mindstalk | Why microtransit sucks (Reply)

A long post here (2200+ words), but I don’t see a natural division point. Also, apparently this journal style doesn't display lists nicely. I may need a new style, sigh. (No, actually it's DW adding p tags to the li tags.)

First I have to define microtransit. A narrow definition could be “door to door, shared rides, summoned by mobile app, coming quickly, operated or subsidized by a transit agency to cost riders roughly transit fares.” A broad definition could be simply “door to door, shared rides.” Without the shared rides it’s just a taxi, without flexible door to door service you don’t have the appeal.

And the idea is appealing. Boosters suggest it will have almost the convenience of taxis, while having the cost-sharing of buses. Flexible shared rides have a long history of being tried, and a more recent plethora of mobile-enabled microtransit pilot projects… which, allegedly, all fail. Ridership struggles to get above 2 riders per vehicle-hour, when even coverage-oriented suburban buses average 10. (That’s 10 boardings per hour, not necessarily 10 riders present at once.) And costs per trip are high, $20+ or more.

Is this inevitable? I think it is. Imagine a simple case: n riders are picked up, then dropped off in some random order (this avoids the complexity of dropping off some riders and picking up new ones.) I haven’t done a perfect proof, but I’m pretty sure that on average there will be n-1 detours before you get off, with a worst case of 2n-2. If n=2 and detours take 5 minutes, that’s not so bad. If detours are 15 minutes that’s worse, especially since 1/4 of the time you get 2 detours.

(What do I mean by detour? Deviation from the shortest-time trip that would meet your needs, e.g. a standard drive or taxi ride.)

And we need to ask, why are there 2 passengers? When you pulled out your phone to request a ride from A to B, what happened? There are 4 possibilities:

no match, and you get a subsidized taxi ride.
the system makes you wait a long time before it finds an appropriate match. That’s not very responsive.
the system matches you quickly with someone, at the cost of a long detour to pick them up or drop them off. That’s slow.
the system quickly matches you to someone closely compatible with your trip.

The last case sounds ideal. But suppose you had been delayed, and actually requested a ride 20 minutes later. Would you still have been matched with someone? Would your prior ‘partner’ have been matched with someone else? If there is such a reliable supply of people making compatible trips, wouldn’t it likely be simpler to just have a bus?

And what if the supply of riders increases, so you have 3 ride requests in short order? At 2 riders per car, there’s little economy of scale; the agency needs to pay for more drivers and vehicles. Also, you’re splitting the driver’s expenses among just 2 riders. By contrast, even a small bus can easily carry 30 people – lots of cost-sharing and spare capacity.

“Ah, just have bigger microtransit vehicles!” But now n=3, and you have 2 detours on your way. Average 10 minutes, max 20, if it’s still 5 minutes per detour.

At n=6, minivan size, there are now 5 detours on average, with 25-50 minutes of delay. This is rapidly getting less appealing.

And again, this was a simple model of picking up people then dropping them off. It would be more efficient for the agency to allow picking up replacement riders, but that leads to even more detours for the current riders.

Are there forms of flexible transit that do work? Yes – and they show what’s wrong with microtransit.

School buses: this is probably the one most familiar to my readers. In elementary school, a bus came to my door, as it did for other students, before finally heading off to school. Note:

it condenses one end of trips, as everyone’s going to the same school; no “drop-off detours” in the morning. Likewise it is not door to door service, but door to school.
the simple case does apply; kids are only picked up.
despite those advantages, it can be rather slow, with the time from pickup to school being rather longer than a drive would be, unless you’re the last student.
But it does transport a lot of kids per adult, and lets parents off the hook, so, yay.

In 7th and 8th grade (ages 12-13) my new school still had school buses, but told us to walk to a neighborhood stop where it could pick 3 of us up at once. Presumably that saved a fair bit of time, though I lack the memories to evaluate that. (Also I was going to a much further school than before, so it would have taken longer anyway.)

In high school they just had us take public transit, subway and all.

Airport shuttles: There’s another form that we used to have in the USA: airport shuttles, especially SuperShuttle in Los Angeles. You would go from your plane to a shuttle stop, waiting not just for a SuperShuttle van but one going in your general direction (Downtown, Glendale, Long Beach, Santa Monica, etc.) The shuttle would leave if filled quickly, or might circle up to 3 times hoping for more passengers. Then it would head out, going to each person’s address. So even with you all going in the direction of Glendale, some major detours might happen before you got home, but at least there was no risk of someone new being picked up along the way. There were discounts for hotel destinations (so even if you weren’t staying at a hotel, you might use one as your stop to save $5-10.) Going to the airport, you would call the company the day before, tell them your flight time, and be told a pickup window, which might be fairly early to account for people they’d be picking up after you.

Again, it was not door to door service, but door to airport service – one end is anchored in a major source/sink of trips.
It could be annoying and add a bunch of time to your trip, especially if you lived further out and there were a bunch of other people, or there weren’t other people and you circled the terminals waiting for more.
It was cheaper than taxis, but not cheap. Around $20+ in the 1990s, from LAX to the Caltech “hotel”, high 20s or low 30s for houses. In the mid 2010s, high 30s to a Glendale house. Oh, and 10-20% tip was expected on top of that.
The main thing that made it work was that after spending hundreds of dollars on a plane ticket, and multiple hours in your flight, the extra time and money of the shuttle didn’t seem so bad. And you did get taken to your door – important with luggage – and anyway, public transit alternatives didn’t even exist then. (And still aren’t good.) The lack of transit also meant that there really was a good supply of passengers; whereas in Chicago, a lot of people would simply take the train that ran right into the airport.
I use past tense because by late 2019, SuperShuttle finally surrendered to Uber/Lyft stealing its lunch. Which is a pity because SuperShuttle had been a sustainable business since 1983… unlike Uber and Lyft, which are finally and barely profitable as of the last year. But when VC money let them offer comparable prices for direct service (or at least less-shared, a la Lyft Line), naturally everyone would take them.

Notice the similarity with school buses: the model works, but one end is anchored in a major destination rather than being door to door, and it’s still annoyingly slow and expensive.

Share taxi, a very common pattern in the developing world. Wikipedia lists many examples in brief, but I’ll describe the colectivos I saw in La Serena, Chile: cabs with a designated destination would pick up people downtown, and head off to the destination, letting people off along the way, possibly making limited detours for them. (I did see the colectivos myself, but my friends told me what they’d been told of the operation; none of us ever took one.) Then, presumably, making its way back, and getting flagged down along the way by people wanting to go in. You can imagine variations like the destination being a big source of passengers itself, or maybe people calling in from phone and a cab being dispatched mildly to pick them up.

But note that this is not a door to door service. In the narrow mode it’s simply a bus, except that these often aren’t scheduled, but wait until they’re full before leaving (obviously a big risk of delay, there.) Even in a “detour to drop off”, mode, one end of the route is anchored in a major source of passengers. If you want to go between arbitrary points, either you get an actual private taxi, or you take a colectivo in then out again, with all the joys of hub-and-spoke transfers. I would also guess that if detours are made, they are at the judgement of the driver, so that if you ask him to 10 blocks off the main route he says “no, you can get off here.”

Paratransit: Here we return to my previous post. Conventional paratransit (when it works as promised) is microtransit, in my broad sense, though it sacrifices the rapid response. With an allowable one hour leeway on either side of your request to schedule a trip, including a 20-minute arrival window for pickup, it is basically case 2 from my list above: you request a ride, and the system allows itself a lot of time to try to match you with someone. Except it’s also case 3, as even the match may involve a long detour – and probably does, since not that many people are paratransit dependent, and the people who are don’t make that many trips (in large part because it’s terrible), so the chance of a close match is pretty low. You might even get case 1, basically a taxi ride, which is great for you, but not for the agency’s cost per ride.

Of course, here the special condition is that some people literally cannot take the regular bus, so we are (in theory) committed to providing an alternative. But I think the choice of sucky and expensive service, or nice and really expensive service, is inherent. If the agency commits to attempting ride efficiency, that has to mean wide matching windows and long detours, and still not having many riders at once. If the agency commits to a really good experience, that basically means disability-friendly taxi service. In the extreme case, issue everyone their own chauffeur. In a less extreme case, have enough drivers on hand to meet typical peak demand.

Except, that demand isn’t fixed. If you look at current paratransit use, and hire enough drivers to make that nicer and more responsive, then people will want more rides, necessitating more drivers.

Uber and Lyft: if microtransit is so terrible, why do Uber and Lyft keep offering shared ride options: UberPool, Lyft Line, Uber Shared, Lyft Shared? Well:

from a cynical PR perspective, their main function may be simply to exist and deflect criticism, even if never used; the companies can say “look! we’re trying to be efficient!”
they’re opportunistic. Probably most rides aren’t even requested as shared. Even if you do request a share, the system spends maybe a couple minutes trying to find someone with an acceptable (8 minutes, by one Uber Shared page) detour, and then if it can’t then you just get a regular ride.
currently it needn’t cost them much to make the offer. Looks like current Uber pricing is to start with a tiny discount, and then if it finds a match, you can get up to 20% off. Which means Uber would end up collecting 1.6x the revenue for what is almost a single trip.
and I'd bet that many if not most actually shared rides are due to trip concentrators, like heading downtown or to a big game.

Pilot programs: if microtransit is so terrible, why do agencies keep trying? I dunno. Executives easily seduced by tech companies promising that they can solve geometry and economics with an app? Executives liking their Uber experience, and generously wanting to share that with the public, without thinking things through? Executives desperate for a gimmick to make transit sexier? Executives desperate for a way to claim coverage for a low density area, without the expense of actual bus service? (That might be the one use case for narrow microtransit: a half-bluffing, last-resort service, that gets replaced by a bus ASAP if people start actually using it seriously.)

Finally, some quotes:

A single new rider risks wrecking your schedule: “If I’m the passenger, the detour works if the driver is detouring for me, but not if the driver is detouring for someone else in the car.”

And: “I think we underestimate the degree to which the attraction of fixed route service is its predictability. All technology-driven ways to make it less predictable may be, on balance, a net negative for customer utility.”

Or: “Deviating to pick up one more passenger wastes the time of all the passengers already aboard the vehicle”

And: “almost all of the arguments for micro-transit ignore this idea, and almost immediately go into the benefits (e. g. front door service) without any of the costs (delays caused by front door service).”