In San Francisco, a Glimpse of the Future of Transit

What if transit agencies were as nimble in providing service as riders are in using it?

I recently moved to San Francisco, and I was lucky enough to find an apartment that’s only a few short miles from work. As a lifelong public transport commuter, I made a point of checking out the transit options.

It didn’t look great; they all involved bus-to-bus transfers.  All the literature suggests that people really hate transfers, especially for services like buses that tend to have longer and less reliable headways.

So I turned to apps. The real time information powering Citymapper, Transit App, and Moovit is not new, but I didn’t have much use for it when I lived in New York. My daily commute was so obvious I didn’t need additional information – I walked to the A train.

In San Francisco, things changed. There are dozens of different combinations that might prove to be my fastest route on any given day (see below). Which one is faster depends on a long list of variables – traffic congestion, the different wait times for express or local service – and, critically, the chances that my dropoff time from one route is convenient to my pickup on another. This last point is critical, because knowing it in advance directly eliminates one of the issues with bus transfers: not knowing how long you’ll wait for the second bus.

Screen Shot 2017-05-04 at 6.36.32 PM.png

That’s important, because I have 3 or 4 different options for the first bus I board in the morning (for those in San Francisco – the 1, 1BX, 38, 38R, 2, and 33). With real time information in hand, I can hop on whatever first bus leg will get me to work quicker, including knowledge about the future location of the second bus. After experiencing it for the first time, it’s clear that real time information and predictive software is already helping to make transit more attractive, even in places where things like bus-to-bus transfers are required.

But at a more profound level, it’s highlighting changes to transit operations that may be on the horizon. In San Francisco, I no longer really care about routes in any conventional sense. I don’t care about a schedule, I don’t really care about ‘stops’, and I don’t really care where one bus drops me off or another picks me up. As long as those things can be reliably conveyed to me when and where I need them, they don’t need to be fixed in time or space.

In one important sense, bus routes are a way to communicate information about service to riders: “Stand here, get this bus, go there”. That makes sense when you don’t have a more dynamic way to communicate, which we didn’t when we invented bus routes 200 years ago. It’s obviously also a logical structure in places where you have or need dedicated infrastructure, as streetcars once did and as my old subway commute in New York City still does.

Clearly, though, routes have had a lot of staying power even beyond those places. Despite enormous urban development changes, most of the bus routes in major cities still look a lot like the streetcar routes that they replaced almost 100 years ago.

Now, sophisticated riders are able to hop between routes, knitting together new kinds of service with real time information. But that’s only a one-way flow – riders are getting smarter.

But what if transit agencies were as nimble in adapting service? Could buses decide to skip stops or turn local into express based on how many riders were already in the vehicle and where they were going? How about dynamically rerouting buses to avoid traffic congestion? What about deciding on the number vehicles in service based on real time demand?

Before you can even start thinking about that kind of dynamic service, agencies need to overcome one obvious hangup: a lack of fine grained data on their own riders. Riders are getting a lot of information from agencies via new APIs and apps, but not much is flowing back the other way.

That’s one reason I was excited to read about a new MBTA partnership with Transit App, which seems to be bridging this gap and providing user information back to the agency. That’s an exciting first step. If agencies are going to get more nimble and more able compete with the private car (let alone the private autonomous car), much more of that will be required. Without fine grained information on how its own riders are using their service, I can’t see how public transit agencies are able to thrive in a future marketplace of on-demand options.

Just to point out an obvious constraint to changing service patterns – you can’t start skipping stops and dispensing with the idea of routes until everyone is using a device that can convey information in real time (a smartphone, or something like it). Today, I suspect the number of daily bus riders using an app on a regular basis is still quite low.

And another disclaimer – in places where buses are already fast, frequent, and full, dynamic service changes probably won’t produce all that much benefit. But that’s a small slice of today’s transportation picture. Only about 2% of travel in the US is completed on public transportation. I think there are a lot more places that would benefit from nimbler bus service, and that’s a good thing for the potential growth of shared rides in the future.

All of these concepts are not wholly new – the International Transport Forum has modeled the most efficient possible public transport structure for Lisbon. While it kept the metro in place, it eliminated traditional fixed route bus service and replaced it with on-demand minibuses and shared taxis. A recent report by ITDP and UC Davis argued that transit will have to evolve and use smaller, on-demand vehicles where they make sense, in order to meet our long term climate objectives.

But what’s changed for me since I moved to San Francisco is that my commute is starting to give me just a hint of what that future might look. It’s exciting to see up close.

Montreal to NYC by Bus and Rail

The trip from Montreal to New York is surprisingly inconvenient. I had time to ponder potential improvements during a recent 12 hour train ride home. Here’s what I came up with.

I’m from Montreal but live in New York City, so I frequently make the trip between the two. For those who haven’t had the privilege, there are currently two ways to make the trip for those who can’t afford a flight or don’t want to drive: the bus, and the train. Both could use an upgrade.

First, the train. The train’s scheduled run time is about 11 hours, although my anecdotal experience suggests that an additional hour is not an uncommon delay. That means the trip is frequently as long as 12 hours, almost double the length of the driving alternative. The other struggle is frequency – there’s only 1 train per day, leaving in the morning and arriving in the evening.

Why is the train so unimaginably slow? There are two basic reasons. One is that the track north of Albany is slow and single tracked. New York to Albany is nearly half the mileage of the whole trip but takes only 2 hours and 20 minutes at its fastest. The stretch from Albany to Montreal takes 8 to 9 hours, despite covering roughly the same distance.

The bus is quicker, and completes the whole trip in a scheduled time of about 8 hours at its fastest. There are two reasons the bus is slower than a car – it makes a mysteriously long (30-45 minutes) service stop in Albany, and it takes longer to cross the border than a passenger car typically does. As with anything that travels on roads, it can also suffer from significant delays getting into and out of the congested core of Manhattan.

Is there a way to improve the two options to make them more competitive with the car? Although some have proposed a multi-billion dollar HSR program, there’s a much quicker approach that would yield enormous benefits for both riders and transport operators at minimal cost.  See the table below for a summary of the current situation.

Road Distance Train Travel Time Bus Travel Time
New York – Albany 150 miles 2h 20m 3h+ 
Albany – Montreal  220 miles 8h 30m 4h 30m+

The train is quickest south of Albany. It also provides the additional benefit of a reliable and congestion-free approach to Manhattan’s urban core. As an added bonus, it also has spectacular views of the Hudson for its entire length. The bus wins by an enormous time margin north of Albany. Why not combine the two to get the best of both?


The proposal is simple – passengers at Albany would transfer to a bus to complete the trip to Montreal (or vice versa) and shave about 4-5 hours off of of today’s train journey. Transfers between modes can be a pain, but they don’t have to be. The Albany train station could be configured to make the bus transfer quick and easy, cross ticketing (ie a single ticket for bus and rail) would make it seamless, and the buses could be timed to depart immediately after train arrivals. The best case scenario could get the total travel time below 7 hours and get travel time roughly to parity with the private car, with the added benefit of a congestion free approach to Manhattan’s Penn Station.

Aside from the clear passenger travel time benefits, there would be lots of operational gains for both bus and rail services to this change. As of right now, one of the train’s sources of delay is that it swaps out an electric locomotive for a diesel locomotive at Albany to complete the trip north to Montreal. This typically takes as long as 30 minutes. If the train ran only between Albany and New York, it could stay exclusively on electric power and gain back a large number of wasted man hours.

Even more beneficial, the two train sets currently providing 1 daily round trip on the 12-hour Montreal to New York segment could run 4 round trips on the the 3 hour trip from Albany to New York. Just by turning the trains back at Albany, you’d get 4 times more service over the Albany to NYC line using the same train equipment and crew. That’s a significant increase to the 13 daily round trips currently being offered on that line. 

Much like turning back trains at Albany improves Amtrak’s efficiency, preventing buses from entering New York City would makes buses more efficient. The traffic around New York City is severely detrimental for the operating efficiency of buses. Aside from making bus travel slower on average (which reduces the number of passenger miles a bus can serve in a given hour), congestion is unpredictable. That unpredictability requires bus operators to build additional slack into the bus schedule for buses heading into New York City. Keeping buses on the relatively uncongested Albany-Montreal branch would allow operators to tighten up schedules and run more frequent service at the same cost.

I used the word ‘operators’ deliberately – there’s no reason that multiple bus operators shouldn’t be encouraged to run the new Montreal-Albany route. Given that the goal here would be to attract choice riders, bus operators could compete to offer the best service – luxury buses, large seats with more legroom, etc. 

Even for those who believe that enhanced rail infrastructure is the ultimate solution, this intermediate and flexible step could help build ridership for mass ground transportation (ie, alternatives to air and private car travel) over time, helping make the justification for further investment more obvious.

Another advantage of creating the bus/rail combination is that service levels could be dramatically improved and tailored to demand.  Amtrak currently offers 1 train per day – it’s sometimes sold out at holidays, but it’s likely not terribly well utilized at most other times of year.  With the new model, there would now be 17 trains per day running between Albany and New York, and each could be combined with a bus leg to Montreal. This list of 17 departures would create a whole new realm of options. As just one example, a bus operator could launch a service from Montreal at 1AM to catch the 5:05AM train from Albany to NYC that arrives in NYC at 7:30AM.

There’s one other improvement that would make any option dramatically better: a dedicated border crossing at the Montreal bus station. There’s precedent for this – airline passengers already clear customs at the Montreal airport, even when departing for the United States. Clearing customs at the bus station (rather than the border crossing) would allow for much more certainty in bus travel times (by avoiding an unpredictable border), allowing for further tightening and cost efficiency for bus operators and better experience for riders. I’m sure there are 100s of obstacles to getting that done, but it’s a no brainer from a transport operator and rider experience perspective.

So what would all this cost? I don’t know, but infinitely less than a rail upgrade would cost to get anywhere near the travel time benefit described above. There are definitely obstacles to getting this done, but I think there’s something to the idea of a combination bus-rail alternative between NYC and Montreal.