Mobile Edition
Print
Get the Mag
Weekly UpdatesThis article tests and calibrates an often repeated assumption about the revenue benefits of reducing dining duration. This assumption is that a reduction in dining duration yields a proportional increase in revenue, so that, for example, a 20 percent reduction in dining duration would yield a 25 percent increase in revenue. This article's simulation-based study of over twelve hundred restaurant scenarios finds that, on average, the revenue bump experienced by reducing the dining duration is less than one-quarter of the amount predicted by the common assumption. Even in the most favorable circumstances, the revenue bump is less than one-half that predicted by the assumption. Thus, while reducing dining duration might result in a marginal increase in revenues, managers should not count on a substantial revenue bump.
The assumption that reducing dining duration will make a substantial contribution to the revenue-enhancing goal of revenue management is tested in the simulation-based experiment described in this article. The concept being tested is that a reduction in dining duration will mean that more guests can be seated in a particular day part. The result will be an increase in revenue. After simulating more than twelve hundred scenarios of possible restaurant operation, the unavoidable conclusion is that any revenue gain from reducing dining duration will be relatively small. To gain much of an increase, the conditions must be perfect in the sense that guests must cooperate by arriving in a timely fashion so that they are waiting to fill seats as they become available. Although some revenue increase should occur, the matter is far more complex than one might originally expect.
Keywords: restaurants, revenue management, capacity management, simulation
**********
Over the last half dozen years or so, restaurant revenue management has received considerable attention. Much of the work on the topic has come from authors based in the School of Hotel Administration at Cornell University. Professor Sheryl Kimes, the most prolific author on the topic, has written extensively about the tools, or levers, that can be applied to manage restaurant revenue. These tools include managing duration and managing price. In this article, I address duration management.
Although duration management seems to make sense, I was curious to determine the extent to which reducing dining duration translates to increased revenues, if at all. Kimes and her coauthors have argued on several occasions that reducing duration would translate into a corresponding proportional increase in revenue (Kimes 1999; Kimes, Wirtz, and Noone 2002; Kimes 2004a, 2004b). Under this assumption, which I will refer to as the "duration reduction assumption," a 20 percent reduction in dining duration would lead to a 25 percent increase in revenue. My aim in this article is to investigate the conditions that affect the validity of this assumption. If it is correct, restaurateurs should pursue this avenue, but if the assumption does not hold, I would suggest that restaurant operators use other revenue management approaches.
To investigate this matter, I developed an extensive simulation-based study of restaurant performance. The results indicate that environmental factors affecting the validity of the assumption include demand intensity, the length of the peak demand window, and customers' willingness to wait for service. I also find that in general, restaurants will capture only a fraction of the revenue expected if the assumption were to hold. Thus, as I explain below, I cannot see how duration management is a viable revenue management tactic, although managers might wish to control table duration for other reasons.
In the remainder of this article, I present a review of the relevant literature, describe the design and present the results of my simulation study, discuss my findings, and offer concluding remarks.
Existing Studies
Restaurant revenue management can be categorized into issues related to managing capacity, issues related to managing demand, and issues related to implementing those concepts. Professor Sheryl E. Kimes has been the most prolific author on the topic, with articles stretching back to 1998. Indeed, to my knowledge, the first use of the term restaurant revenue management appears in a 1998 article by Kimes and her coauthors (Kimes et al. 1998).
Some restaurant revenue management articles cover a range of related topics (Kimes 2004a), while others address specific issues, and still others focus on implementation. Specific topics that have been addressed include capacity management (Sill 1991; Sill and Decker 1999), forecasting (Hu, Chen, and McCain 2004), performance evaluation (Reynolds 2004, Reynolds and Thompson 2007), table mixes (Kimes and Thompson 2004, 2005), table combinability (Thompson 2002, 2003), and duration-related issues (Kimes, Wirtz, and Noone 2002; Kimes and Robson 2004; Noone and Kimes 2005; Noone et al. 2007). Implementation-focused articles have used specific restaurants as examples (Kimes et al. 1998; Kimes 2004b). I direct those readers interested in an overview of restaurant revenue management to Kimes's (2004a) report on the topic published by the Cornell Center for Hospitality Research.
To my knowledge, the effects of dining duration on revenue have been addressed at least four times. In the first such study, Kimes (1999) uses an example where a restaurant has one hundred seats, a four-hour peak window, and a sixty-minute mean dining time: "If the meal time can be reduced to 59 minutes, the restaurant can handle an additional 6.8 customers ... a 1.7 percent increase" (p. 19). The 1.7 percent increase in capacity is the same percentage increase one attains by comparing the old dining duration to the new dining duration (60/59*100 percent).
Kimes, Wirtz, and Noone (2002) and Kimes (2004a) both reference the first Kimes (1999) article, but provide a slightly different example. In this case, the basic scenario is the same: a one hundred-seat restaurant, a four-hour peak window, and a sixty-minute mean dining time. However, in this case, "if the dining duration could be reduced to 50 minutes ... revenue would increase ... [by] 20 percent" (Kimes 2004a, p. 14; Kimes, Wirtz, and Noone 2002, p. 222). The 20 percent revenue increase is the same proportional increase that the old dining duration represents compared to the new dining duration (60/50*100 percent).
Kimes uses the example of a restaurant with annual sales of $2,358,874, having $861,797 of the sales in the ten "hot" weekly hours. She states that "if dining duration could be decreased from fifty-three minutes [the existing value] to forty-eight minutes, even if seat occupancy remained the same, the annual revenue potential would increase by $89,771, or 3.8 percent" (Kimes 2004b, p. 61). Since this extra revenue would be coming from the peak periods only (that is, the only periods where the restaurant was operating at capacity), the percentage increase in revenue during the peak periods is $89,771/$861,797, or 10.4 percent, which is the same percentage increase one gets by dividing the old duration by the reduced duration (53/48*100 percent = 10.4 percent).
In all these cases, what I define as the "dining reduction assumption" is being applied as a simple capacity calculation to estimate the revenue benefits associated with reducing the dining duration. My focus in this article is testing the validity of that assumption.
Design of the Simulation Study
A significant problem with the duration reduction assumption concerns how time savings are accumulated from party to party. For example, the assumption would suggest that if a mean dining of sixty minutes is shortened to fifty minutes, then the savings accumulated on the original five parties would enable a sixth to be served in the same length of time (i.e., five parties times sixty minutes per party = six parties times fifty minutes per party). However, the issue becomes when those time savings materialize. In reality, shortening the dining duration from sixty minutes to fifty minutes really only means that you can seat six parties consecutively at a table in the same time that you previously could seat five parties consecutively. Obviously, unless demand at the restaurant persists sufficiently long enough to capture that sixth party, the actual revenue increase will be lower than what the assumption predicts.
To accomplish my stated purpose of testing the validity of the duration reduction assumption, I developed a simulation-based study to investigate the revenue increase that could be achieved by reducing dining duration. Simulation has been used several times in the restaurant revenue management literature (cf. Kimes & Thompson 2005; Thompson 2002, 2003). An overview of the use of simulation in hospitality contexts can be found in Thompson and Verma (2003).
As listed in Exhibit 1, the study contained nine factors, eight of which were environmental, or largely outside the control of management, and one, the reduction in dining duration, within the control of management. I included four of the environmental factors to ensure I examined a wide range of restaurant environments. These factors are the size of the restaurant, the mean party size, the variation in dining duration within party sizes, and the variation in dining duration across party sizes.
Restaurant size had three levels: 50 seats, 150 seats, and 450 seats. I selected these values to ensure that the study was representative of a significant range of restaurants. Similar restaurant sizes have been used in earlier studies (50-200 seats, Thompson 2003; 230 seats, Kimes and Thompson 2004; 240 seats, Kimes and Thompson 2005). Further, restaurants of 50 to 450 seats constituted 89.7 percent of he sample restaurants in Thompson's (2007) study of table mixes of real restaurants. The two mean party sizes I used were 2.5 and 3.5 people. While the lower average party size is more consistent with the values reported in the literature (approximately 2.6 customers per party, Kimes and Robson 2004), I include the larger party size for thoroughness. Exhibit 2 lists the probabilities of each size party that I used to yield the two mean party sizes.
FEED