Microtransit Doesn’t Scale

How to minimize damage from predictable failures

In our last issue, I promised to examine how driving automation might transform transit operations, a topic my co-authors and I spend a good deal of time on in our forthcoming book The End of Driving. One area where we foresee a significant impact is on microtransit. To make that argument, though, a clear baseline is necessary.

So today, let's talk about microtransit, why it generated (and generates) so much excitement, and—to lay my cards on the table—why I think it can't succeed with human drivers.

Some readers may be asking themselves, just what is microtransit?

Let me explain. You may remember that, about a decade ago, everyone was excited by the success of Uber. The question on every entrepreneur's lips was “what if Uber, but for X”? So: what if Uber, but for transit?

It was an obvious question to ask, because traditional fixed-route buses are clearly inefficient.

To be cost-effective, buses need to carry many passengers, so that many fares can offset the costs of running the bus. But passengers all have different origins and destinations. This means that to the extent that a bus tries to maximize the convenience of some riders, it is also maximizing the inconvenience of others, and discouraging the latter group from riding. (A transport planner would call this the efficiency versus coverage trade-off.) The best way to solve this is to run ‘main line’ buses along the densest corridors often, and ‘feeder’ buses along less-dense corridors occasionally. This approach maximizes both revenue and coverage, but the trade-off is a fundamental mismatch between where and when people actually want to travel versus where and when buses can economically operate. Many potential riders find themselves too far from bus stops, travelling at inconvenient times, or heading to destinations that require multiple transfers.

Meanwhile, the taxi model—door-to-door service that goes exactly where you want, exactly when you want—offers everything that fixed-route transit lacks. The problem, of course, is cost. Taxis are prohibitively expensive for most trips because each vehicle serves only one passenger or group at a time.

Microtransit purports to be an improvement on both models, one that combines the convenience of taxis with the efficiency of shared rides.

Instead of running big buses, a microtransit service runs shuttles, typically 6-to-15-seat vans. These operate within defined zones, pick up multiple passengers with similar destinations, and—using sophisticated routing algorithms—optimize their routes in real time. Instead of mostly-empty feeder buses and overcrowded main-line buses running on fixed schedules, microtransit offers flexible vehicles that go exactly where people want to travel when they want to go, while still sharing costs across multiple passengers.

So just like Uber or Lyft, a user summons a vehicle to their location via app, and it takes them where they want to go. Just like a shared Uber or Lyft, other people are in the vehicle, and the trip includes detours to pick up or drop off these other passengers. But just like public transit, the cost of the trip is capped, making it affordable for patrons.

The appeal of this model is obvious, particularly for transit agencies required by law to provide expensive paratransit service to elderly or disabled passengers. Rather than launching something wholly new, agencies could expand something they were already doing—giving some riders a pickup from where they were and taking them where they wanted to go—and do it better, engaging more riders while serving coverage mandates more efficiently. And riders should be happy, because they are finally getting everything they have always wanted from transit: maximum convenience for minimum cost.

Or that is the promise, at least. Unfortunately, the promise has not been kept.

A decade of real-world experience reveals a consistent pattern: microtransit is fundamentally inefficient. The thinker on transit we most admire at Changing Lanes, Jarrett Walker, warned early that flexible service faces geometric constraints no technology can overcome—“ridership is the death of flexible service”—and subsequent evidence has brutally validated his prediction. Unlike fixed routes where additional passengers represent nearly pure revenue gain, each new microtransit rider requires roughly proportional additional resources. This inescapable reality means flexible service will always be inefficient compared to fixed-route service.

However, that same decade of experience permits us to say that microtransit has some potential. It can work, but only in markets, or geographic areas, too sparsely-populated for conventional transit. Today’s successful microtransit operations integrate with existing operations, are sized to actual demand, and target specific gaps rather than attempting wholesale replacement of transit networks.

Rolling Boulders Uphill

Surveying the landscape, we find that most instances of microtransit in the wild were costly failures.

Los Angeles Metro's microtransit pilot ballooned to nearly eight times the per-passenger cost of low-performing fixed routes, at $47 per ride in FY 2023, though recent improvements brought this down to $29 per trip by early 2024. Sacramento's SmaRT Ride, after expanding to nine zones, was ultimately shuttered when costs hit $47 per ride. Park City, Utah cancelled its on-demand service after reaching the conclusion that fixed-route service would deliver the same benefits at $664,000 less per year.

But these are just anecdotes. Lukas Foljanty brings the goods: in his annual market analysis of microtransit services—the latest, for 2024, tracks 1,300+ active ones—shows that the average service carries just 2.3 passengers per vehicle hour, with only one-third achieving three or more. Given that an average of three is, as Foljanty notes, the minimum threshold for viable operations, that means that most active microtransit operations today aren't viable.

Even where microtransit reached those numbers, like Milton, Ontario’s Go Connect service, which had stabilized at 4.2 passengers per vehicle-hour, it was considered merely cost-neutral versus low-ridership fixed routes, not a superior alternative. (It’s relevant that Milton did not renew the pilot when it expired in 2016.) The performance threshold of four-plus boardings per vehicle-hour appears to be critical: cost-per-trip typically remains well above conventional bus service unless passenger density exceeds this level, a threshold most pilots never reach.

But why should this be so?

The Curse of Success

It’s not a mystery. Jarrett Walker laid it out in 2019: microtransit’s promise of high ridership is the thing that destroys its viability. Let's imagine that Exurbia, almost entirely made up of low-density suburban areas where fixed routes would perform poorly, decides to offer a flexible service. At first, it is a great success. Demand is low, so Exurbia's shuttle-van serves very few people, but it serves them very well. With minimal demand, the van meanders through residential streets, carrying perhaps two to three passengers per hour but delivering them exactly where they want to go across Exurbia's wide neighbourhoods. At this utilization level—just meeting the minimum viability threshold most practitioners consider necessary—costs per trip remain manageable.

But because they're so thrilled, word gets around, and ridership grows. Now the van faces an impossible geometry problem: if significant numbers of people want hourly service, it cannot physically reach every neighbourhood every hour. To maintain coverage and ridership satisfaction, Exurbia must add more vans.

That solves the immediate problem, but generates another: cost. Three vans serving the same area means triple the driver costs, triple the maintenance, and triple the fuel expenses. Since labour represents the largest transit operating cost—and each van requires its own driver regardless of passenger load—the per-passenger subsidy escalates dramatically. Whereas once, with modest ridership, Exurbia's cost per trip was manageable at around $20, comparable to low-performing fixed routes, adding vehicles to meet demand drives costs to $40+ per trip.

This is microtransit's core constraint: unlike fixed routes, where additional passengers add identical fare revenue but minimal marginal costs, each new microtransit passenger requires roughly the same resource investment as the first. A bus can accommodate five or thirty-five passengers with identical driver and vehicle costs. But microtransit's individualized service means serving thirty-five scattered passengers requires dramatically more resources than serving five.

That’s assuming, of course, that five are actually served, but microtransit’s inability to actually pool rides consistently is now evident. Montgomery County's official evaluation of its Ride On Flex service found that 90% of all trips carried only a single passenger, with an average of just 1.14 passengers per trip; barely above solo occupancy, sophisticated routing algorithms notwithstanding. Consequently, microtransit provides convenience to riders, but at significant cost to providers.

There is a prominent example of this dynamic not far from where I live. In 2017, the Ontario town of Innisfil launched Canada's first Uber-transit partnership. Rather than introduce a fixed-route service, the Town paid for residents’ ridehail trips on Uber within Innisfil, limiting the co-pay (the ‘fare’) to $3. This initially seemed not only innovative but also cost-effective. But it has been a victim of its success; subsidies ballooned from $100,000 in 2017 to $1.8 million by 2024, forcing fare increases and monthly ride caps. What began as a cheaper alternative to buses now costs more per trip than fixed-route service would have… a service that, not coincidentally, the Town is now considering.

You might wonder why agencies continue pursuing microtransit despite the fact that theory, and practice, both predict failure. The failure of flexible service in places as diverse as Sacramento, Berlin, and Innisfil is not secret. So why do microtransit offerings not only persist, but continue to emerge? It’s because institutional pressures, political incentives, and regulatory requirements are stronger than the evidence.

Transit agencies operate under legal mandates that persist regardless of economic efficiency. The Americans with Disabilities Act requires complementary paratransit service at costs of ~$40 per trip, consuming about a fifth of operating budgets while carrying fewer than 3% of riders (Canada has parallel requirements). Many agencies also have a political mandate to serve low-density areas where fixed routes already cost $15-to-$25 per passenger; a powerful engine of the Endless Emergency that makes most transit service so difficult to improve.

These mandates create a predictable cycle that perpetuates microtransit adoption despite evidence of failure. Coverage requirements provide initial justification for flexible service experiments. Federal innovation grants—easier to secure than funding for proven investments like frequency improvements—cover startup costs while deferring questions about long-term sustainability. This creates perverse incentives where agencies can more easily fund experimental microtransit than expand proven fixed-route service. The result is that transit managers rationally pursue funding they know they can get rather than projects they know would work.

Of course, there are also market players, keen to sell their services to transit operators, who elide awareness of the problem. Those firms sidestep the core difficulty by promising to transcend it through superior technology, smarter operations, or more sophisticated targeting. Companies promise that artificial intelligence will optimize routing so efficiently that the old rules no longer apply; algorithms will bundle passengers with supernatural precision, turning every van into a virtual bus route. It would be nice if it were true, but I'm unconvinced that this is a problem that can be solved by thinking harder. Scattered origins and destinations in low-density areas simply cannot generate the passenger volumes needed to offset the service costs, at least today. No amount of computational power can make three passengers per hour economically equivalent to thirty.

These perennial forces—political pressure, funding for innovation, and marketing hype—are important to understand, because their consequence is that microtransit experimentation will continue in the future, even though the model has been shown, largely, not to work.

Some Natural Rejoinders

At this point, you may have questions. Even if microtransit costs more than traditional service, isn't it still worth paying for if riders prefer it? The answer is no, but justifying that answer requires understanding when public transit subsidies make sense.

I am on record as thinking that they don’t, but I am aware my opinions are heterodox. Let’s reason from within conventional wisdom today.

As per conventional wisdom, it’s justifiable to depart from laissez-faire in transport to achieve public goals that would not otherwise be met. Four of those goals are reducing congestion, cutting emissions, preventing road incidents, and ensuring that everyone, not merely the wealthy, has mobility options. In dense urban areas, all four of these apply, and (for example) fixed-route bus service hits all of them.

That said, the public purse isn’t bottomless. If we’re going to subsidize operations to achieve those goals, there will nonetheless be a limit. So what level of subsidy is appropriate? If my math is good, it seems that US transit systems collectively averaged $6.88 in subsidy per trip in 2023.¹ That’s a useful benchmark. If microtransit can't beat that figure while delivering meaningful social benefits, we can get more social utility by doing what we usually do, i.e., fixed-route service.

As we have seen, microtransit can’t get there.

You might also wonder why only the public faces this problem. If microtransit fails, why do private ridesharing services, like Uber Pool, succeed?

The short answer is that they don’t, which is why they largely don’t exist any more. Uber suspended Pool throughout North America in March 2020, initially citing COVID concerns; it has yet to bring it back at scale. Lyft killed Lyft Line and formally dropped all shared rides in April 2023. But the pandemic was merely a handy pretext: the underlying economics were already deteriorating before the pandemic. In Chicago, for instance, requests to share fell from 27 percent to 12.8 percent over 2019. Many people just won’t share rides with strangers, and lack of take-up increased cost margins, meaning that the savings differential on a private versus a shared trip declined, leading to even less take-up, and so on. Even when pooling worked, results were modest: observed mean occupancy for pooled trips was approximately 1.3 passengers per ride, which is barely above solo occupancy.

The geometric constraints that doom flexible public transit also doomed private ride-pooling. The difference is that private companies can exit when the math doesn't work.

Approaches That Minimize Damage

But some public provision not only persists, but succeeds (I won’t say flourishes.) The most successful operations make a virtue of necessity by using one flexible service to cover paratransit and remote-area coverage obligations, spreading costs across multiple funding streams. Golden Empire Transit merged general public rides with ADA paratransit, achieving 60% better vehicle utilization while drawing Medicaid reimbursements that quadruple the agency’s typical farebox-recovery rates. This integration approach recognizes that if you're already operating expensive, low-productivity service for legal compliance, expanding it on the margin to serve additional riders can improve overall efficiency.

Another useful approach focuses on specific market gaps where microtransit's flexibility provides value despite higher unit costs: first-mile connections to rail stations, off-peak coverage when fixed routes become sparse, or rural coverage where fixed-route buses also cannot operate sustainably. This targeted approach works because instead of competing with efficient fixed routes, it's competing with no service at all or with even more expensive alternatives like individual taxi vouchers. The service fills a genuine coverage gap; there’s no need to gild the lily with claims about revolutionizing transport efficiency.

Put another way, Innisfil's strategy of building transit ridership and political support through subsidized ridehail was sound, using the sizzle of an Uber partnership to demonstrate to residents and politicians the latent public demand for transit service. The Town's error was treating the program as an end, rather than a means. Having proven that residents valued public transit, Innisfil should have transitioned to fixed-route service much earlier, before costs became unsustainable.

From all this, we can see that microtransit is like a specialized medical procedure; it costs more than routine care, and should only be deployed where necessary… but, in some cases, it is necessary. The agencies achieving financially-tolerable operations share a pragmatic approach: they've sized services to realistic demand, integrated operations with existing requirements, and accepted that flexible service comes with cost premiums that will always exceed efficient fixed routes. They use a ~$7 per trip benchmark as their ceiling rather than their starting point, recognizing that microtransit costs of $15-to-$20 per ride can never be justified in terms of efficiency, but only in terms of other goods.

This analysis of human-driven microtransit establishes the crucial baseline for evaluating the potential impact of driving automation. The central question will become whether automation can overcome the limitations that make microtransit a niche service today. We will consider that question in detail later this year.

Respect to

Emma McAleavy

,

Kevin Kohler

, and

Jeff Fong

for feedback on earlier drafts.

---

Changing Lanes on the Road

As a reminder, later this year, on 22 September 2025, I will be speaking at the Ontario Transit Funding Forum in Toronto.

At the Forum, co-hosted by Transport Futures and Toronto Metropolitan University's TransForm Lab, I’ll join other expert speakers from government, business, academia, and NGOs to discuss revenue and funding policies that can support efficient transit operations while increasing ridership. I hope to see you there!

You may register here, but note that Transport Futures has generously offered a 10% discount to paying subscribers of Changing Lanes! If that’s you, and you would like to attend, please contact me via Substack for details.

1. As per Exhibit 16.1 on p. 148, in 2023 US agencies had $57.7B in operating expenses against $10.1B in fare revenue, while delivering 6.9B trips (Exhibit 11.1 on p. 103). Expenses less revenue divided by trips comes to $6.88 per trip.