Ride-hail and delivery under demand, fleet, and promise constraints.
The problem.
A live ops team is making three plans simultaneously — where vehicles should be, which orders to commit to, and which promises already made still hold. The window is minutes, not hours, and the cost of a bad commit shows up in the next hour, not next quarter.
Dispatch heuristics keep the system alive. They do not explain why the 17:40 wave was bad. The post-mortem moves slower than the next shift.
The shape of the decision.
Variables: order-to-vehicle assignments, repositioning moves, batching decisions. Constraints: time windows · capacity · driver hours · service-level agreements. Objectives: completion · ETA reliability · driver utilization.
How Alpha Z helps.
Dispatch becomes a formulation, not a heuristic. The algorithm returns assignments with the binding constraint named — so when an ETA slips, the system can say which promise had to give and why.
What the system learns about which formulations work in which demand regimes becomes a structured insight in the library — the next surge starts with the right model, not the right guess.