Decision SupportIntegrated revenue management approaches for capacity control with planned upgrades
Highlights
► Simultaneous consideration of capacity control and planned upgrades in revenue management. ► Derivation of new structural insights that facilitate the control process. ► Proposition of two new dynamic programming decomposition approaches. ► Extensive simulation study based on real-world car rental data. ► Outperformance of planning approaches widely-used in practice today.
Introduction
Capacity control, which is considered to be the key component of modern revenue management, is concerned with the task of optimally selling a fixed perishable capacity within a given time horizon by controlling the availability of the products that make use of this capacity (see, e.g., Talluri and van Ryzin, 2004). However, traditional capacity control often ignores the fact that many firms in the service industry offer several products that are substitutable in the sense that the seller could fulfill a certain product request with a more desirable substitute from a prespecified set of alternative products. Examples include airlines selling economy, business, and first class seats and car rental companies offering many types of cars differing in size and features.
In this paper, we consider the situation when this substitution is provided at the original product’s price, which is called an upgrade. It is assumed that customers always accept upgrades to superior products if they are offered these at no extra cost. Conversely, if customers are urged to voluntarily buy a higher value product – in practice, this usually occurs at the time of fulfillment – this is termed an upsell. Upgrading and upselling can be beneficial if the selling firm faces a mismatch between supply and demand. This mismatch often occurs when capacity decisions have been determined for the long term, which is usual in traditional revenue management scenarios, while firms experience stochastic and seasonal demand. We assume that the capacity mismatch is relatively transient or not very pronounced; customers will therefore not strategically adapt to upgrades.
Two important aspects of upgrading are fairness and scope (see, e.g., Gallego and Stefanescu, 2009). Informally, upgrades are defined as fair if upgrade priority is given to customers who purchase higher quality products. Specifically, no customer should receive a higher quality upgrade than another who had originally bought a higher quality product. Regarding the scope of upgrades, two basic models can be distinguished: Full cascading, which allows upgrades to any higher quality product, and limited cascading, which is less flexible as upgrades are allowed only to the next higher quality product.
The research presented in this paper is motivated by an industry project conducted with a major car rental company. In fact, the car rental industry is one of the most important users of upgrades. As Geraghty and Johnson (1997) note, it is essential for car rental companies to undertake market segmentation by not only offering different products using the same resources but also by offering many different car types. This is especially important as some typical fencing conditions known from the airline industry, for example, advanced purchase restrictions, do not seem to work well in the car rental business (see, e.g., Carroll and Grimes, 1995). However, by offering many different car types, car rental companies cannot directly substitute one type for another but must consider a certain predefined upgrade hierarchy. The resulting supply-side substitution opportunities are inevitable as, although desirable, there still is no tight integration of fleet management and revenue management decisions in the car rental industry (see, e.g., Lieberman, 2007). Furthermore, as the differences in the cost are usually quite small, car rental firms have a common policy to acquire fewer economy cars than required and more compact or full-size cars. Through upgrading, the latter cars offer more flexibility if the demand situation changes.
Although the above examples illustrate an obvious need to explicitly incorporate upgrading opportunities in revenue management’s capacity control process, there is not much theoretical work on developing appropriate approaches. Practical implementations usually resort to rather simple heuristics, such as successive planning, which means that upgrade contingents on higher-valued resources are determined first and that, subsequently, the new virtual capacity is considered fixed for standard capacity control. Therefore, one of the contributions of this paper is the proposition of integrated dynamic programming decomposition approaches for capacity control with planned upgrades. Our approaches extend the well-known dynamic programming decomposition from traditional capacity control to additionally exploit the opportunities of planned upgrades. Furthermore, we show that the approaches are applicable to real-world scenarios in car rental revenue management and that they outperform common practice procedures, such as the successive planning of upgrade contingents and capacity control decisions.
The paper is structured as follows: Section 2 provides a brief review of related literature. In Section 3, we present two dynamic programming formulations for capacity control with planned upgrades that are adopted from the literature. Based on these formulations, as our first contribution, we theoretically derive new structural insights that facilitate the control process under certain conditions that often occur in practice, such as single-leg airline capacity control and daily car rental capacity control. Based upon these models and insights, we then propose two different dynamic programming decomposition approaches for capacity control with planned upgrades, which is the second contribution of the paper (Section 4). While the first approach is specifically suited for multi-day revenue management problems that occur, for example, in the hotel and car rental businesses, the second approach is more general and can be applied in arbitrary network revenue management settings. In Section 5, we discuss the computational results from a simulation study based on real-world data obtained from our car rental industry partner. These results demonstrate the decomposition approaches’ practical applicability and their relative performance compared to other common control procedures, such as successive planning. We conclude with a summary of the key results (Section 6). The proofs of all propositions as well as complementary tables with additional information on the simulation study are provided in the Online Appendix.
Section snippets
Literature review
There is an extensive literature on revenue management in general. For surveys, see, for example, Belobaba, 1987, Weatherford and Bodily, 1992, McGill and van Ryzin, 1999, Chiang et al., 2007, as well as the textbooks by Talluri and van Ryzin, 2004, Phillips, 2005.
While most revenue management publications consider only one type of resource and thus do not mention any supply-side substitution, a few authors have proposed approaches that integrate planned upgrades with capacity control. Alstrup
Model formulations
In this subsection, we restate two general dynamic programming formulations of the revenue management problem with planned upgrades that have been proposed by Gallego and Stefanescu (2009). We consider a firm offering n products. For each product k ∈ {1, … , n}, pk denotes its predefined price that is valid throughout the sales horizon. The sales horizon can be sufficiently discretized into T time periods so that no more than one buying request arrives in each period t ∈ {1, … , T}. Each request asks for
Dynamic programming decomposition approaches for the multi-day capacity control problem
In this section, we focus on multi-day capacity control with planned upgrades, as faced, for example, by hotels and car rental companies. We first slightly change the notation used in Section 3 to make the modeling approach more comprehensible for this application field (Section 4.1). Note that we restrict ourselves to considering the ad hoc variant in which the upgrade decision is made at the time of sale. Because the dynamic programming (DP) formulation is intractable for multi-day problems
Computational results
In this section, we report the results of a simulation study that demonstrates the applicability and relative performance of the dynamic programming decomposition approaches derived above. The simulations are based on real-world demand and capacity data provided by a major car rental company. To protect the interests of this company and to guard against the release of sensitive data, we have held back (or modified) certain details. The experiments were conducted on an Intel Xeon processor-based
Conclusions
In this paper, we address the problem of integrating revenue management’s capacity control with upgrade decision making. We derive new structural properties for an integrated dynamic programming formulation (DP) of this problem that was recently proposed by Gallego and Stefanescu (2009). Depending on the field of application, these new properties simplify the resulting control process under certain conditions, for example, when only one leg or day is considered. In particular, we prove the
Acknowledgements
The authors thank the editor and three anonymous referees for their detailed and constructive comments that helped improving this paper.
References (39)
- et al.
Booking policy for flights with two types of passengers
European Journal of Operational Research
(1986) - et al.
Network revenue management with inventory-sensitive bid prices and customer choice
European Journal of Operational Research
(2012) - et al.
Dynamic control mechanisms for revenue management with flexible products
Computers & Operations Research
(2010) - Belobaba, P.P., 1987. Air Travel Demand and Airline Seat Inventory Management. Ph.D. Thesis. Flight Transportation...
- et al.
Simulation-based booking limits for airline revenue management
Operations Research
(2005) - et al.
Evolutionary change in product management: experiences in the car rental industry
Interfaces
(1995) - et al.
An overview of research on revenue management: current issues and future research
International Journal of Revenue Management
(2007) - et al.
The pool control system
Interfaces
(1977) - et al.
Separable approximations for joint capacity control and overbooking decisions in network revenue management
Journal of Revenue and Pricing Management
(2009) - et al.
A dynamic programming decomposition method for making overbooking decisions over an airline network
INFORMS Journal on Computing
(2010)
Revenue management of flexible products
Manufacturing & Service Operations Management
Revenue management saves national car rental
Interfaces
The passenger-mix problem in the scheduled airlines
Interfaces
A reinforcement learning algorithm based on policy iteration for average reward: empirical results with yield management and convergence analysis
Machine Learning
A reinforcement learning approach to a single leg airline revenue management problem with multiple fare classes and overbooking
IIE Transactions
Simulation optimization for revenue management of airlines with cancellations and overbooking
OR Spectrum
Revenue management in the car rental industry: a stochastic programming approach
Journal of Revenue and Pricing Management
Cited by (41)
Optimal reservation control strategies in shared parking systems considering two types of customers
2023, Computers and Operations ResearchAutonomous mobility of a fleet of vehicles for precision pesticide application
2021, Computers and Electronics in AgricultureA greedy approach to improve pesticide application for precision agriculture using model predictive control
2021, Computers and Electronics in AgricultureA review of revenue management: Recent generalizations and advances in industry applications
2020, European Journal of Operational ResearchCitation Excerpt :Two basic forms of upgrading can be distinguished (see Gallego & Stefanescu 2009): Full cascading (multistep models) allow the seller to fulfill the demand for a product with any higher-quality product, while limited cascading (single-step models) allow an upgrade only to the next higher-quality product. Steinhardt and Gönsch (2012) and subsequently Gönsch and Steinhardt (2015) analyze the dynamic programs with full cascading upgrades in the context of car rental and passenger air transport, respectively. Steinhardt and Gönsch (2012) prove that if only a single rental day (i.e. a single leg) is considered, opportunity cost is monotonous with regard to the upgrade hierarchy.
How much to tell your customer? – A survey of three perspectives on selling strategies with incompletely specified products
2020, European Journal of Operational ResearchIntegrating pricing and capacity decisions in car rental: A matheuristic approach
2018, Operations Research PerspectivesCitation Excerpt :In fact, due to the inherent flexibility of the fleet, this sector is often studied in revenue management, especially as far as capacity allocation is concerned. For example, Guerriero and Olivito [8] derive different acceptance policies for car rental booking requests while Steinhardt and Gönsch [9] integrate these approaches with operations issues related with planned upgrades. As for pricing, it is considered as an emerging tool used by practitioners to manage demand, since it is increasingly easy and cheap for companies to dynamically and swiftly change the prices through online booking channels [10,11].