Hotel Management With Foreign Collaboration

A Comparative Revenue Analysis Of

Hotel Yield Management Heuristics

By

Timothy Kevin Baker

Bass Hotels and Resorts, Inc.

Three Ravinia Drive, Suite 2900

Atlanta, GA 30346-2149

E-mail: baker.tim@hiw.com

and

David A. Collier

The Ohio State University

1775 College Road

Columbus, OH 43210

E-mail: collier.4@osu.edu

A Comparative Revenue Analysis Of

Hotel Yield Management Heuristics

ABSTRACT

Yield management is the dynamic pricing, overbooking, and allocation of perishable assets across market segments in an effort to maximize short-term revenues for the firm. Numerous optimization heuristics for allocation and overbooking exist for the airline industry, whose perishable asset is the airline seat. When an airline departs, no revenue is gained from the empty seat(s). In the hotel industry, the perishable asset is the hotel room--once a room is left empty for a night, that night's revenue cannot be recaptured. The literature on yield management heuristics for the hotel industry is sparse. For the hotel operating environment, no research has adequately (1) integrated overbooking with allocation, (2) modeled the phenomenon of hotel patrons extending or contracting their stay at a moment's notice, or (3) performed a realistic performance comparison of alternative heuristics.

This research develops (1) two hotel-specific algorithms that both integrate overbooking with the allocation decisions, (2) a simulation model to model realistic hotel operating environments, and compares (3) the performance of five heuristics under 36 realistic hotel operating environments. Seven conclusions are reached with regard to which heuristic(s) perform best in specific operating environments. Generally, heuristic selection is very much dependent on the hotel operating environment. A counterintuitive result is that in many operating environments, the simpler heuristics work as well as the more complex ones.

Subject Areas: Heuristics, Service Operations, and Simulation

A Comparative Revenue Analysis of

Hotel Yield Management Heuristics

INTRODUCTION

Yield management is the dynamic pricing, overbooking, and allocation of perishable assets across market segments in an effort to maximize short-term revenues for the firm. Yield management provides some success stories and a few non-publicized failures. For example, American Airlines increased its revenues by an estimated $1.4 billion over the 1989-91 time frame. Net profits after taxes were $892 million during that same period (Smith & Leimkuhler, 1992). Hertz rental cars, Marriott International hotels, and the Royal Caribbean Cruise Line have also benefited from yield management implementations (Lieberman, 1992). Hertz increased its annual revenues by 1-5% annually. Marriott improved its 1991 revenues by $25-35 million. Royal Caribbean Cruise Lines obtained a revenue increase in excess of $20 million for one year. However, these benefits are not easy to obtain (Hanks, Cross, & Noland, 1992; Lieberman, 1992). One hotel chain spent over $1 million in implementing a yield management system, only to discontinue using the system because it was not appropriate for their properties. One airline attributed losses in excess of $10 million due to errors in its yield management models.

The main objectives of this research are to: (1) develop modified heuristics for the hotel operating environment, (2) test and compare five heuristics, and (3) help hotel managers select the appropriate reservation control heuristic that best suits their operating environment. These heuristics are tested under the most realistic hotel-specific operating conditions to date. For example, the operating environments in this research include the following realistic characteristics: (1) room demand modeled as a Poisson process with varying demand intensities and timing, (2) daily cancellation rates for the demand that is accepted, (3) no-show rates, (4) errors in estimating no-show rates, (5) stayover extension-contraction levels--probabilities that the customer will extend or contract his or her stay a certain number of days at a moment's notice, (6) errors in estimating the stayover extension-contraction levels, and (7) room rate differences between regular and discount customers. These factors are defined in Tables 5 and 6 and permissible values are listed in subsequent tables.

These factors are combined into 12 different operating environments based on interviews with hotel managers (Hyatt 1993; Marriott 1992). The operating environments are differentiated by: (1) whether the hotel is primarily for business or resort purposes, (2) the quality of the data needed for yield management analysis--good (bad) data implies small (large) errors in estimating no-show rates and stay lengths, and (3) whether the primary customers are transients, groups, or a mixture. See Table 1 for a summary of these operating environments. When these twelve operating environments are combined with three levels of demand intensity, 36 hotel environments are defined. Each heuristic is tested in each operating environment and demand intensity combination to provide a realistic and complete test bed for the implementation recommendations in this paper. Also, a hotel may have an operating environment that is "between" those in Table 1 (e.g., its demand intensity may be between low and medium). In this case, the hotel should look at all nearest operating environments, select the best performing heuristics from those environments, and then perform further testing on all of these heuristics.

Table 1 About Here

The simulation used to test the relative performance of the heuristics defines two customer categories and rolls forward in time day by day, performing a heuristic optimization run each night for a 14-night horizon. It accepts or rejects each reservation request