On the value of booking data for upsell decision-making in revenue management

In passenger aviation and many other areas of transportation, it is common practice to offer customers who have booked a ticket for a lower compartment free seats in higher compartments at a discount before departure, a practice known as upselling. For example, economy class customers are offered a seat in business class for a small surcharge a few days before take-off. Obviously, it matters to whom to offer an upsell and at what price. In this paper, we address this decision problem in a generic fashion for revenue management settings. We assume that the company has disaggregated booking data about the customer’s initial choice of a product from a provided offer set. This data contains information about individual customers’ preferences and may be leveraged to decide on upsell prices. To this end, we propose an optimization approach based on an expectation model, in which customers’ response probability is represented as a conditional probability formally consistent with their initial buying decision in a multinomial logit model. We present variants of the approach based on different levels of exploitable customer-specific booking data. In a numerical study, we investigate the value of this data usage and upselling in general to the company. Upselling in conjunction with knowledge of the customers’ original offer sets and customer segments, substantially increases revenues. Furthermore, the study demonstrates that the proposed approach can lead to larger revenue benefits than a naive benchmark approach which statistically decouples the customers’ upgrade acceptance decision from their original choice during the purchasing process.