  Throw in mudslides, general reliability issues, and the ease of express service from Everett on ST 510, and you have a system that structurally disincentivizes people from trying the train. More after the jump.

What about RailPlus between Seattle, Tukwila and Tacoma?  Wouldn’t it be great to offer the equivalent of mid-day Sounder service without requiring any additional track easements or operating costs?  Last September I asked Vickie Sheehan at WSDOT about the idea and to ask how many seats are usually available between Seattle and Tacoma:

The average load factor is not the deciding factor when making arrangements to accept commuter tickets on Cascades, or Amtrak tickets on Sounder trains.  The factors to be considered involve the adequacy of the existing services and the number of days trains are at or near capacity, rather than the average load factor.  Current demand for travel on the Seattle to Portland route is much higher than between Seattle and Vancouver, B.C. and is growing. Ridership projections do not show sufficient space on trains to make the Rail Plus fare an option to be pursued at this time between Seattle and Tacoma. In fact our projections show that we shall soon be in a position where demand outstrips supply.

While I understand the point about capacity, it is my understanding that Amtrak ticketing procedures overrepresent their load factor to make fewer seats seem available.  According to Amtrak staff, seats cannot turn over between either Vancouver BC-Seattle or between Seattle-Portland.  A rider from Everett to Seattle, for example, ‘ghostrides’ the train between Vancouver BC and Everett and that seat cannot be sold again until south of Seattle.  Thus a ‘sold-out’ train will always have many empty seats.  So while theoretical capacity may be reaching its limits, a more functional ticketing system with the ability for multiple turnovers would allow commuters to sit in seats destined for other folks between Tacoma and Portland.

South Sounder has an endpoint problem.  The 55 minutes between Freighthouse Square and King Street aren’t competitive either on time or cost with the 590-series, whereas from Puyallup northward Sounder is the fastest commute available.  Adding RailPlus would improve Tacoma ridership by providing the option of a 45-minute train between Seattle and Tacoma.  I understand this will only happen, if at all, after the Pt Defiance Bypass when Amtrak moves to Freighthouse Square.

It’s an attractive vision;  commuters using Amtrak and interurban non-commuters on Sounder.   Maybe Sounder is too cheap, but Amtrak’s $13-$20 between Tacoma and Seattle is way too much.  Can we find a better synergy between South Sounder and Amtrak?

50 Replies to “Improving RailPlus”

  1. I’m trying to figure out how/why Amtrak would have a ticketing system like that. It’s as if the people who built their software went out of their way to make it broken.

    1. Amtrak’s ticketing system dates from the early 1970s. They’ve had replacement of it as an urgent priority item for about 4 years. Enough said?

      (The current system was an improvement over the non-computerized, non-centralized system used prior to the early 1970s, where piles of physical tickets were allocated to individual stations, so if the allotment of tickets ran out, stations had to call other stations on the phone to see if they had any tickets they could send over…)

      1. I know that the Amtrak system handles repeated turnover of seats. And I would be very, very surprised if it were impossible to turn over a Seattle-Olympia seat for an Olympia-Portland seat.

        But they may be hitting some sort of hard-to-alter limit on the number of separate segments their system can handle, given that it’s a mainframe program written in the 1970s for a smaller number of routes (not including Cascades), and the paper system used before that did not handle all possible seat turnovers. So, for instance, Seattle-Tacoma might be programmed as one segment.

      2. One of the reasons they do this might be becasue they use open seating on the train. ARROW can do seat assignments, however it wouldent be good PR if a couple got on a train, and one seat was in car 3 and the other in car 5, even though there were 2 open seats. If you were to increase turnover on the train and still maintain the higher quality of service you’d have to switch over to assigned seating onboard. Actually, i find Sounder quite competitive with travel on I-5. First off its far more comfortable than being stuck on a bus lurching through traffic, since you’re always moving. you are free to move about the train where your pretty much stuck in the bus seat, the train is seldom late compared to who knows what on i-5. Sure it may take a few minutes longer endpoint to endpoint but the quality of the travel is well worth it.

      3. It wouldn’t make sense to have specific reservation-time coach seat assignments on a train anyway.

    2. “I’m trying to figure out how/why Amtrak would have a ticketing system like that. It’s as if the people who built their software went out of their way to make it broken.”

      Now that’s funny!

      Do you realize how much work it takes to put all those bugs into a system? Why do you think programmers get to spend so much time ‘debugging’ problems.

      Job Security!

      “It’s not a bug, it’s a feature!”


      “We didn’t bother testing it, since we only changed 8 lines of code.” This after a software failure that knocked out AT&T’s network (about 20 years ago).

      “That’s what I like about standards where I work, there are so many to choose from!”

  2. “According to Amtrak staff, seats cannot turn over between either Vancouver BC-Seattle or between Seattle-Portland. A rider from Everett to Seattle, for example, ‘ghostrides’ the train between Vancouver BC and Everett and that seat cannot be sold again until south of Seattle. Thus a ‘sold-out’ train will always have many empty seats. ”

    Which Amtrak staff told you that? And is there anyone from Amtrak who can confirm that? I’m pretty sure, based on personal experience and observations as a passenger, that it’s not true…

    1. It does not make sense to me either. I don’t see why their software could not know that a person going Everett to Seattle is getting off in Seattle, and that seat is then available for sale south of Seattle. I’ve also been on a few sold out trains that were quite sold out, with people overflowing into the bisto car.

      1. That’s what I’m thinking. You’d have to go out of your way to write software that broken. No-one would ever do that… right?

      2. According to their site, the system only handles two separate segments, one north of Seattle and one south. Thus, they *can* accommodate the case you describe (someone switching in Seattle), but if someone goes from Everett to Tacoma, say, the system will book their seat as if they were travelling from Vancouver, BC to Portland/Eugene.

        Why does it work this way? Well, scheduling is what’s known as an NP-complete problem, which is another way of saying that it’s very computationally hard (and not so easy for humans, either). In other words, programming a system that could handle seats that could turn over at any stop would have been significantly more complicated (and thus more expensive).

        My guess is that, historically, Amtrak simply hasn’t had the kind of ridership that would make a full scheduling system. There are very few routes that even sell out, and most of the ones that do don’t have much “through” ridership. Think of the NEC — either you’re riding between Boston-NYC-Washington, or you’re getting off before or after New York.

      3. This limitation isn’t credible.

        I know for a fact for sleeping car reservations, Amtrak’s system is fully capable of turning over the same sleeping car space multiple times on the same train.

        I find it implausible that for unassigned coach space on a train like the Empire Builder which travels from Seattle to Chicago via many intermediary points that if someone were to buy a ticket from Seattle to Leavenworth or Wenatchee, that the space would be unavailable for reservation for the entire rest of the trip to Chicago. And if the reservation software can deal with intermediate trips on the Empire Builder, why couldn’t it do the same on the Cascades?

      4. Bollocks. Nothing is computationally hard about managing a list of N seats, each with a variable-length list of endpoint pairs representing bookings, for any value of N you’ll ever encounter in real life. I don’t care if it’s NP-complete any more that any programmers with jobs care about the Pumping Lemma.

      5. Bruce,

        First, your claim about computational difficulty isn’t correct. I wrote a traveling salesman solver in college, and if I remember correctly, the largest problems we were able to solve were on the order of 100 cities. They were certainly closer to 100 than 1,000. Depending on the amount of seat turnover, and the algorithms used, the problem could easily have been intractable — especially on the kinds of computers that Amtrak used in the 80s when their seat assignment software was originally written.

        But the real point I’m trying to make is in response to this:

        You’d have to go out of your way to write software that broken.

        In this case, that’s simply not true. It’s harder — as in, requires more design and implementation work — to write a scheduling program which handles splitting seats across different trip segments than to write one which doesn’t. And it’s harder to write one that handles the general problem (i.e. turning over seats at any station) than to write one which handles a fixed number of segments, e.g. one north of Seattle and one south.

        You’re free to criticize Amtrak for not having spent the money to write better software, but it’s incorrect to assert that they went out of their way to cripple it.

      6. Amtrak’s reservation program, Arrow, is based on an airline reservation program that dates back to the 70’s. Airline reservation systems all work on segments, so unless every train route was broken up into individual segments between each station it would be very difficult for this type of software to deal with seat turnover. I’m sure that Amtrak tries to limit the number of segments on each route to keep things from getting too complicated.

      7. Thanks for confirming my suspicions, Zed.

        I’d still be a bit surprised if Sounder South was programmed as a single segment. But I’d also be very surprised if Seattle-Tacoma was *not* all part of one segment; if you’re minimizing the number of segments, there’s absolutely no reason to put a segment break at Tukwila, none whatsoever. Tacoma, maybe, Olympia-Lacey, probably, but Tukwila, no way.

    2. Any Amtrak agent who thinks they know how the system works is delusional, or lying.

      1. Whether Amtrak’s system can handle instantaneous repsponse to seat availability is an issue for future e-ticketing, where the conductors will ‘check in’ the passengers on board.

        Right now, printing the physical ticket is the only way the system knows who is supposed to be on the train.

        Seating does become available on a station-station segment if someone who purchased a ticket on the Internet or over the phone, but didn’t get down to the train in time to print their ticket (and presumably, missed their train).

  3. On the whole the notion that space on Cascades trains is preferentially available for riders traveling a greater distance than that covered by Sounder makes sense – hence higher fares on Cascades to discourage short distance riders from taking away seats from someone riding a longer distance, and not honoring commuter tickets.

    What I see RailPlus providing for Everett & Edmonds riders is a later departure home than the current last Sounder train at 5:35pm (if I read it correctly.) The same thing isn’t relevant on the route toward Tacoma, where Cascades skips most stops anyway, and the last Cascades train leaves in between two Sounder trains.

    As to the perfomance and cost of Sounder vs ST buses between Tacoma and Seattle, if there is available capacity on Sounder trains, it is economically nuts to run more expensive-to-operate buses with cheaper fares that are redundant to Sounder. The fares should be set at the same level between bus & train and buses should feed Sounder trains. And to preclude anyone from arguing that fully burdened Sounder costs are higher than ST buses – my assumption here is that existing Sounder trains have excess capacity so the marginal cost of those train seats is zero, and the marginal cost of operating rush hour buses which have to deadhead back empty or be parked in Seattle and the driver deadhead is decidedly not zero. Further the ST bus costs include zero charge for the freeway capacity.

    Lastly, there is a question of what should be the train fare between Seattle and Tacoma on Cascades. If the reservation system question above can be resolved, and if there are excess seats available (for example due to riders going Tacoma – Portland and intermediate points) then the fares between Seattle and Tacoma should certainly more competitive.

    1. the marginal cost of operating rush hour buses which have to deadhead back empty or be parked in Seattle and the driver deadhead is decidedly not zero.

      Rightly so. That’s why they run in revenue service in the reverse-peak direction.

      my assumption here is that existing Sounder trains have excess capacity so the marginal cost of those train seats is zero

      And by that logic, the marginal cost of excess capacity on ST Express is also zero, and we’re back to the difference of $6.26 (source).

      1. Actually, many of Metro’s peak-only routes do deadhead, the 218 for example. You have to get at least a few extra riders for it to be worth the hassle of putting the bus in service and making the stops.

      2. one of the wonky things about the way bus service is scheduled, is you have revenue hours and non revenue hours. i dont know the details behind it, but i do know some has to do with reporting to the feds. Its more expensive to run a bus in revenue service than non revenue, part of it is that theres a cost to provide the ADA service, if nessasary (Express and limited stop routes are exempt from this requiement as far as i know, which may be why metro has so many express routes). plus when you account for it you are accounting for that many more hours of service, when really the cost for operating the coach at the end of the day might not be that much more, its still a diffrent pool of service hours so to speak. which goes back to the feds and how your money is alloted and so on so forth… big accounting nightmare i’m sure.

      3. Bruce: Metro has tons and tons and tons of deadheads. It’s not really a fair comparison here, since Sounder’s “garages” are a few hundred feet from the terminals.
        And technically the two reverse trains on the South line are deadheads, but ST correctly made the assumption that they can monetize the deadhead and provide service to a few hundred.

        Another problem with turning deadheading buses into revenue routes (at least in Metro’s case) is that the bus arrives at the terminal at different times. For example, two afternoon trips on route 197 deadhead back to the U-District after reaching Twin Lakes P&R. You could run an afternoon Twin Lakes-UW express, but because it arrives at potentially different times every day it’s hard to get people to use it. But Metro does allow customers to ride deadheading coaches (with coaches headed in to North Base as an exception).

      4. Tim, I don’t know what data you are looking at, but if you look at the 590 series buses from Tacoma/Lakewood to Seattle, between 5 and 9 am there are 70 trips scheduled northbound on the five routes combined, and approximately 9 trips scheduled southbound. So that’s 60 trips which are either deadheading, or where the bus remains in Seattle and the driver has to deadhead. Plus with the unpredictability of I-5, ST would have to schedule significant recovery time to provide reliable service, whereby the Sounder reverse trips reverse in 10-20 minutes.

        Operating 60 one-way trips incurs significant marginal costs.

      5. They don’t all continue as the same thing. Look at all routes in the 590 series. Look also at route 577 and 578.

        If you want to do a trip-by-trip analysis, look at the trip details page on One Bus Away and look down at the bottom for “continues as”.

      6. Tim, 577/578 for the most part is a 2-way route so the buses turn around. The 590 series for the most part is a peak direction route with headways as short as 2 minutes in the peak direction, and with a small amount of reverse service. There aren’t 60 extra sounthbound trips. The 60 extra northbound trips either deadhead back or lay over in Seattle for the evening commute, and it is expensive, inefficient service to provide – just like the Kent & Auburn peak-only express buses.

      7. Carl, look at the data. Most 577s turn around and become 577s. Some 590s turn around and become 578s. During peak, 578 only operates in the non-peak direction.

      8. Tim, here is the data.

        In the morning peak 14 northbound 577 trips arrive in downtown Seattle between 532am and 929am. There are no northbound 578 trips in the morning peak.

        The first 8 of the 14 trips turn around into southbound 577 trips. The other 6 do not and either deadhead home or are stored. There are also 8 route 578 southbound trips departing between 5am and 9pm, the first of which deadheads into Seattle because it leaves at 5am before any northbound bus has arrived. So 7 possible northbound route 590/592/593/594/595 buses return in revenue service as 578, and the other 54 either deadhead or are stored. (70 northbound trips, 9 southbound 590/594, 7 southbound 578.)

        77% of route 590/592/593/594/595 either deadhead empty or are stored, and 43% of the 577 trips also.

    2. “On the whole the notion that space on Cascades trains is preferentially available for riders traveling a greater distance than that covered by Sounder makes sense – hence higher fares on Cascades to discourage short distance riders from taking away seats from someone riding a longer distance, and not honoring commuter tickets.

      What I see RailPlus providing for Everett & Edmonds riders is a later departure home than the current last Sounder train at 5:35pm (if I read it correctly.) The same thing isn’t relevant on the route toward Tacoma, where Cascades skips most stops anyway, and the last Cascades train leaves in between two Sounder trains.”

      Then explain the deal with MTA Maryland’s Penn line and Amtrak’s Northeast Regional trains. The Penn Line runs all day, both directions and later than the certian trains on the NE Regional that are availble for commute to and from Baltimore/DC and Perryville. The trip itself is shorter than that of the South Sounder.

  4. KC metro and CT have an annoying habit of running peak-period-peak-direction-only express buses that essentially duplicate service already provided by the Sounder trains.

    For example, KC metro routes:
    152, 157, 158, 159, 162

    And CT routes:
    405, 406, 417

    While a few of these routes do provide a couple of trips for those that missed the last Sounder train, most of these routes simply provide redundant service between downtown and the same stations that are already served by the trains. If KC metro and CT are serious about balancing budgets, routes like these need to go away.

    1. Does sounder have enough capasity to handle the extra riders? some of them serve stops that sounder dosent, how do you handle that (Auburn P&R for example). Just food for thought. P.S. i do agree there needs to be some condolidation inbetween ST Express and Sounder service and duplicate service offered even by ST and the local agencies. Again, it goes back to how do you address the capasity issues, stop location issues, trip leingth, and service quality, etc presented by combining routes is a whole topic itself.

      1. In the case of the 152–not only Auburn P&R, but stops in the CBD and along the SODO Busway. Unless your office is really close to King Street Station, it is, in some cases, a wash between hopping on the next bus to King Street and riding one of the routes that operates parallel to the Sounder.
        Also, from the bus you get to point and laugh when BNSF runs freight interference. And the Sounderites get to point and laugh when a peanut butter factory catches fire.

      2. Pretty much every Metro Auburn-downtown Seattle and Kent-downtown Seattle (except route 150) is a peak-only express which has to deadhead somewhere, and many of which can only make a single trip in a shift due to the length of the route. If Sounder has capacity, it would substantially more cost effective for Metro to have a local route that picks up the riders and drops them at the Sounder station. 1-2 buses can make multiple trips.

        While some riders may prefer a 1-seat bus ride, it may be a diseconomic luxury. There is plenty of capacity on buses at 4th & Jackson or buses & Link at International District that can distribute riders downtown (some of these buses are empty at the start of their routes such as 255 and 545).

        If Metro doesn’t get funding to maintain service, these redundant buses should be the first to go, or they should charge a substantial peak express fare (in New York the subway is a flat $2.25, the express buses from the Bronx and Staten Island cost $5.50.)

        One thing that currently impedes making the change is that the Metro bus fares are cheaper than Sounder. It would make planning the system so much simpler if fares were aligned between common points regardless of mode or operator, e.g. Kent – Seattle fare is the same on Sounder, ST Express, Link, and Metro bus or any combination of transfers.

    2. Not to mention that the 416 and 417 are handy when there are mudslides. In addition, these bus routes cover other areas besides the stations.

      1. Running extra buses that duplicate the Sounder every day, just for the 10 times a year or so that there are mudslides sounds like a significant waste. Much cheaper would be to only run those buses on days when a mudslide actually cancels Sounder service. Or, if you can get frequent, reliable, and fast shuttles to Edmonds and Multilteo from Lynnwood with timed connections (assuming sufficient capacity on the 511), maybe for the few times a year that mudslides actually cancel the Sounder, that’s good enough.

        While in an ideal world of unlimited money, running duplicate service to/from downtown has some value, the reality is that money is finite and CT has major budget problems that will require some serious service cuts to resolve, on top of what they already did last year. Routes like 416 and 417 add very little to the value of the network and are very expensive routes to operate, as each trip requires a long drive in unpredictable traffic, plus some 20 miles or so of deadheading.

        To illustrate these invisible deadheading costs, if you’ve ever driven on I-5 northbound just north of 520 on a weekday afternoon and looked at the traffic going the other way, you will typically see the southbound lanes being stop-and-go for miles. Every minute or two, you will see a CT-labeled bus heading south, waiting in that long line of cars. As CT doesn’t operate a single route that’s in service heading south at that time, this means every single one of those buses are deadheading, carrying no one.

        Even though these buses are not carrying passengers, they are still burning fuel and the driver is still getting paid the same about as if the bus were in service. Furthermore, if these buses are to provide reliable service for customers heading in the peak direction, they need to have a long scheduled layover downtown, so that even in the worst-case traffic scenario getting to downtown, customers heading home will still see their bus arrive on time. This further increases the cost (at least the driver pay portion of it), as, even when traffic heading into downtown is light, whatever time gets saved on the road is spent in layover, and the driver is still getting paid for it.

        The bottom line – buses that duplicate the Sounder are an expensive luxury that we can no longer afford to provide, if we want to preserve the important parts of our bus service through this revenue crisis.

      2. If Metro has trouble setting up shadow service when Link fails, and then getting the word out, I have a hard time seeing CT doing any better. If shadow service isn’t built in, will people actually risk taking Sounder?

        That said, I’ll be curious to see how full the various CT runs headed downtown are. If the shadow runs merely head to Lynnwood TC or Mountlake Terrace TC, would that overwhelm the other buses?

        And are there good bus connections to get to the Amtrak reverse-peak-direction runs? or for any Cascade runs that stop at Edmonds or Mukilteo in the future?

        I experienced the trouble with re-routes as a strategy yesterday when I was trying to catch a southbound bus on 1st Ave S yesterday. The Mariners game was apparently coming to an end, but I didn’t know when. So, I made a beeline for Stadium Station and didn’t try to guess whether the bus was coming on 1st or 4th. The brochures and online schedules really need to be specific about which runs will be on re-route. (Though, ideally, the TMP would designate Edgar Martinez Dr and 1st Ave S to need transit lanes.) Likewise, Sounder commuters need to know Plan B ahead of time, or Sounder will lose ridership.

        As a cheaper alternative, maybe have a pair of (low-seniority) drivers each at Mukilteo and Edmonds Stations, and they end up driving only if the mudslide alert is on.

    3. I think you mean the 416 instead of the 405/406. The 405/406 are actually up at the top of the hill at Edmonds P&R, near Hwy 99, not down in the bowl. For people who park-and-ride, both are options but I wouldn’t call them competing service any more than a bus from Lynnwood P&R is competing with the Sounder.

  5. Perhaps related, but I think you can’t buy Multi-Ride passes from Amtrak between Seattle and Tacoma either.

  6. If the passes are good on Amtrak, wouldn’t an Amtrak ticket between Seattle and Edmonds or Everett be good on Sounder?

    1. When the Empire Builder is running late, and people who have taken a Cascades train with a connection to Everett or Edmonds on the Empire Builder, AND if it’s M-F, Sounder will honor their ticket.

      I suppose the payment question is the same sticky one Metro-Pierce-Community-Sound Transit have.

      We would need someone at the Amtrak-Sound Transit negotiator level to comment on this.

  7. Even with Amtrak’s definition of sold out, it seems conceivable that we could make RailPlus give users the right to stand on sold out trains. If there are ghost seats, RailPlus users can take them. If not, well, an hour is a long time for most people to stand, but it’s not totally crazy if there’s something to lean against — people do that all the time on Metro North in NYC and occasionally on Caltrain in the Bay Area.

  8. It’s pretty unfortunate that we have all these different local government agencies providing transportation options in the same corridor and they can’t seem to work together to consolidate pricing / ticketing / scheduling.

    1. The ticketing owner for Amtrak is federal. We’d have a much better chance of fixing it if they were local!

  9. I used to use Rail Plus right after it started up, this would have been back around 2004-ish? I found it handy if I just missed the last Sounder, because then I could just catch the 5:30 Amtrak instead. But I found that unlike the Sounder, the Amtrak was often delayed, which made it not as good a commute. The beauty of Sounder, of course, is that it’s not subject to traffic problems on I-5 like the buses are. So if the train is late all the time, it’s not really that much better than the bus, is it?

    Also, isn’t the 516 running later than it used to? 5:30 was super convenient, but 6:50 is a bit late unless you’re staying downtown for happy hour or something. There just aren’t as many commuters trying to leave downtown at almost 7PM as there are at 5:30. That could account for the lack of Rail Plus users.

    One other thing about Rail Plus: one of Senator Murray’s receptionists used to ride the Sounder with me, and I wonder if she might have been the one who suggested that the senator work on making Amtrak and Sound Transit work together on this? I know she was involved in brokering that deal, or at least, she took credit for being involved at the time it was announced. I have no idea if that’s true, but I like to think that senators listen to their receptionists. It’s the idealist in me. :)

  10. You’ve pointed out the problems: timing, connections, but there’s also the confusion as to how it works. I tried to understand it once, but gave up!

    1. It’s simple!

      If you have a MONTHY PASS good on SOUNDER, (you have to know how your ORCA card is loaded), and you want to travel between Seattle, Edmonds, and Everett (Amtrak does not stop in Mukilteo), you go to a Sounder TVM, choose your destination station, then choose RAIL-PLUS.

      If you have the correct pass, the machine will issue a ticke as shown above in the picture.

      The E-purse does not allow upgrades, and regular Sounder tickets are not accepted.

Comments are closed.