Revenue management based on dynamic programming in unrestricted and simplified fare structure

Abstract

In the past few years, many low-cost airlines have been created around the world. Legacy carriers used to apply revenue management (RM) to fully-restricted fare structures with segmented demand. These low-fare competitors with simplified or unrestricted fare structures often manage to capture an important part of the demand in the markets they enter, forcing legacy carriers to change their fare structures. However, because of this change in fare structures, most RM forecasters and optimizers that were previously used by the legacy carriers are no longer effective. The goal of this thesis is to describe and test two RM methods based on dynamic programming in fare structures with few or no restrictions. The first one, the "Lautenbacher" approach (DP-LB), was developed for fully-restricted environments but, when associated with an appropriate forecaster, may be used in less-restricted fare structures. The second one, the "Gallego-Van Ryzin" approach (DP-GVR), was developed for unrestricted fare structures. In this thesis we describe these two methods, and analyze the performance of both in comparison with the performance of traditional RM methods used by legacy carriers, based on results obtained with the Passenger Origin-Destination Simulation (PODS). In our simulation DP-LB leads to results very close to those obtained with traditional RM methods. In unrestricted fare structures DP-GVR appeared to be theoretically appealing. The simulation results obtained with DP-GVR against a competitor using no RM or against a competitor using the same RM method were better than those obtained with other RM methods. But against advanced competitors, DP-GVR got worse results than other RM methods.

(cont.) We showed that this variability in the performance of DP-GVR was related to its sensitivity to forecasts of probabilities of sell-up, forecasts that are difficult to estimate because they depend on the RM methods and actions of the competition. Further study with improved estimation of probabilities of sell-up will be required to determine whether this method can lead to stable improvement over traditional RM methods in unrestricted fare structures.