Rounding out my series on microtransit and demand-responsive transit (DRT):

Key papers and definitions

I eventually stumbled upon this survey, Errico et al., 15 pages which explained some things better than Koffman (2004), 54 pages, which was the classic paper everyone cited but which I found didn't really clarify some of my question. Also it's shorter. In particular, since it was going for a unifying framework, Errico was happy to point out similarities between different modes, confirming my intuition that "this thing seems like this other thing with a small modification."

Like we got a better definition of zone route:

"Vehicles operate in a demand responsive mode and cover a given zone. Usually, those systems have one or two scheduled stops, corresponding to the departure and the end of the line. Departure and end of the line can be placed at the same location. The difference with Point Deviation is the number of scheduled stops, since the latter usually has other stops besides the two terminals"

And

"the Request Stops system appears to be the closest to fixed-line transit, while Zone Route is the closest to DAR [Dial A Ride], at the other extreme of this range."

"Request Stops can be seen as a Route Deviation where only a few deviations are possible. Zone Route can be seen as a particular case of Point Deviation where only one or two scheduled stops are present. Also, Zone Route is similar to a Demand-Responsive Connector, in case the scheduled stops are transfer points to traditional transit. Route deviation can be seen as a point deviation with a more restricted operation policy, and so on."

I think I am still lacking a concrete example of Zone Route, but I'm imagining something like DAR, except that every hour it visits a mall or train station or such, so if you're at that location you know you can wait and hop on.

Time Windows

I found Errico most valuable and accessible for the first part, giving clarifying definitions. Next part was their unifying model, which might be interesting; final part was reviewing various technical papers, which I had no interest in. Their model did propose using time windows of non-zero width, with zero width being conventional scheduled departure times. The idea is that the window defines earliest and latest departure times, with the bus being allowed to leave the stop at any time within the window.

So say the window is 10:00-10:05. If the bus arrives earlier than 10, perhaps because no deviations have happened, it waits, and you can trust that getting to the stop by 10:00 will let you catch it. Conversely, if you're already on the bus and wanting to get off, you can trust that you won't get there later than 10:05. And the window allows some shuffling of time between segments as needed; if the first segment has no deviations while the next has more deviations than usual, you can leave the stop at 10:00 to use the extra 5 minutes.

Koffman was interesting too, if deficient in some definitions; it does cover several real world examples. I noticed that a common theme was ADA avoidance, i.e. the regulatory requirement to provide ADA paratransit along fixed-route bus service is a significant disincentive to providing fixed-route service at all.

The Winnipeg transit connector caught my eye. Going toward the station, it picks people up at their houses, per prior request; going away, it simplifies things by dropping people off at various defined stops. The idea being that people shouldn't have to wait for the shuttle in winter, but can walk home after being dropped off. Also neat was the low-tech planner: a printed map of the area and its stops; a boarding passenger says where they want to be dropped off, the driver circles the stop, and figures out a route before leaving the station.

Merrill-Go-Round

I also found this detailed description of the Merril-Go-Round (PDF), a flexible transit system that lasted some 30 years. The paper is from the first year of service but I think it was stable for much of the period. It's allegedly a point-deviation system, but I feel it has more in common with route-deviation ones; the history of the service suggests a political allergy to the 'route' label due to failed marketing. There were 10 scheduled checkpoints, half a mile apart; two buses and 30 minute headways, which is rather amazing, implying just 3 minutes to get between stops including any deviations. Deviation did have a surcharge: 25 cents basic fare, 40 cents for one-door service, 50 cents for two-door (e.g. deviating to pick you up and drop you off.) A neat twist was that the surcharge was only paid once, so if the shuttle came to pick up a group of 3 people, they would pay 3x25 + 15 cents, not 3x40 cents. In that first year it was getting 6-20 riders/hour, which is very good for DRT.

At some point state funding cuts dropped it to hourly service, and covid or something turned it into pure DAR.

Other Resources

In the unlikely event that you want to learn even more, I've got you covered. A 2010, 97 page guide to DRT, giving statistics, design criteria, and more real-world examples. One thing it noted was that flexible service is most viable with time-insensitive trips, i.e. not so good for getting to work or school regularly.

And then a 2019, 200 page State of the Practice, which had one set of numbers I saved: Denver's RTD (which seemed a star of urban flexible transit for a while), wants 25 boardings per vehicle-hours to justify fixed bus service every 15 minutes, 40 boardings for 10 minute service.