
A traveler with a reservation needs a car of the promised class at a particular hour. An AI model can forecast rental demand from booking patterns, but the desk must also account for late returns, cleaning and vehicles already promised to someone else. A monthly demand estimate and a noon availability count serve different purposes.
Forecast a station, a class and a time
Agita Ferstere’s Station-based car rental demand study in Research and Technology proposes grouping the fleet by class and combining forecasts for monthly demand at a station. The underlying company data are private. That horizon can inform fleet planning. For the desk, record the time a car becomes ready, the requested class and whether the customer received a substitute. A monthly total can be accurate even when the station runs out on Saturday morning.

A returned car can still be unavailable
Start with cars ready now. Add returns expected to clear inspection before the pickup window. Subtract committed pickups in the same class. An expected return is uncertain, and physical presence alone cannot establish readiness. Preserve the forecast as it stood when the desk made the promise so later cancellations do not quietly improve its apparent accuracy.

Whose reservation was actually fulfilled?
Compare the promised handover with the recorded time and class. Count upgrades separately from receiving the class requested. For demand, remember that a sold-out station may stop taking bookings. A different study of Car rental price prediction research from the BCML Lab estimates rental prices. A good price estimate supplies little evidence about the chance of a car being ready.

For the trip-energy forecast behind an electric rental’s arrival reserve, see How AI Estimates the Energy an Electric Car Needs.
Try it with the evidence
Balance a noon rental desk
Source notes. Illustrative station ledger. Six compact cars are ready at 10 a.m. Four more are due back before noon, but only two of those are expected to clear inspection by noon. Nine compact pickups are committed. No other transfers are booked.
Prompt. Calculate the expected noon shortfall. Repeat if only one return clears inspection. What later record should the desk save to evaluate its AI forecast?
Check your reading
Eight cars are expected ready against nine pickups, a shortfall of one. With one cleared return the shortfall is two. Save actual ready times and the time and class handed to each customer, alongside the forecast issued at 10.
Illustrated field notes
From reservation to handover






