Cornell University School of Hotel AdministrationThe Scholarly Commons

Articles and Chapters School of Hotel Administration Collection

10-1999

Developing a Restaurant Revenue-managementStrategySheryl E. KimesCornell University, sek6@cornell.edu

Deborah I. BarrashCornell University, dib3@cornell.edu

John E. AlexanderCBORD Solutions

Follow this and additional works at: http://scholarship.sha.cornell.edu/articles

Part of the Food and Beverage Management Commons

This Article or Chapter is brought to you for free and open access by the School of Hotel Administration Collection at The Scholarly Commons. It hasbeen accepted for inclusion in Articles and Chapters by an authorized administrator of The Scholarly Commons. For more information, please contacthlmdigital@cornell.edu.

Recommended CitationKimes, S. E., Barrash D. I., & Alexander, J. E. (1999). Developing a restaurant revenue-management strategy [Electronic version].Cornell Hotel and Restaurant Administration Quarterly, 40(5), 18-29. Retrieved [insert date], from Cornell University, School ofHospitality Administration site: http://scholarship.sha.cornell.edu/articles/462/

Developing a Restaurant Revenue-management Strategy

AbstractPropounding theories is one thing, but too often the intended beneficiary hasn’t the time or tools to checktheir usefulness. Here’s a case where researchers worked with a local restaurant to test their ideas, makerecommendations for improvement, and track the results.

Keywordsrevenue management, restaurant industry, strategies

DisciplinesFood and Beverage Management

CommentsRequired Publisher Statement© Cornell University. Reprinted with permission. All rights reserved.

This article or chapter is available at The Scholarly Commons: http://scholarship.sha.cornell.edu/articles/462

18

Developing a RestaurantRevenue-managementStrategyby Sheryl E. Kimes,Deborah I. Barrash, andJohn E. Alexander

Sheryl E. Kimes, Ph.D., is an associateprofessor at Cornell~ School of HotelAdministration <<sek6@cornell.edu)>,where Deborah 1. Barrash is a doctoral-

degree student «dib3@Cornell.edu».John E. Alexander is the president ofCBORD Solutions and co-owner ofCoyote Loco Cantina <<}ea@cbord.com».

Propounding theories is one thing, but too often the intended beneficiary hasn’t the

time or tools to check their usefulness. Here’s a case where researchers worked

with a local restaurant to test their ideas, make recommendations for improvement,

and track the results.

s explained in a paper pub-~ lished last year in this our-nal, the goal of restaurant revenuemanagement (RRM) is to maxi-mize revenue per available seat hour

(RevPASH) by manipulating priceand meal duration.’ In the first pa-per of the current series discussingthe steps of RRM,2 co-authorKimes discussed how RevPASHcan be measured and used to evalu-ate restaurant performance, and shepresented a five-step approach forimplementing RRM. In this paperwe explain how we developed arevenue-management strategy fora 100-seat casual restaurant in

Ithaca, New York.

Restaurant revenue managementcan be defined as selling the rightseat to the right customer at theright price and for the right dura-tion. The determination of &dquo;right&dquo;entails achieving both the mostrevenue possible for the restaurantand also delivering the greatestvalue or utility to the customer.Revenue management, or yieldmanagement, is commonly prac-ticed in the airline industry and, to asomewhat lesser degree, the lodgingindustry. Companies implementingrevenue management report in-

1 Sheryl E. Kimes, Richard B. Chase, SunmeeChoi, Philip Y. Lee, and Elizabeth N. Ngonzi,"Restaurant Revenue Management: ApplyingYield Management to the Restaurant Industry,"Cornell Hotel and Restaurant Administration Quar-terly, Vol. 39, No. 3 (June 1998), pp. 32-39.2 Sheryl E. Kimes, "Implementing Restaurant

Revenue Management: A Five-step Approach,"Cornell Hotel and Restaurant Administration Quar-terly, Vol. 40, No. 3 (June 1999), pp.16-21.

at CORNELL UNIV on September 15, 2014cqx.sagepub.comDownloaded from

19

creases in revenue of 2 to 5 percentover the results of prior procedures. 3

Restaurant operators can manipu-late two main strategic levers tomanage revenue: price and mealduration.4 Price is a fairly obvioustarget for manipulation, and manyoperators already offer price-relatedpromotions to augment or shiftpeak-period demand (e.g., early birdspecials, special menu promotions).5More-sophisticated manipulations ofprice include day-part pricing, day-of-week pricing, and price premi-ums or discounts based on party ortable size. Managing meal duration(i.e., speeding table turns) is a bitmore complicated, as we discuss in amoment. For example, meal durationdepends in part on the efficiency ofthe restaurant’s service cycle, as wellas on the idiosyncrasies of customer-arrival patterns and diners’ decidingto linger (or not) after the meal. 6

To develop an RRM program,managers should (1) establish thebaseline of performance, (2) under-stand the drivers of that perfor-mance, (3) develop a revenue-management strategy, (4) implementthat strategy, and (5) monitor thestrategy’s outcomes. In this paperwe discuss and illustrate how toestablish the baseline and understandits drivers, and how to develop arevenue-management strategy. In thenext paper we will discuss imple-mentation and evaluation issues.

The Study SiteAs part of our research we devel-

oped an RRM system for a small,casual restaurant in Ithaca, NewYork. Coyote Loco, a 100-seat res-taurant, serves California-styleMexican food, and is about a milefrom the Cornell University campus.Its average check is approximately$14. The restaurant has a full bar,and its signature margaritas are par-ticularly popular. The restaurant isopen from 11:00 AM to 11:00 PM

every day, and has a happy hourfrom 4:00 to 6:00 PM on Mondaythrough Friday. A manager is alwayson duty.

The menu comprises fewer than12 appetizers, 30 to 40 entrees, andabout half a dozen desserts. Serversare assigned stations that consist ofapproximately four tables. CoyoteLoco uses a Micros 2700 POS and

maintains. an electronic journal of alltransactions. The kitchen is smalland employs four line cooks (includ-ing one manager), two dish-machineoperators, and three food runners.

Food-preparation work is performedduring off-peak hours in a basementwork area. The main restaurant has72 seats (two 2-tops, fifteen 4-topsand one 8-top). The bar area, whichis used for dining on busy nights(and which also serves as the de factowaiting room for not-yet-seateddiners), has 27 seats (nine 2-tops andnine bar stools). During Ithaca’s s

warm-weather months, an outdoorpatio provides seating for an addi-tional 66 customers.

The owners and managers of

Coyote Loco agreed to help us ob-tain the data necessary to begin anRRM system. Moreover, they metwith us regularly to discuss imple-mentation alternatives and concerns.We began our work at the restaurantin September 1998 and have beenable to track its performance overtime. We followed the five-step ap-proach discussed in the first RRMpaper to develop a RM system andstrategies for Coyote Loco/

In the next section we describethe types of data and analysis neces-sary to establish a baseline, the toolsthat can be used to understand ac-tual service-cycle performance, andoperational tactics that are part of arevenue-management strategy. Weuse our experience at Coyote Locoto illustrate the discussion.

Step 1: Establish the BaselineTo develop an RRM program, res-taurant operators must collect de-tailed information on arrival pat-terns, meal times, and RevPASHpatterns. Before collecting data,several issues must be addressed-

including the appropriate levels ofdata, the sources of those data, andthe potential limitations and prob-lems associated with the data.

Level of detail. At a minimum,managers must collect arrival, meal-duration, and revenue data on aday-part basis. Without this level ofdetail, RRM will not have the nec-essary information with which towork. Hourly data can help theoperator to understand demandand revenue patterns, and they canhelp provide sufficient informationto fine tune the RRM system.Quick-service or family restaurantswith short meal times (i.e., servicecycles) may prefer to capture data

3 See, for example: Barry C. Smith, John F.Leimkuhler, and Ross M. Darrow, "Yield Man-agement at American Airlines," Interfaces, Vol. 22,No. 1 (1992), pp. 8-31; and Richard D. Hanks,Robert G. Cross, and R. Paul Noland, "Discount-ing in the Hotel Industry: A New Approach,"Cornell Hotel and Restaurant Administration Quar-terly, Vol. 33, No. 3 (June 1992), pp. 40-45.4 Ibid.; see also: Sheryl E. Kimes and Richard B.

Chase, "The Strategic Levers of Yield Manage-ment,"Journal of Service Research, Vol. 1, No. 2(1998), pp. 156-166.5 Identification and implementation of classic

revenue-management techniques is explained ina series of articles by Bill Quain et al. that waspublished in Cornell Hotel and Restaurant Adminis-tration Quarterly: "Revenue Enhancement, Part 1:A Straightforward Approach for Making MoreMoney," Vol. 39, No. 5 (October 1998), pp. 41-48;"Revenue Enhancement, Part 2: Making MoreMoney at Your Hotel," Vol. 39, No. 6 (December1998), pp. 71-79; "Revenue Enhancement, Part 3:Picking Low-hanging Fruit&mdash;A Simple Approachto Yield Management," Vol. 40, No. 2 (April1999), pp. 76-83; and "Revenue Enhancement,Part 4: Increasing Restaurant Profitability," Vol. 40,No. 3 (June 1999), pp. 38-47.6 For a discussion of the effects of altering the

service-cycle time, see: Christopher C. Muller,"A Simple Measure of Restaurant Efficiency,"Cornell Hotel and Restaurant Administration Quar-terly, Vol. 40, No. 3 (June 1999), pp. 31-37; and,for an analysis of arrival times, see: Brian Sill andRobert Decker, "Applying Capacity-managementScience: The Case of Browns Restaurants,"Cornell Hotel and Restaurant Administration Quar-terly, Vol. 40, No. 3 (June 1999), pp. 22-30. 7 Kimes, op. cit.

at CORNELL UNIV on September 15, 2014cqx.sagepub.comDownloaded from

20

for 15-minute periods. Such detailcan be helpful; however, the opera-tor must be sure that the additionalwork required to collect this infor-mation is worth the added benefit.

(In general, operations with shortdining times will benefit from moredetailed analysis.)

Sources of data. Data can becollected from the restaurant’s POS

(point of sale) system, through timestudy or actual observation, or bycustomer surveys. POS systems gen-erally collect a myriad of transactiondata such as arrival time, meal time,and customer spending, but con-verting that POS data to a usableform is often difficult and time-

consuming. Time study or observa-tion can provide detailed data onarrival time, course timing, meal

time, and revenues, but requires sub-stantial time to train observers and

actually to collect the data. Surveyscan be used to gauge customer reac-tion and sensitivity to price, servicequality, and overall atmosphere, butadministering surveys properly anddeciphering the data take a lot oftime and careful training.

Potential data problems. Asjust noted, data sources generally areimperfect. While POS data containdetailed information on all transac-

tions, for example, the opening timeof a check may not reflect the

guests’ true arrival time (how longdid they wait to be seated?) and theclosing time of the check may notaccurately indicate when the guestsleft the dining room (did the cashierring up the check as soon as the

table was vacated?). The type ofcontrol system used (server bankingor cashier banking) can also influ-ence the accuracy of check-closingtimes. For instance, restaurants thatuse cashiers may find that all checksare closed at the end of the cashier’s sshift. This renders the POS datauseless for duration estimates.A time-study approach to col-

lecting data can yield accurate ar-rival and departure times, and cangenerate a good level of detail re-garding meal timing (drinks to ap-petizers to entrees to desserts), butwe found that it is difficult for theobserver to record accurately alltransactions for multiple parties.Moreover, while POS systems pro-vide information on all transactions,time-study data cover only a sample

at CORNELL UNIV on September 15, 2014cqx.sagepub.comDownloaded from

21

of transactions. That needn’t be a

major drawback, but care must betaken to ensure that a representativesample of meal periods and days arebeing investigated and recorded.

Coyote Loco’s baseline. Thefirst thing we did at Coyote Locowas to collect data from its POS

system. Once we had these data, weperformed time studies of the din-ner period for six nights. The result-ing data were analyzed to develophourly arrival rates, meal times, andRevPASH. All results presented inthis paper are from October 1998,our first full month of observationat Coyote Loco.We did not want to go through

all of Coyote Loco’s POS transac-tions manually, so we used softwarecalled Monarch, a report writer, toextract the required data. We ex-tracted data on the date, the checknumber, the transaction time, theparty size, and the transactionamount. Each party at the restaurant

usually had multiple transactions fortheir meal-including when thecheck was opened, when orderswere entered, and when the checkwas closed. (In a few cases therewere just two transactions: when thecheck was opened with the entireorder and when it was closed at theend of the meal. Any voided checkswere excluded from the study.)

The usable data were then trans-ferred from Monarch to Microsoft

Excel, where the multiple transac-tions were condensed into a singlerecord for each party. Each recordcontained information on the date,the check number, the starting time,the closing time, the party size, andthe check amount for each party.Data analysis to find the number ofhourly arrivals, the mean and stan-dard deviation of meal duration, andthe hourly RevPASH was per-formed using Microsoft Excel andMicrosoft Access software.

Arrivals. We determined and

graphed the distribution of the

number of arrivals for each hour ofeach day during October. As shownin Exhibit 1, Coyote Loco was busi-est on Friday and Saturday nightsbetween 6:00 and 9:00 PM. Sundaywas the next busiest night, with theother nights of the week beingmuch slower. The lunch-time busi-ness was relatively slow every day.

One of the problems with theactual-arrivals data that we collectedis that the number of arrivals doesnot represent the true, uncon-strained demand. (Unconstraineddemand can be defined as the num-ber of customers a restaurant couldhandle if its capacity was unlim-ited.)’ For instance, some customerswho wanted to dine at the restau-rant may have been turned away,walked out after waiting a bit, orsimply observed that the restaurantwas busy and never stopped at all.A variety of methods, both math-

ematical and managerial, have beenused by hotels and airlines to esti-mate this unconstrained demand. To

gauge the true demand, for example,restaurants can have someone countthe number of guests who walk outand track the number of reserva-tions that are turned away duringbusy periods. Doing so will providean estimate of the unconstraineddemand. At Coyote Loco, we foundthat approximately 20 customerswalked out each night on Fridaysand Saturdays.

Meal duration. Analysis of thePOS data gave nightly averages ofhow long it took parties to finishtheir meals, but we were also in-terested in the meals’ standard devia-tion (a measure of the variation ofguests’ meal times). We were inter-ested in reducing both the averagemeal time and the variability in themeal time.

To create a revenue-

management program

managers must, at a

minimum, collect arrival,

meal-duration, and revenue

data on a day-part basis.

8 Eric B. Orkin, "Wishful Thinking and RocketScience: The Essential Matter of CalculatingUnconstrained Demand for Revenue Manage-ment," Cornell Hotel and Restaurant Administration

Quarterly, Vol. 39, No. 4 (August 1998), pp. 15-19.

at CORNELL UNIV on September 15, 2014cqx.sagepub.comDownloaded from

22

Those results were illuminating.On average (using October 1998data), meals took a little over onehour-with a standard deviation ofabout 30 minutes. Statistically, the30-minute standard deviation repre-sents an unusually high variability indining time. That indicates thatCoyote Loco’s managers had littlecontrol over the length of the mealand that customers may have con-sidered the meal time to be incon-sistent (i.e., sometimes it took 35minutes to eat and sometimes ittook an hour and a half to eat-forreasons that weren’t always underthe customer’s control).When calculating the mean and

standard deviation, we eliminatedthose transactions that took under20 minutes and over 3.5 hours. Thishad the effect of removing all take-out business and bar sales, whichgave us a more reliable estimate ofthe mean for the dining room.

The mean and standard deviationvaried by day of week and time ofday (Exhibit 2). Dining times variedfrom 51 minutes on Saturday nightsat 9:00 PM to one hour and 17 min-utes on Tuesday evenings at 5:00 PM.The standard deviation varied from18 minutes on Sundays at 7:00 PMand 9:00 PM to 41 minutes on Tues-

day at 5:00 PM.RevPASH. To calculate hourly

RevPASH, we first calculated therevenue from all the transactions

that began during that hour anddivided by the number of availableseats. As expected, the RevPASHvaried, ranging from $0.76 on Mon-days at 5:00 PM to $7.33 on Fridaysat 8:00 PM. The highest RevPASHwas achieved on Fridays and Satur-days between 6:00 and 9:00 PM.Another busy period was on Sun-days between 6:00 and 7:00 PM.All other nights of the week had alower RevPASH than did Fridays,Saturdays, and Sundays, and lunch-time business yielded a surprisinglylow RevPASH-usually less than$1.00 (Exhibit 3).

Meal-course timing. A timestudy was performed during dinnerfor six nights in an attempt to un-derstand the timing of the meal andto observe any possible bottlenecksin the system. We first sketched thefloor layout of the restaurant andmade sure that we were familiarwith the table-numbering system.Data were collected on the follow-

ing variables:(1) The table number.(2) Number ofguests. This created a

problem at times because notall members of the party wouldbe present when the party was

initially seated.(3) Time guests were seated. Note

that this time is different thanthe time the guests arrived.Due to the high volume ofbusiness during dinner, it was

almost impossible to identifywhen particular parties arrivedand when they were seated.

(4) Timeguests weregreeted. Thiswas the time the server ap-proached the table to greetthe guests and take drinkorders.

(5) Time the appetizer was delivered.Not all guests ordered appetiz-ers, so this time was not re-corded for all customers.

(6) Time the entrée was delivered.(7) Time dessert was delivered. Fol-

lowing completion of theentree, the server approachesthe table and asks whether the

guests would like coffee ordessert. Not all guests ordereddessert, but for those that did,we timed when the dessertorder was delivered.

(8) Time the guests left. This wasdefined by when the guestsleft the table.

Student observers armed withwatches sat in an unobtrusive spotand timed the elements and trans-actions listed above. They used asimple form to record the times.

The time studies helped verifythe results from the POS data (Ex-hibit 4). As expected, average diningtimes were slightly longer thanthose found with the POS databecause the time study measuredthe time between when the cus-tomer arrived and departed thetable rather than when the checkwas opened and closed.

Approximately 15 to 20 minuteselapsed from when the server firstapproached the table to when appe-tizers were delivered. Entrees weredelivered a little over 20 minutesafter the appetizer was delivered.For those customers who ordered

dessert, that took about another halfhour. On average the total diningtime (arrival at the table to depar-ture) was an hour and 12 minutes,with a standard deviation of about25 minutes.

at CORNELL UNIV on September 15, 2014cqx.sagepub.comDownloaded from

23

The time studies (left) helped to verifythe results from the POS data (above).Average dining times recorded duringthe time study were slightly longer thanthose defined by the POS data becausethe time study measured the timebetween when the customer arrived anddeparted the table rather than when thecheck was opened and closed.

at CORNELL UNIV on September 15, 2014cqx.sagepub.comDownloaded from

24

Summary of findings. Wewere not at all surprised to find thatFriday and Saturday nights werebusy and profitable, but the lowRevPASH and head counts that werecorded for the other nights and alllunch periods were unexpected. Theaverage meal time of almost an hourand a quarter seemed right, but wewere alarmed at the high standarddeviation of the meal time. Armedwith this knowledge and the resultsof the time study, we decided toproceed to the next step and studythe possible causes of our findings.

Step 2: Understand the CausesA variety of tools can be used tohelp managers understand the un-derlying causes behind operationalproblems, including service blue-prints,9 process analysis,&dquo; andfishbone diagrams.ll Those tech-niques are fairly simple to imple-ment and have been widely used intotal-quality-management programs.

Service blueprints can be used tographically illustrate a service pro-cess. The steps in the process are

mapped and the connections be-tween steps are identified. One ofthe key strengths of the serviceblueprint is the identification ofpotential delays and failure points.

Fishbone diagrams, sometimesreferred to as cause-and-effect dia-

grams, can be used to help managersunderstand the causes of a problem.The head of the fish is the problemand the bones of the fish are the

possible causes. Typically, possible

-

Icauses are divided into the four

categories of methods, materials,labor, and equipment, but thesecategories are not meant to berestrictiveThe causes at Coyote Loco.

We developed a blueprint for Coy-ote Loco and tried to identify po-tential sources of failure (Exhibit 5).Following this, we developed afishbone diagram to try to identifythe possible causes for the mostimportant failure points.

The service process at CoyoteLoco is typical of many casual res-taurants except that tables are pre-set with chips, salsa, and water. Thesteps involved in Coyote Loco’sservice (its blueprint) and the po-tential failure points in each of thosesteps are illustrated and discussed inExhibit 5.We decided that the biggest

problem at Coyote Loco was thelength and variability of dining time.Reducing the mean dining timewould be difficult without first re-

ducing the standard deviation of themeal time.l3 If we could reduce the

variation, we felt that we could alsoreduce the average meal time.We developed a fishbone diagram

in which the problem was definedas the high standard deviation ofmeal duration, and the problemcategories were personnel, informa-tion, equipment, methods, andproducts (see Exhibit 6, overleaf).For each of the categories, we de-tailed the possible related causes. Wefound that the main issues influenc-

ing the standard deviation of mealtime at Coyote Loco were related topersonnel and methods. Althoughthere were some equipment andproduct issues in the kitchen, mostof the dining-time problems that werecorded during our six nights ofobservations were in the front of the

house and involved personnel andprocedural issues. Those two sets ofproblems are discussed below.

Personnel Problems

Training and variable skilllevels. The combination of incon-sistent training and variableemployee-skill levels contribute toinconsistent service. The restaurantoffers no formal training programfor either new hires or existing em-ployees. In addition, standard oper-ating procedures are not well devel-oped, and what does exist is notclearly communicated to employees.

Host stand and greet time.The host is responsible for the flowof guests into the dining room-one of the most important functionsin the restaurant. At Coyote Loco,however, the host was also respon-sible for take-out orders, seating, andother nonessential jobs. Therefore,during peak hours of operation, thehost was not able to attend to the

primary responsibilities of greetingguests and managing dining-roomflow.

Procedural DifficultiesGreet time. Once guests are

seated, a server approaches the tableand offers to take a drink order. Thetime for this varied widely (mean:two minutes and thirty seconds,standard deviation: two minutes).The inconsistency occurred becauseof lack of communication betweenthe host and server, servers’ lack of

attentiveness, and servers’ beingoccupied with other activities.

Bussing tables. Coyote Locoemploys servers, food runners, andbussers. Even when customers were

waiting for a table, however, tablesoften sat for several minutes before

being cleared. One of the majorcauses was the inconsistency ofserver pre-bussing. When serversdidn’t remove items from the table,bussers had more work and were

delayed in resetting the table andseating the next party.

9 G. Lynn Shostack, "Designing Services ThatDeliver," Harvard Business Review, January-February 1984, pp. 133-139.10 Sheryl E. Kimes and Stephen A. Mutkoski,

"The Express Guest Check: Saving Steps withProcess Design," Cornell Hotel and RestaurantAdministration Quarterly, Vol. 30, No. 2 (August1989), pp. 21-25.

11 See: D. Daryl Wyckoff, "New Tools forAchieving Service Quality," Cornell Hotel andRestaurantAdministration Quarterly, Vol. 25, No. 3(November 1984), pp. 78-91; and U. Apte andC. Reynolds, "Quality Management at KentuckyFried Chicken," Interfaces, Vol. 25, No. 3 (1995),pp. 6-21.

12 See: Wyckoff, 1984; and Apte and Reynolds,1995.

13 Eliyahu M. Goldratt and Jeff Cox, The Goal(Great Barrington, MA: North River Press,1984).

at CORNELL UNIV on September 15, 2014cqx.sagepub.comDownloaded from

25 at CORNELL UNIV on September 15, 2014cqx.sagepub.comDownloaded from

26

Reservations. Although reserva-tions constitute less than 25 percentof Coyote Loco’s business, we no-ticed that they can create a problemduring busy times. When a tablebecomes available near the reserved

time, the host &dquo;holds&dquo; the table evenwhen other guests are waiting to beseated. The held table may sit emptyfor some time, especially when thereserving party arrives late. We ob-served tables that were left empty forover half an hour even when other

guests were waiting.Table management. The host is

assigned multiple duties and cannotadequately concentrate on tablemanagement. Because of the greatvariability in diners’ meal duration,estimates of the waiting time offeredto arriving customers are unreliable.

In addition, there is usually a timelag between when a table is availablefor seating and when the host canfind and seat the next party.Communication with food

runners. Coyote Loco uses foodrunners to deliver items to tables.Since the food runners do not takethe orders from customers, yet occa-

sionally deliver food to the tables,they sometimes deliver the wrongfood to the table, which causes fur-ther delays in the service process.

Suggestive selling. Althoughsuggestive selling is a good tech-nique for increasing the averagecheck, it may not be appropriate atall times. We saw that during peakperiods when customers are waiting,the servers were nonetheless still

suggesting appetizers, desserts, and

additional drinks. The additional

profit from those items may not beas high as that of serving anotherparty.

Step 3: Develop a Revenue-managementStrategyAfter identifying the causes of un-even service and unpredictable mealtimes, we developed recommenda-tions for Coyote Loco’s managersand owners. Our intent was to keepthe recommendations as simple aspossible, with an aim toward in-creasing RevPASH. We realized thatto succeed we needed the owners’and managers’ commitment, andthat we also needed to present ourideas in a way that would be appeal-ing to Coyote Loco’s front-of-the-house employees.

at CORNELL UNIV on September 15, 2014cqx.sagepub.comDownloaded from

27

Some of our recommendationswere appropriate for all levels of de-mand while others would depend onhow busy the restaurant was. Belowwe outline our overall recommenda-tions and then detail the tactics of the

revenue-management strategy we

developed.Overall recommendations:

(1) Improve training. Front-of-the-house employees need a formaltraining program on service pro-cedures and revenue-managementtactics. (Specific information on

those procedures and tactics aregiven below.) .

(2) Develop standard operating proce-dures. The service-blueprint andfishbone diagrams helped illus-trate the lack of standard oper-ating procedures, and so werecommended the following:

. Greet time. All parties should begreeted within one minute ofbeing seated.

. Appetizer delivery. All appetizersshould be delivered within fiveminutes of ordering. ,

· Entree delivery. All entrees shouldbe delivered within 12 minutesof ordering or five to seven min-utes after the appetizer has beencleared.

· Course timing. Entrees should notbe delivered until the appetizerhas been cleared.

· Pre-bussing. When possible, serversshould remove any unnecessarydishes from the tables. Serversshould always be carrying some-thing when they walk back tothe kitchen.

at CORNELL UNIV on September 15, 2014cqx.sagepub.comDownloaded from

28 at CORNELL UNIV on September 15, 2014cqx.sagepub.comDownloaded from

29

. Payment timing. Servers shouldbring the check to customerswithin two minutes of the re-

quest. After the dessert has beenordered (or refused) the servershould prepare the check forimmediate delivery to the cus-tomer upon request (any addi-tional items ordered can alwaysbe added and the check re-

printed before delivery). Theentire payment process shouldtake less than five minutes.

. Clear time. All dirty tables shouldbe bussed continuously duringservice (especially at the end ofthe meal) so that when the cus-tomer leaves the table, there willbe a minimal number of itemsthat still need to be removed. Alltables should be cleared and resetwithin one minute of customer

departure.(3) Improved table management. Man-

agers need to improve their abil-ity to track which course a tableis on and be able to accuratelypredict when the table will beavailable for the next party.Improved communication be-tween servers and the hostand a manual tracking systemshould help with better tablemanagement.

Revenue-management Strategiesfor Coyote LocoUsing principles drawn from otherindustries (see the box on page27), we analyzed Coyote Loco’shourly RevPASH data, classifiedeach hour as either hot, warm, orcold, and developed revenue-management strategies (Exhibit 7).Demand was cold for all lunch pe-riods, on Mondays and Wednesdaysfrom 5:00 to 6:00 PM and 8:00 to10:00 PM, and on Thursdays andSaturdays after 9:00 PM. Demandwas hot on Thursdays and Sundaysfrom 6:00 to 7:00 PM and on Fridayand Saturdays from 6:00 to 9:00 PM.

After identifying cold and hotmeal periods, we developed strate-gies to help Coyote Loco increaseits RevPASH (each of those strate-gies is discussed in Exhibit 7). Avariety of options is available, andexecution of each of the optionsvaries according to demand levels.For example, suggestive selling wasconsidered a good tactic duringcold periods, but an undesirabletactic during hot periods. For man-agers to use these tactics, therefore,they must know when they are in(or about to enter) a cold or hotperiod and be able to alert employ-ees of the situation.

A Plan of Attack

Restaurant operators can use the

methodology that we just presentedto develop a revenue-managementstrategy for their restaurant. By es-tablishing the baseline of revenueperformance and then understand-ing the causes of that performance,managers can develop revenue-management tactics to help increaseRevPASH during both cold and hottimes. For example, during coldtimes managers will want to con-centrate on raising the averagecheck, while during hot times man-agers should focus on increasingtable turns and serving as manypeople as possible.

Coyote Loco provided us with anexcellent study site, and we aregrateful to its employees, managers,and owners for the chance to testour ideas. In the next paper we willdiscuss the actual implementation ofrevenue-management strategies atCoyote Loco and present the impactof those tactics on its RevPASH andfinancial performance. We’ll exam-ine methods of establishing appro-priate incentive and training pro-grams for servers, managers, and

bussers, and present suggestions onhow to monitor the success of the

revenue-management system. CO

We found that the restaurant

required one set of strategies

for slow times and an entirely

different set for busy times.

at CORNELL UNIV on September 15, 2014cqx.sagepub.comDownloaded from