Skip to main content

Predicting Demand from Sales Data: Unconstraining in the Car Rental Industry

Avis Budget LogoIn Revenue Management applications, the sales process is often managed by imposing booking limits on certain booking classes so as to reserve capacity for late-booking, high-value customers. Once a limit is reached, in constrains demand observations for that class; but using the observed sales data as demand estimate will always be an underestimation that can potentially severely hurt revenues.


In his M.Sc. project in the MSOR programme 2011/12, Darayus Vakil pursued the question of how to estimate this unobserved demand (i.e., "to unconstrain demand") in a context where bookings in each class are intermittent, under the supervision of Dr Arne Strauss and Dr Wenjuan Zhang at WBS and in close contact to Dr Dong Li at the car rental company Avis Budget EMEA.

The proposed technique is simple and intuitive, and can easily be implemented in a spreadsheet. It is based on Croston's method, which was modified to account for the observation that the intervals of demand arrivals tend to become smaller as we near the pickup date. It also possible to estimate demand when booking classes are opened and closed various times.

As Dr Li informed us, implementation has been made a top priority project at our collaboration partner Avis Budget EMEA.

The solution was convincing and pragmatic.

A warm thank you to the minds who have been designing this solution, to whom we wish to continue to shine in their academic career.

Carlo Pacifico, 7 September 2012
Mediterranean Revenue Manager
France, Italy and Spain
Avis Budget EMEA