The airline pricing problem [article]

Jonas Rauch, Universitäts- Und Landesbibliothek Sachsen-Anhalt, Martin-Luther Universität, Sebastian Sager
2018
Airline revenue management can be separated into two major areas -pricing and inventory control -with the joint objective function of maximizing revenue. Pricing defines optimal fare products, each being a combination of a price, segmentation rules and attributes such as rebooking conditions, and assigns each fare product to a booking class. Inventory control then optimally controls availability of each booking class as a function of expected demand and remaining inventory. Both have a strong
more » ... pact on an airline's profit and therefore play a critical role in its success in a highly competitive market. While inventory control has been the subject of extensive research in the past decade, the pricing side has gotten very little attention in the scientific RM literature. In fact, many authors use the term revenue management and inventory control synonymously. As a result, whereas capacity control is highly automated based on sophisticated forecasting and optimization methods, pricing decisions in industry practice are mostly taken manually with little decision support. One explanation for the underrepresentation of pricing in the RM literature is its complexity: It cannot be analyzed in isolation, but always has to be considered in combination with inventory control, because every pricing decision potentially changes the optimal booking class availabilities. In this thesis we formulate the joint airline pricing and inventory control problem as a two-level optimization problem. Existing publications on this topic focus on analyzing structural properties of the problem under very limiting assumptions regarding the customer choice model and often using deterministic inventory control schemes. In contrast, in this thesis we analyze pricing for a general class of stochastic customer choice models and in combination with dynamic inventory control, with the goal of numerically solving the resulting pricing optimization problem. To this end we conduct a sensitivity analysis of the customer choice model, allowing to numerically compute the gradient of booking probabilities with respect to prices and product attributes. In addition we derive an adjoint equation of the inventory control dynamic program for a single flight, which allows to efficiently compute the gradient of expected revenue with respect to the pricing and demand input parameters. Combining both we are then able to apply gradient-based optimization methods to solve the joint pricing and inventory control problem. Transferring concepts form network inventory control, the pricing methodology is heuristically extended to the network case with a large number of flights connected by transfer traffic. As a by-product an improved optimization and control mechanism for network inventory control is derived, which in a simulation study shows significant revenue gains over the traditional method. iii
doi:10.25673/4740 fatcat:dvawfnbvlzhqnhckv7ne3uaaka