Title: A simulation study: comparing the Markov decision process approach to the expected marginal seat revenue b heuristic under differing demand behaviour

Authors: Michael Frank, Martin Friedemann, Anika Schroder

Addresses: Revenue Management System Development (FRA ID/I), Lufthansa German Airlines, Lufthansa Base, Frankfurt/Main 60546, Germany. ' Revenue Management System Development (FRA ID/I), Lufthansa German Airlines, Lufthansa Base, Frankfurt/Main 60546, Germany. ' Controlling Central Europe, TUI Germany, Karl-Wiechert-Allee 23, Hanover 30625, Germany

Abstract: In this article, we compare Markov Decision Process (MDP) to expected marginal seat revenue b (EMSRb) under realistic demand behaviour by implementing both in a stochastic simulation model. First come first served as well as an upper bound, serves for reference. We give a short explanation of the investigated methods, describe the simulation setup, in particular the demand model, and carry out sensitivity analyses regarding settings of the demand generator and the optimisation methods. Computational results show that MDP outperforms EMSRb in all investigated scenarios but we raise the question how much improvement is needed to justify the additional effort of the more complicated procedure.

Keywords: dependent demand; EMSRb; expected marginal seat revenue b; MDP; Markov decision process; revenue management; stochastic simulation; upselling; demand behaviour.

DOI: 10.1504/IJRM.2009.024161

International Journal of Revenue Management, 2009 Vol.3 No.2, pp.133 - 147

Published online: 27 Mar 2009 *

Full-text access for editors Full-text access for subscribers Purchase this article Comment on this article