Dynamic Pricing for Network Revenue Management: A New Approach and Application in the Hotel Industry

Dynamic pricing for network revenue management has received considerable attention in research and practice. Based on data obtained from a major hotel, we use a large-scale numerical study to compare the performance of several heuristic approaches proposed in the literature. The heuristic approaches we consider include deterministic linear programming with resolving and three variants of dynamic programming decomposition. Dynamic programming decomposition is considered one of the strongest heuristics and is the method chosen in some recent commercial implementations, and remains a topic of research in the recent academic literature. In addition to a plain-vanilla implementation of dynamic programming decomposition, we consider the variants proposed by Erdelyi and Topaloglu (2011) and Zhang (2011). For the base scenario generated from the real data, we show that the method based on Zhang (2011) leads to a small but significant lift in revenue compared with all other approaches. We generate many alternative problem scenarios by varying capacity-demand ratio and network structure and show that the performance of the different heuristics can be strongly influenced by both. Overall, our paper shows the promise of some recent proposals in the academic literature but also offers a cautionary tale on the choice of heuristic methods for practical network pricing problems.