Airline revenue management : sell-up and forecasting algorithms

Abstract

Recent technological improvements have allowed airlines to implement sophisticated Revenue Management systems in order to maximize revenues. Computational capabilities make it possible to perform network-based analysis of supply and demand and therefore to increase the gains achieved with the help of "0- D control" Revenue Management algorithms. However, the more commonly used and cheaper flight leg-based algorithms have not yet been used to the best of their potential and can still benefit from better modeling of passenger behavior. Our first purpose in this thesis is therefore to evaluate the benefits of incorporating sell-up models into current leg-based airline Revenue Management algorithms. Another question we would like to try and address is whether it would be possible to improve the leg-based models to reach revenue gains comparable to those of O-D control algorithms. To try and achieve this goal, we improve the modeling in our leg-based Revenue Management algorithms by accounting for the possibility of sellup, that is the probability that a passenger will accept a more expensive ticket than originally desired if seats are not available at the lower fare. In addition, previous research has shown that there are revenue gains to be achieved through better forecasting, therefore, we also evaluate the use of better forecasting methods and quantify their revenue impact. In particular, we focus our efforts on understanding the impact of the unconstraining models on revenue gains by using various detruncation methods and comparing their effect.