RateHighwayPROJECT APOGEEMenu
Mission Control in Houston during the Apollo 11 lunar surface EVA.

Designation NB-01 · Lab notebookCar rental only

The week as one object.

This notebook asks whether rate, utilization, and length-of-rent mix can be evaluated as one decision. It is research, not a product.

S69-39593 · Mission ControlApollo 11 EVA, 20 July 1969. NASA. Public domain.

Notebook · not a module

Today, a revenue manager may view demand curves, competitor rates, and fleet levels in separate tabs. RateMonitor already applies pricing rules at market speed within the floors, ceilings, and utilization limits set by the operator. This notebook asks whether the whole week can be scored as one decision instead of as three separate results that can cancel one another.

Two results that can look successful but are not

Filling the fleet at too low a rate and holding a high rate while cars stay idle are both failures. A useful measure is contribution compared with fleet size: invoiced revenue divided by average daily fleet. This keeps average daily rate (ADR) and utilization in the same view. RateHighway already links them in practice: raise rates when the fleet is tight, lower rates when demand is weak, and always respect the operator’s floor.

Utilization rules create a local feedback loop. Fleet teams decide how many cars are at a location. Revenue managers set price. The rules respond when utilization crosses a threshold. But they do not ask which rental should use the remaining car: a one-day online travel agency (OTA) booking, a corporate weekly rental, or a one-way rental that moves the car to another location.

A week can show 95% utilization and higher ADR while still losing contribution. This can happen when rate, demand, and rental length are optimized one after another. They should be evaluated together.

Pickup is a constraint

Reservations build toward a pickup date. Downtown demand often books late; airport demand often books earlier. The pickup model estimates the final demand from bookings already on the books and the usual booking pace:

F = OTB + (1 − c(L)) · B

Here, OTB means on-the-books reservations. c(L) is the share usually booked at lead time L, and B is the seasonal baseline. Pickup models are good at completing a familiar booking curve. They are less reliable when the type of demand has changed. They estimate a number; they do not score the whole week.

Melt means reservations that cancel or do not arrive. It varies by booking type: prepaid and pay-on-arrival reservations do not behave the same way. Length of rent (LOR) also matters. Short and long rentals have different costs and occupy different days. A one-day and a seven-day rental should not disappear into one average. The week is a set of connected rental periods, not a list of separate days.

A direction, not a paper

The mathematics we are exploring

Treat the week as one decision x = (r, u, λ): rates across locations, vehicle classes, days, and channels; utilization over time; and the mix of rental lengths. The objective is contribution across the network that actually shares cars. Pickup is a constraint. Rate, utilization, and rental-length mix are controls in one decision, not separate tabs.

x = (r, u, λ) J(r, u, λ) = ∑_{s,k,t} [ π_{skt}(r) · u_{skt} · m(λ) − κ_{skt}(u, λ) ] subject to, for each station s, class k, day t, lead L: F_{skt}(L) = OTB_{skt}(L) + (1 − c_{sk}(L)) · B_{skt} u_{skt} = occ(F, fleet, λ) ∈ [0, 1] λ ∈ Δ^{|LOR|} r ∈ [r̲, r̄] Sequential RM is the broken decomposition r* = argmax_r R(r | û, λ̂) u* = argmax_u U(u | r*) λ last, if at all. The laboratory refuses the order. Rate, occupancy, and LOR mix are two (three) actuators of one decision. Empty cars at a high rate and full cars at a low contribution print as the same failure.

J

Contribution across the locations that actually share cars. It is not ADR alone, utilization alone, or a score the operator cannot explain.

λ

The mix of rental lengths. The model chooses the mix instead of measuring it only after the week is over.

F = OTB + (1 − c(L)) · B

The pickup model still applies. It limits what can still arrive, but it is not the objective for the whole week.

This research extends ideas RateHighway already uses: utilization as an action signal, LOR as an analytical view, and RateMonitor as the execution layer. A connected objective must still include channel cost, vehicle-turn cost, upgrades, and the future value of the remaining fleet. The operator must be able to understand and challenge the result.

This is a research notebook. It is not an announced product or release. The operator still decides.

All notebooks → · Possible futures →