A customer type discovery algorithm in hotel revenue management systems

Knowing customer types and their purchase behavior helps revenue management experts to estimate the demand and finally devise a better sales strategy to improve the revenue. Inferring customer types from sales transactions and availability data is a challenging topic in RM. In this paper, we proposed an approach to discover customer types using a classic linear ordering problem. Our linear ordering-based market discovery approach (LMD) comprises three steps: generation of an initial solution, evaluation of the solution by a choice-based model, and finally creation and addition of a new customer type. The number of different customer types is factorial in the number of alternatives and should be pruned. Here, the customer types are pruned based on observed sales and offered-sets, instead of business assumptions or applications. To evaluate the proposed method, a real publicly available dataset of five hotels is used. The results show that LMD outperforms the other available approaches in the literature and improves the log-value results of all datasets by approximately 6%.