Innovative Data-Driven Demand Estimation Strategies in Revenue Management

Understanding customers’ choice behavior and decision-making process is crucial in many industries, including hotels, airlines, and retail, as it allows firms to offer the right product to the right customer. This has resulted in a growing focus on accurately estimating customer demand for a product or service in the revenue management space, which can serve as a valuable input to optimal pricing, inventory, and assortment decisions. However, much of the existing work in the literature has not been able to comprehensively address some common challenges that are observed in practice: unobservable no-purchases, heterogeneous customer preferences, and the impact of factors beyond prices such as online reviews. In this dissertation, we focus on data-driven demand estimation problems in revenue management and develop novel statistical approaches within a practical framework, using hotel context as the main application area. Specifically, in Chapter 2, we develop a novel statistical estimation method to uncover “true” demand from (censored) sales transactions data when a firm cannot directly observe customers who choose not to purchase any product. Chapter 3 extends this approach to overcome two hurdles simultaneously: unobservable no-purchases and non-homogeneous customer populations with varying preferences, developing a practical estimation and segmentation methodology. While these two chapters focus on the impact of price on customer demand as the main attribute of interest for computational studies, Chapter 4 provides a fresh perspective on the competitive effects of online reviews, specifically the sentiment of text reviews relative to competition, with a large-scale empirical study utilizing the actual bookings data of a hotel chain