Mobility and transportation
Sizing an on-demand mobility fleet with simulation
Routing algorithms and a discrete-event simulation that let the business size its fleet for real demand variability rather than peak assumptions.
27%
Reduction in required fleet size
300K+
Trips scheduled by the routing algorithms
Client
Global automotive manufacturer, mobility services division
Basis
Partner experience, prior role
The insight
“Sizing for peak buys reliability at a price the business model cannot carry. Sizing for average fails exactly when riders need the service most.”
Exhibit
Peak sizing buys the last few points of service at a steep price
Share of daily trips served, by fleet size
Illustrative, computed from stated assumptions: hourly Poisson demand with morning and evening peaks; each vehicle completes 2.5 trips per hour. Peak sizing covers the busiest hour at 99% confidence (43 vehicles); 31 vehicles serve 97% of daily trips.
Fleet size was the largest capital decision
The manufacturer was building on-demand and autonomous mobility services, where every trip request had to be matched to a vehicle and routed in near real time. Fleet size was the largest capital decision in the business model.
Neither peak nor average was the right answer
Demand for on-demand trips varies sharply by hour, day, and location. Sizing the fleet for peak demand would make the service reliable but uneconomic; sizing it for average demand would leave riders waiting at exactly the moments that mattered most. Spreadsheet capacity models could not capture how routing, dispatch, and demand variability interact.
Testing fleet sizes in simulation before buying vehicles
We built custom vehicle routing algorithms to dispatch and schedule trips, and a discrete-event simulation that replayed realistic demand through the full dispatch and routing logic. The simulation allowed fleet size, service-level targets, and dispatch policies to be tested together before any vehicles were committed. The routing work also produced a patented dynamic programming method for route optimization.
A 27% smaller fleet at the same service
The simulation identified a 27% reduction in required fleet size while maintaining service, and the routing algorithms scheduled more than 300,000 trips.
What made the difference
01
Route every trip
Custom vehicle routing algorithms dispatched and scheduled trips, producing a patented dynamic programming method along the way.
02
Replay realistic demand
A discrete-event simulation pushed realistic, variable demand through the full dispatch and routing logic.
03
Decide jointly
Fleet size, service targets, and dispatch policy were tested together before any vehicles were committed.