From lab to street: Why multimodal is the key to affordable demand response

The promise of microtransit was simple: provide the convenience of a taxi with the accessibility of public transit. But as agencies move from small pilots to city-wide deployments, the math is starting to break. 

Traditional microtransit functions as a heavily subsidized transportation network company (TNC), offering point-to-point service that mimics a taxi. However, microtransit is hitting a wall. While popular with riders, the unit economics of microtransit are fundamentally unsustainable, especially when compared to the cost per rider and farebox recovery for fixed-route bus and rail service. When a vehicle is dedicated to a single trip from origin to destination, it creates significant deadheading and keeps passengers per revenue hour (PPRH) in the low single digits and cost per rider in the mid double digits. 

With improved operation, funding and optimization, the average cost per rider (UPT) for directly operated demand response services for U.S. transit agencies decreased to $51.50 in 2024, per the U.S Department of Transportation’s National Transit Summaries & Trends report. But for many agencies, subsidies exceeding $30 to $50 are a non-starter for long-term operations. The industry is trapped in a binary choice: offer high-quality microtransit and eat the cost or offer inexpensive fixed route with coverage gaps. 

Microtransit becomes multimodal 

The solution isn't to abandon microtransit, but to evolve its role from a standalone service to a high-frequency feeder for existing fixed routes. By using microtransit exclusively for the first- and last-mile trips to bring riders to and from fixed-route stops, agencies can leverage the high capacity and more affordable unit economics of bus and rail for most of the trip. 

However, the optimization problem suddenly becomes much more complex. For multimodal to work, it needs optimized transfer timing, waiting time, traffic and dynamic rerouting. Missed transfers or misestimates can lead to cascading delays and compounding errors that eat into multimodal’s cost reduction and rider trust. This difficult technical barrier is why multimodal has historically delivered only marginal cost improvements. 

To bridge this gap, researchers at the Socially Aware Mobility Lab at Georgia Tech and at the University of Michigan have spent eight years developing the optimization to make multimodal mobility work at scale. 

To solve these enormous problems quickly, an algorithm called column generation can be used to focus only on the most promising route combinations instead of checking every possibility. Bilevel optimization is another technique that’s used to integrate latent demand, a famous chicken and egg problem for transit. By solving both problems simultaneously, this approach balances affordability for the agency and service quality for the rider. Optimization techniques like these do the heavy lifting to make multimodal service work at scale. 

Powerful as these optimization tools are, they have the strongest impact when shaped by real-world input from the very agencies that will run them. Transit operators bring invaluable insights like rider habits, unwritten service rules and day-to-day constraints that improve these models far beyond what simulation alone can achieve. 

This collaborative process ensures the system isn’t just theoretically efficient, but truly reliable and scalable in practice. Only then does the math translate into measurable wins for both budgets and riders. Deployments with agencies like the Metropolitan Atlanta Rapid Transit Authority (MARTA) in Atlanta and University of Michigan Campus Transit in Ann Arbor, Mich., put these optimization advances to the test, demonstrating that high connectivity with fixed routes led to improved service metrics and significant cost reduction. 

The operational reality 

During those research pilots, the impact of multimodal optimization was immediate. At the University of Michigan, existing high frequency bus routes allowed for an average wait time of just three minutes. In larger urban environments like Atlanta, the ability to predictably feed fixed routes contributed to a 66% month-over-month increase in ridership: monthly ridership surged by 10 times over six months. 

These metrics suggest that riders aren't inherently opposed to transfers; they are opposed to unreliable ones. For decades, the transfer penalty, or the idea that riders will inherently choose a slower direct route over a faster one that requires a transfer, has governed transit planning. But riders in transit deserts don’t have the luxury of alternative transit options.  When optimization reduces the gamble of a failed transfer, riders are less opposed. 

The research to reality pipeline is often where good ideas go to die, but for public transit, it is where the next generation of service must be born. We can no longer afford to view the high cost of microtransit as a necessary evil to provide first- and last-mile coverage. 

The goal of transit has always been to provide the best mobility for the most people at the lowest cost. For too long, demand response and fixed-route service have been siloed into two different budgets and two different goals. The solution to the high cost of microtransit isn’t shrinking service areas or increasing budgets to throw money at microtransit’s unsustainable unit economics. It’s about bridging these two worlds. Only through multimodal can microtransit move into a scalable, sustainable reality.