Kimes rest.revenue mgt.1.26.04

CHR Reports

The Center for Hospitality ResearchU N I V E R S I T YA T C O R N E L L

RestaurantRevenueManagement

ORNELLC

by Sheryl Kimes, Ph.D.

2 • Cornell Center for Hospitality Research

Restaurant Revenue Management

School of Hotel Administration Cornell University Ithaca, NY 14853

ARAMARKDK Shifflet & Associates

ehotelier.comGlobal Hospitality

Resources, Inc.Hospitality World

The Center for Hospitality ResearchU N I V E R S I T YA T C O R N E L L

Thank you to our generous corporate supporters!

AIG Global Real Estate Investment

Bartech Systems International

Cendant Corporation

Cornell Hotel Society Foundation

Four Seasons Hotels and Resorts

Kohinoor

Marsh’s Hospitality Practice

Nestlé

Willowbend Golf Management

Wyndham International

Lodging HospitalityNational Hotel Executive

MagazineResort+Recreation

RestaurantEdge.comShibata Publishing Co. Ltd.

AIG

Partners and Sponsors

MARSHThe world’s #1 risk specialist™

hospitalitynet.orgHotel Asia Pacific

Hotel ChinaHotel Interactive, Inc.

Hotel ResourceLodging Magazine

ORNELLC

FriendsSmith Travel Research

The Hospitality Research Group of PKF Consulting

The Lodging ConferenceTravelCLICK

WiredHotelier.com

Cornell Center for Hospitality Research • 3


THE PRINCIPLES OF REVENUE MANAGEMENT CAN BE APPLIED TO RESTAURANTS, given that the restaurant’s

unit of sale is the time it takes for a complete meal cycle, rather than just the meal itself. Moreover,

restaurants have classic characteristics that invite revenue-management strategies (those characteristics

being relatively fixed capacity, perishable inventory, a demand inventory, time-variable demand,

appropriate cost structure, and segmentable customers). When a restaurant’s operation is gauged by

the time-related measure called revenue per available seat-hour, or RevPASH, managers can analyze

operations and menus to improve that statistic. Using RevPASH allows managers to capture more of

the restaurant’s actual performance in their analysis than does average check or typical food- or labor-

cost percentages.

Restaurateurs have available two general sets of strategic levers to build RevPASH, which is the

goal of restaurant revenue management. Those key levers are duration management and demand-

based pricing. Pricing approaches involve setting prices according to customers’ demand characteris-

tics, such as whether they are willing to dine off peak or whether they are not as concerned about

price as they are about the dining experience. Pricing strategies must be approached carefully to avoid

the appearance that the restaurant seeks to gain at the expense of customers (which customers view as

unfair). Typically, this means adjusting menus to offer discounts and specials that, while they offer

more value to the customer, may well make as strong a contribution to revenue as other, higher-price

menu items that cost more to serve. That is the province of menu engineering.

Duration management helps restaurateurs gain control of the most erratic aspect of their

operation, which is the length of time customers sit at a table (including the rate at which customers

will arrive to occupy that table). Among the tactics available for duration management are reducing

the uncertainty of arrival, reducing the uncertainty of duration, and reducing the time between meals.

Whether the restaurant accepts reservations or serves customers as they arrive, its manager needs to

have a sense of when customers are most likely to appear. That is a matter of creating a forecast

based on the restaurant’s history and of carefully managing reservations (if the restaurant accepts

them). Although a restaurateur cannot directly control the customer’s use of a table, careful process

control and analysis can make the restaurant’s operations (including menu design, kitchen operation,

and service procedures) as effective as possible for moving the meal along, and perhaps indicating to

the customer when it is time to leave.

As an example, Chevys Arrowhead, a Phoenix-area restaurant, used revenue-management levers

to improve its revenue through process control. Seeking to augment revenue and also to improve

customer service, the restaurant analyzed its operations and its customers’ characteristics. It found

that its table mix (mostly 4-tops) was inappropriate for its customer base (mostly singletons and

couples). It also found that it could tighten up its post-meal procedures, particularly those involving

settlement. The restaurant was reconfigured, servers were retrained, and certain key positions were

added. The result was an increase in revenue (from higher occupancy) that paid for the increased

capital costs in one year. The revenue improvement in this instance was to guests’ advantage, since

menu prices were not changed as part of this revenue-management implementation.


by Sheryl E. Kimes

Executive Summary



CHR Reports:Restaurant Revenue Management

is produced for the benefit of the hospitality industry byThe Center for Hospitality Research at Cornell University

Gary M. Thompson, Executive DirectorGlenn Withiam, Director of Publication Services

Restaurant Revenue Managementis CHR Reports, Vol. 4, No. 2 (February 2004)

Single copy price US$50.00

Copyright © 2004 by Cornell University

Advisory Board

Roger Cline, Chairman and CEO, Roundhill HospitalityRichard Cotter, EVP, Hotel and F&B Operations, Wynn ResortsBjorn Hanson, Ph.D., Global Hospitality Industry Managing Part-

ner, PricewaterhouseCoopersJo-Anne Kruse, SVP, Human Resources, Cendant InternationalCraig Lambert, President, Fieldstone GroupMark V. Lomanno, President, Smith Travel ResearchGordon S. Potter, Ph.D., Associate Professor, Cornell UniversityJanice L. Schnabel, Senior Vice President, Marsh’s Hospitality

PracticeDavid A. Sherf, SVP, Real Estate and Investment Analysis, Hilton

Hotels CorporationJudy A. Siguaw, Ph.D., J. Thomas Clark Professor of

Entrepreneurship and Personal Enterprise, Cornell UniversityBarbara Talbott, EVP Marketing, Four Seasons Hotels and ResortsElaine R. Wedral, President, Nestlé R&D Center and Nestlé PTC

New MilfordR. Mark Woodworth, Executive Managing Director, The Hospitality

Research GroupPeter Yesawich, Ph.D., President and CEO, Yesawich, Pepperdine,

Brown & Russell



My objective in this report is todevelop a framework for such a theory andto discuss and demonstrate how that theorycan work in practice. After reviewing thenecessary conditions for revenue manage-ment, the strategic levers available forrevenue management, I will explain howthose strategic levers, along with sometactical tools, can be applied to restaurants.

Defining Revenue ManagementRevenue management is the application ofinformation systems and pricing strategies toallocate the right capacity to the right cus-tomer at the right place at the right time.1 Inpractice, revenue management has meantdetermining pricing according to predicteddemand levels so that price-sensitive custom-ers can achieve a favorable price by purchas-ing at off-peak times, while price-insensitive

customers will be able to make their pur-chases at the peak times that they desire.The application of revenue management hasbeen most effective when it is applied tooperations that have the following character-istics: relatively fixed capacity, perishableinventory, a demand inventory, time-variabledemand, appropriate cost structure, andsegmentable customers.2 As I explain next,these attributes are generally found in someform or another in the restaurant industry.

Relatively fixed capacityRelatively fixed capacityRelatively fixed capacityRelatively fixed capacityRelatively fixed capacity. A restaurant’scapacity can be measured by number ofseats, kitchen size, menu items, or staffinglevels. Most restaurant operators’ ap-proaches to optimizing revenue primarilyinvolve filling the seats to capacity andturning tables as quickly as possible, but thateffort can be limited by the kitchen, themenu design, or staff members’ capabilities.

RESEARCH IN REVENUE MANAGEMENT has traditionally addressed the theoreticaland practical strategic problems facing airlines and hotels, among otherindustries, but it has given little consideration to the restaurant industry. The

restaurant business is similar enough to hotel and airline operations that restaurantsshould be able to apply revenue-management-type practices in a strategic fashion, butthe applications have so far been mostly tactical. A broad theory of revenue manage-ment would permit restaurant operators to gain the benefits of strategic revenuemanagement that they currently lack.

1 B.A. Smith, J.F. Leimkuhler, and R.M. Darrow, “Yield

Management at American Airlines,” Interfaces, Vol. 22, No. 1

(1992), pp. 8–31.

Restaurant Revenue Managementby Sheryl E. Kimes

2 S.E. Kimes, R.B. Chase, S. Choi, E.N. Ngonzi, and P.Y. Lee,

“Restaurant Revenue Management,” Cornell Hotel and Restaurant

Administration Quarterly, Vol. 39, No. 3 (June 1998), pp. 32–39.



Seating capacity is generally fixed overthe short term, although restaurants havesome flexibility to crowd a table with anadditional seat if necessary, and therestaurant’s cost of adding additional capac-ity in the forms of tables or seats (say, byreconfiguring the dining room or seatingdiners in the lounge) is lower than that ofmany businesses that typically use revenuemanagement. Most restaurants have a fixednumber of tables, but can vary the numberof seats depending on the mix of party sizes.In addition, some restaurants might increasecapacity during pleasant weather by usingoutdoor dining.

Perishable inventoryPerishable inventoryPerishable inventoryPerishable inventoryPerishable inventory. One might thinkof a restaurant’s inventory as being its supplyof raw food, but most of that is not perish-able until it is removed from the freezer or issitting on the receiving dock. Instead, arestaurant’s inventory is best thought of astime—or, in this case, the time during whicha seat or table is available. If a seat is notoccupied for a period of time, that part ofthe restaurant’s inventory perishes. This isthe key to the strategic framework that Ipresent here, and it is the element that Ibelieve has been missing in most approachesto restaurant revenue management. Insteadof counting table turns or revenue for a givenday part, restaurant operators shouldmeasure revenue per available seat hour(RevPASH). This measure captures the timefactor involved in restaurant seating.

Demand inventoryDemand inventoryDemand inventoryDemand inventoryDemand inventory. Demand can beinventoried either by taking reservations orby creating queues of waiting guests. Mostindustries that employ revenue managementuse reservations (or advance sales) to create ademand inventory. Reservations are valuablebecause they give an operator the opportu-nity to sell and control his or her inventoryin advance of consumption (often withadvance payment for that consumption). Inaddition, companies that take reservationshave the option to accept or reject reserva-

tion requests. During high-demand periods,operators may choose to reject low-valuerequests, for instance, while during low-demand periods, managers may choose toaccept such requests.

While many restaurants take reserva-tions, a majority of restaurants do not do so,preferring instead to manage a queue whendemand exceeds supply. Indeed, whilereservations help a restaurant sell andcontrol its inventory, they are not withoutproblems. As I discuss later, no-shows, late-shows, and short-shows are all problems inthe restaurant industry, which is why somerestaurants choose to rely on walk-in busi-ness rather than take reservations.

Time-variable demandTime-variable demandTime-variable demandTime-variable demandTime-variable demand. Setting asidecarry-out activities as a separate business,restaurant demand consists of guests whomake reservations and guests who walk in.Both forms of demand can be managed,albeit with different strategies. Strategicdifferences notwithstanding, guests whomake reservations and those who walk inconstitute an inventory from which managerscan select the most profitable mix of custom-ers. To do this, however, restaurant opera-tors must forecast customer demand andmanage the revenue generated from thatdemand.

Restaurant demand has two compo-nents: namely, the timing of the demand andthe duration of that demand (that is, howlong the meal lasts). As in most businesses,customer demand varies by time of year, dayof week, and time of day. For restaurants,dinner demand may be higher on weekends,during summer months, or at particulartimes during the lunch or dinner periods.Restaurant operators must be able toforecast time-related demand so that theycan make effective pricing and table-alloca-tion decisions.

A special factor for restaurant opera-tors is that they have to reckon with thelength of time a party stays once it is seated.



This factor is analogous to a hotel’s having toforecast the number of guests who will stayan additional (unscheduled) night, but thehotel still is selling an integral room-night tothe stayover guest and not dealing with theoften-unpredictable period that diners willstay at a table. If restaurant managers canaccurately predict meal duration, they canmake better reservation decisions and givebetter estimates of waiting times for walk-inguests.

Appropriate cost structureAppropriate cost structureAppropriate cost structureAppropriate cost structureAppropriate cost structure. Like hotels,restaurants have a cost structure that featuresrelatively high fixed costs and fairly lowvariable costs, although it’s true that a menuitem’s food-cost percentage is usually higherthan the variable-cost percentage associatedwith a hotel room. Like hotels, restaurantsmust generate sufficient revenue from eachsale to cover variable costs and offset at leastsome fixed costs. Nevertheless, restaurants’relatively low variable costs allow for somepricing flexibility and give operators theoption of reducing prices during low-demand times.

Segmentable customersSegmentable customersSegmentable customersSegmentable customersSegmentable customers. Like hotels,restaurants have some customers who areprice sensitive and others who are not. Forexample, certain customers (for instance,students, families with small children, orpeople on fixed incomes) may be willing tochange their dining time in exchange for adiscounted price. Conversely, other custom-ers are not at all price sensitive and are oftenwilling to pay a premium for a desirable tableat a desirable time. Restaurant operatorsneed to be able to identify these two seg-ments and design and price services todifferentiate them and meet their needs.

Measuring Success:The Case for RevPASH

Restaurant managers are typically evaluatedby such measures as the check average andthe food- and-labor-cost percentage that themanager has been able to achieve. 3 Neither

of those measures captures sufficientinformation about a restaurant’s revenue-generating performance.

Having a restaurant manager concen-trate only on a high average check, forinstance, is equivalent to a hotel’s focusingsolely on achieving a high average room rate.Just as ADR omits consideration of occu-pancy, a restaurant’s revenue performancecannot be evaluated without information onseat occupancy. A high average check mayeven be an indication of detrimental prac-tices in times of strong demand if, forexample, customers are encouraged to lingerover their meal with coffee and dessert whileother parties wait for a table.

Similarly, a manager’s achievingspecified food-cost and labor-cost percent-ages is laudable, but that does not tell theentire story. In particular, the margin is not ameasure of profitable use of a restaurant’scapacity. A restaurant manager can do agood job of maintaining margins and still beunprofitable, especially since an overempha-sis on margins can lead to a propensity tofocus unduly on minimizing costs. Again,reducing cost is fine, but not when thatcauses reduced revenue due to disgruntledcustomers.

The extent to which available seats areoccupied is another commonly appliedmeasure of success, since a busy restaurant isgenerally a revenue-producing restaurant.Relying on seat occupancy as a measure ofsuccess suffers from problems similar tothose of relying on hotel-room occupancy (inthe absence of consideration of ADR),because high use does not necessarily meanhigh revenue. A restaurant can run at 90-percent of capacity and still not make moneyif menu items are sold at too low a price, forexample, or, more generally, if checkaverages are too low.

3 Much of this section comes from S.E. Kimes, “Implementing

Restaurant Revenue Management: A Five-step Approach,” CornellHotel and Restaurant Administration Quarterly, Vol. 40, No. 3

(1999), pp. 16–21.



Because it embraces capacity use andcheck averages, revenue per available seat-hour (RevPASH) is a much better indicatorof the revenue-generating performance of arestaurant than are the commonly usedmeasures that I just discussed. RevPASHindicates the rate at which revenue is gener-ated and captures the trade-off betweenaverage check and facility use. If occupancypercentage increases even as the averagecheck decreases, for instance, a restaurantcan still achieve the same RevPASH.Conversely, if a restaurant can increase theaverage check, it can maintain a similarRevPASH with slightly lower seat occupancy.

Exhibit 1 gives an illustration of thisprinciple. The four hypothetical restaurantsin the exhibit all have the same RevPASHfor the hour in question ($7.20), but eachachieves it in a different manner. RestaurantA has a seat occupancy of 40 percent and anaverage check of $18, while Restaurant Dhas a seat occupancy of 90 percent but anaverage check of $8. Restaurants B and Calso achieved a RevPASH of $7.20, but withvarying seat occupancy and average-checkstatistics.

The easiest way to calculate RevPASHis to divide revenue (or profit) for the

desired time period (e.g., hour, day-part,day) by the number of seat-hours availableduring that interval. For example, assumethat a 100-seat restaurant makes $3,000 onFridays between 6:00 and 8:00 PM. ItsRevPASH for those hours would be $15($3,000/100 seats/2 hours).

Managing Demand: Strategic LeversRestaurants appear to possess the conditionsnecessary for revenue management, butthere is little evidence that most restaurantsuse a strategic approach for applyingdemand-management mechanisms. 4 Asuccessful revenue-management strategy ispredicated on effective control of customerdemand. I have alluded to the two strategiclevers that restaurant managers have at handto manage demand, and thus, revenue.Those are duration management anddemand-based pricing.

Duration managementDuration managementDuration managementDuration managementDuration management. As I men-tioned above, restaurant operators typicallyface an unpredictable duration of customeruse, which inhibits their ability to managerevenue. To allow for better revenue-management opportunities, restaurantmanagers must increase their control overthe length of time that customers occupytheir seats. To do this, restaurateurs canrefine the definition of duration, reduce theuncertainty of arrival, reduce the uncertaintyof duration, or reduce the amount of timebetween customers’ meals (see Exhibit 2).

Redefining durationRedefining durationRedefining durationRedefining durationRedefining duration. The length oftime that guests use a table is usually mea-sured by the number of minutes or hoursthat they actually occupy that table or by theevents relating to a meal (e.g., by the courseor by the full meal). In either case, therestaurateur must know how long a typicalguest will stay at a table in a given day part ormeal. When duration is defined in terms of

4 S.E. Kimes and R.B. Chase, “The Strategic Levers of Yield

Management,” Journal of Service Research, Vol. 1, No. 2 (1998),

pp. 156–166; Kimes et al., op. cit.

Exhibit 1Sample calculations of RevPASH

Seat Average checkRestaurant occupancy (per person) RevPASH

A 40% $18.00 $7.20

B 60% $12.00 $7.20

C 80% $9.00 $7.20

D 90% $8.00 $7.20

Note: Mean dining times are assumed to be one hour in all cases.



a meal rather than as the time to complete ameal, the operator must be able to predictmeal length so that selling a meal essentiallybecomes selling a certain length of time.

On the surface, restaurants sell meals,rather than explicitly selling time—although afew restaurants actually do sell blocks of time(e.g., seating parties every two hours, with areminder to leave when the time is up). Inreality, restaurant operators sell time in theform of meals of reasonably predictablelength. This could be done directly, intheory, by asking customers how long theywill need the table when they make areservation or request a table, but such anapproach would require a radical change inthinking for both restaurateurs and custom-ers. Even though that approach wouldexplicitly change the definition of durationfrom the meal itself to the time involved ineating the meal, the tactic would put off mostguests—other than those who have a specificdate or appointment after the meal.

Instead, most restaurant operatorsmust keep track of the length of time thatguests occupy a table during given day parts.From these observations, the restaurateurcould determine an average meal length,while also noting the extent of variations inmeal length. That is, the restaurant operatorneeds to know the average length of a meal,plus how close to the average most dinerscome. Wide variation of meal lengths(known as a high standard deviation) makesforecasting more difficult and perhaps callsfor management efforts to make the durationmore consistent.

Uncertainty of ArrivalArrival uncertainty can be reduced throughreservation and arrival-management policiesand by developing and implementing anoptimal table mix. The key to reducingarrival uncertainty is to understand thetiming and volume of customer arrivals. Thisinformation can be used to develop forecasts

Exhibit 2Methods of managing duration

Uncertainty of arrival

Internal measures

Reservations policiesArrival management

Optimal table mix

External measures

DepositsGuaranteed reservationsReconfirmed reservations

Uncertainty of duration

Internal measures

Changes in the service delivery systemLabor scheduling

Menu designCommunication systems

External measures

Pre-bussingVisual signals

Coffee and dessert barCheck delivery



that can assist with the establishment of goodreservations and walk-in policies and withdeveloping an optimal table mix.

Reservation PoliciesAs I said earlier, restaurants that takereservations must contend with possible no-shows, short-shows, or late-shows, whileoperators who do not take reservations mustpredict how many guests will arrive andwhen they will do so. Given a choice, a no-reservation policy is probably preferable, butin many situations, particularly in fine-dining

restaurants, customers expect to be able tomake a reservation, so accepting reservationsis often a market necessity.

Most reservations are made directlywith the restaurant, although on-line reserva-tions are increasingly being made (e.g.,opentable.com, iseatz.com). If a restauranttakes reservations, its managers must decideon the number of tables to allocate to eachtime slot and determine the desired intervalbetween reservations. The number and sizeof tables to allocate to each time slot de-pends on the mix of walk-in and reservationbusiness and on staffing levels. Little re-search exists on the optimal number oftables to allocate to each time slot, but thefocus should be on spreading demandthroughout the meal period, and if possible,shifting demand to off-peak periods (seebelow for a more detailed discussion ofdemand shifting). The desired intervalbetween reservations will be dictated accord-ing to the expected meal duration by partysize.

When customers call the restaurant fora reservation, they risk the possibility ofcalling during hours the restaurant is closed

or of reaching someone who is not knowl-edgeable in how to take a reservation.Restaurants handle this problem in two ways:(1) (1) (1) (1) (1) dedicated reservation agents and (2)(2)(2)(2)(2) on-line reservations.

Dedicated reservation agents not onlyreduce the load on other staff members whomight respond to reservation calls, but alsoprovide increased accessibility and consis-tency. Obviously, retaining dedicatedreservation agents may be cost prohibitivefor small operations, but may be quiteworthwhile for high-volume restaurants.

On-line reservations provide completeaccessibility, consistent service, and anenhanced reach at a minor cost—approxi-mately $1 per person seated. While this costmay seem high to some operators, the cost isrelatively low, given that the labor costsassociated with reservation agents can bereduced and the restaurant can have in-creased exposure to a potentially largermarket.

Reservations are not without problems.The fact that a customer makes a reservationdoes not ensure that the customer will honorthat reservation. Even if the customer showsup, there is no assurance that the customerwill arrive on time or with the promisednumber of customers. Because reservationsare unpredictable, they must be managed—either internally (with techniques not involv-ing customers) or externally (involvingcustomers). The primary internal approachis overbooking, while external methodsinclude requiring deposits, guarantees, andthe reconfirmation of reservations.

No-showsNo-showsNo-showsNo-showsNo-shows. The primary internalapproach to handling no-shows isoverbooking, a technique used by mostcapacity-constrained service industries.Restaurants have typically not usedoverbooking to offset no-shows, but haveinstead relied on walk-in business as abuffer—although this strategy works only ifenough walk-ins arrive at the right time.

Given the choice, a no-reservation policymight be best, but that’s not always possible.



The key to a successful overbookingpolicy is to obtain accurate information onno-shows, cancellations, and walk-in guestsso that managers can set levels ofoverbooking that maintain an acceptablelevel of customer service. A manager can usesimple mathematical models to developappropriate overbooking policies by time ofyear, day of week, and time of day. A goodoverbooking policy balances the cost ofunused tables with the cost of inconvenienc-ing or displacing a party—bearing in mindthat a guest denied a reserved table may notbe especially forgiving. Along that line,restaurants that use overbooking mustdevelop good internal methods for selectingand handling displaced guests. Operators inother industries base their displacementdecision on time of arrival (if customers arelate, their reservation is no longer honored),frequency of use (regular customers arenever displaced), or perceived importance(important, high-spending customers arenever displaced).

As I indicated above, requiring credit-card guarantees and deposits are externalapproaches to reservation management, as iscalling customers to reconfirm reservations.Restaurants have long required a deposit forspecial meals (e.g., Mother’s Day, NewYear’s Eve), although the practice may meetwith customer resistance during low-demandperiods. Similarly, many fine-dining restau-rants in large cities have started to require acredit-card guarantee for reservations onbusy nights. Hotels and airlines have usedguaranteed reservations for many years andhave been able thereby to reduce thenumber of no-shows. One problem withcredit-card guarantees in the restaurantindustry, though, is that unlike hoteliers whocan require one night’s room rate to secure areservation, restaurateurs lack knowledge onexactly how much the reserving party willspend on dinner and so cannot charge thespecific price of the lost meal for no-shows.

Restaurants using credit-card guarantees haveaddressed this problem by charging guestswho fail to honor their reservations a statedfee (typically, $15 to $25 per guest).

Rather than require deposits or credit-card guarantees, some restaurants use a lessobtrusive, more service-oriented method ofreducing no-shows. These operators calltheir customers during the day to reconfirmtheir reservations for the evening to come.The call reminds the customer of thereservation and gives the customer thechance to cancel on the spot, if need be. Thecalls also create a reasonably solid forecast ofthe number of parties who intend to honortheir reservation. For this approach to besuccessful, the incremental personnel costassociated with calling customers should beoffset by the increased revenue associatedwith a reduction in no-shows.

Late-showsLate-showsLate-showsLate-showsLate-shows. Restaurants can managelate-shows by establishing and communicat-ing maximum hold times for tables. Whenthe reservation is made, customers can beinformed that their table will be held for aspecified period after the time of theirreservation—at which time the table will bemade available to any waiting party. Such apolicy, if clearly communicated, can helpreduce late-shows and help protect therestaurant from the resulting idle capacity. Arestaurant with a RevPASH of $30 (or $0.50per minute) would, for instance, lose $60during busy periods (i.e., when customersare waiting) for a 4-top that is held 30minutes for a late-arriving party.

Short-showsShort-showsShort-showsShort-showsShort-shows. Short-shows are moredifficult to handle, especially, say, when thecustomer ordered only appetizers at dinnertime and then abruptly left. In that situation,after all, the customer honored the reserva-tion, but merely left before making a “com-plete” purchase. Theoretically, customerscould be charged a per-person fee for short-shows, but in practice, this policy wouldprobably not be well accepted. Hotels face a



similar problem when customers stay for ashorter time than they have reserved. Earlydeparture fees have met with customerresistance. Some hotels handle this problemby forecasting the percentage of guests wholeave early and overbooking accordingly.

Alternative Reservation PoliciesSome restaurants do not expressly acceptreservations, but do use call-ahead seatingthat allows customers to put their names inthe queue, sometimes with an estimatedarrival time. Such a policy can be effectivefor customers who know how to use it, but

can be confusing and potentially frustratingto those who do not know about the tele-phone policy. For example, if a party hasbeen waiting for a while and sees a later-arriving party being seated, those who arewaiting may be unhappy and frustrated. Evenif the call-ahead seating policy is explained,that may not always placate the dissatisfiedguest. In addition, call-ahead seating candistract the host or hostess from the neces-sary functions of greeting and seating guestsalready at the restaurant.

Some restaurants take reservationsonly for large parties, which allows therestaurant to prepare the table ahead of timeand reduce the wait for that party. Thenagain, if the restaurant does not require aguarantee (either with a credit card ordeposit), it can end up with empty seatswhen it could have been serving waitingcustomers. One operator that I know ofaccepted a reservation from a party of 20 fora Friday night at 7:00 PM but did not requirea guarantee. The manager set up the table

ahead of time and scheduled an additionalserver, but the party never showed up.Meanwhile, the restaurant had turned away anumber of potential guests.

Another approach to reservations is toaccept reservations only during off-peakperiods. Customers placing a high priorityon reservations may choose to book at anoff-peak time and others may be willing toforgo a reservation and take their chances onsecuring a table upon arrival at their desiredtime. Disney restaurants use a “priority”seating approach, in which guests can reservea table only during off-peak times and canthen, upon arrival, be seated at the nextavailable “right-size” table.

Without ReservationRestaurants that do not accept reservationsand rely on walk-ins to fill their seats must beable to forecast the quantity (how manyparties of various sizes) and timing (the timeof day) of walk-ins and must have well-developed arrival-management policies. Inaddition, a table mix that reflects the compo-sition of the party mix (for example, a fairnumber of 2-tops in restaurants that havemany small parties) is essential.

Improved forecastingImproved forecastingImproved forecastingImproved forecastingImproved forecasting. The POSsystem is the best source of information onthe quantity and timing of walk-ins, eventhough it carries the built-in liability ofshowing only when a check is opened andnot when guests arrived or were seated.Nevertheless, POS information can be usedto develop reasonable forecasts of arrivalpatterns by time of day and party size.Sophisticated forecasting methods are notnecessary; even a simple average of thenumber of parties of two that arrived onFridays between 6:00 and 7:00 over the pastthree or four weeks is sufficient. Thisforecast can then be used to improve therestaurant’s seating and greeting functions.

Improved arrival managementImproved arrival managementImproved arrival managementImproved arrival managementImproved arrival management. Thehost or hostess is essentially the restaurant’s

Restaurants that do not accept reservationsmust be able to forecast and manage the

quantity and timing of their demand.



revenue manager. Frequently, though,restaurants place inexperienced employeesin this position. Such an approach may keeplabor costs down, but it can impair revenuesince that particular employee may not bematching parties to tables with revenue yieldin mind. For example, if a host or hostessregularly seats parties of one or two at 4-tops,the revenue effect during busy periods willbe substantial.

Along that line, policies for determin-ing the order in which customers should beseated must also be developed. Mostrestaurants use first-come, first-servedseating, in which the parties that arrived firstare seated before later arrivals, but this canbe a thorny matter. The policy couldbackfire, for instance, if the first party on thewait list is a party of two and the open table isa 6-top. As in the case of Disney, somerestaurants deal with this problem byadapting the first-come, first-served rule toone of first-come, first-served at the next“right-size” table. Although such a policyseems logical, it may diminish customersatisfaction.

Some restaurants assign a variety ofside-work duties to hosts and hostesses,including handling take-out orders andanswering the phone. Such an approach canwork during slow times, since not manycustomers will be arriving at the restaurant,but it can be dangerous during busy periods.During busy periods, the host or hostessshould remain at the reception stand toensure that all guests are greeted promptlyand seated or, if necessary, placed on thewait list. In addition, during busy periods, itmay be wise to have multiple hosts orhostesses or use seaters to take guests totheir tables.

It’s important for the host or hostess tobe able to give accurate wait-time estimatesto arriving guests. Some restaurants prefer tounderestimate wait time (in the hope thatpotential guests won’t be scared off by a long

wait), while others prefer to overestimatewait time (so that customers are pleased thatthey don’t have to wait as long), but areasonably accurate wait time is essential tocustomer satisfaction. If the host or hostess isinexperienced, it is even more important toprovide guidance on how to properly quotewait times.

Optimal Table MixAn optimal table mix is one which providesa set of tables that closely matches the mix ofparty sizes. 5 For example, if 50 percent of allparties are parties of one or two and therestaurant has only 4-tops, the revenuepotential of the restaurant will be dimin-ished. One of the major advantages of atable mix that matches the customer mix isthat it takes much of the guesswork out ofwhich party to seat at which table. In addi-tion, an optimal table mix will minimizecustomers’ waiting time, but might requirestaffing changes. If the optimal table mix hasmore tables than the original mix (as isusually the case), the restaurant may need toschedule more servers than previously. Also,the increased seat occupancy associated withan optimal table mix may increase the loadon the kitchen.

A simple and effective way of deter-mining a restaurant’s optimal table mix isfirst to determine its party mix. Those datacan be obtained through either the POSsystem or through observation. Once this isdone, the operator needs to determine theappropriate table size for each party size.

5 G. M. Thompson, “Optimizing Restaurant-table Configura-

tions: Specifying Combinable Tables,” Cornell Hotel andRestaurant Administration Quarterly, Vol. 44, No. 1 (February

2003), pp 53–60; G. M. Thompson, “Optimizing a Restaurant’s

Seating Capacity: Use Dedicated or Combinable Tables?,” CornellHotel and Restaurant Administration Quarterly, Vol. 43, No. 3.

(June 2002), pp. 48–57; S.E. Kimes and G.M. Thompson,

“Restaurant Revenue Management at Chevys: Determining the

Best Table Mix,” Cornell University School of Hotel Administra-

tion working paper 04-23-03; and S.E. Kimes and G.M.

Thompson, “An Evaluation of Heuristic Methods for Determining

the Best Table Mix in Full-Service Restaurants,” Cornell University

School of Hotel Administration working paper 09-04-02.



This is merely logic: parties of one or twoshould be seated at 2-tops, parties of three orfour at 4-tops, and so forth. Because mostrestaurants have relatively few truly largeparties, it’s probably best to just lump anylarge parties into one category (e.g., 8+people).

Research has shown that an improvedtable mix can increase revenue potential byup to 35 percent without an increase incustomers’ waiting time. In addition, thecombinability and layout of the table mix canmatter. For example, many operators believethat having only 2-tops can lead to betterresults because the tables can be combinedinto any necessary configuration. A simula-tion study by Gary Thompson found,however, that while combinable tables workwell for small restaurants, large restaurantsdo better with a variety of dedicated, non-combinable tables.6

In addition, research has shown thatchanging the table mix each night (asneeded) in a busy restaurant can increaserevenue by 1.2 percent. This has consider-able promise for restaurants that are busy forall meal periods or are busy each night of theweek. If the party-size mix varies by mealperiod or by night, it might be worthwhile todevelop and change the table mix on aregular basis. Some restaurants might alsowant to consider changing the table mix forcertain busy days such as Valentine’s Day orMother’s Day.

Uncertainty of DurationA restaurant operator who has dealt with thearrival-time issue must still be able to predictmeal length, because this controls thenumber of tables available. With thisinformation, operators of reservations-basedrestaurants can decide which reservation

requests to accept, and restaurants with alarge walk-in trade will be better able toprovide accurate estimates of waiting timefor guests in the queue. In addition, areduction in meal duration during busyperiods can increase seat occupancy andtable turnover and can lead to increasedrevenue.

As stated at the outset, one of thedifficulties of implementing revenue manage-ment in restaurants is the fact that theirexplicit unit of sale is a meal (or an event)rather than an amount of time, although onecould argue that the true measure of therestaurant’s product is time. While the likelylength of a meal can be estimated, its actualduration is not firmly set. Reduced diningtimes can have considerable revenue poten-tial during high-demand periods.7 Considera restaurant with 100 seats, a $20 averagecheck, a one-hour average dining time, and abusy period of four hours per day. Duringbusy periods, defined as those when custom-ers are waiting for a table, a decrease indining time can increase the number ofcustomers served and the associated rev-enue. Under the assumptions I just gave, therestaurant could theoretically serve 400customers during its four-hour busy time(240 minutes/60 minutes * 100 seats),assuming all 100 seats were occupied fourtimes for exactly one hour each time. Thatwould result in revenue of $8,000 (400customers * $20 average check). If theaverage dining time could be reduced to 50minutes, the potential number of customersserved would increase to 480 (240 minutes/50 minutes * 100 seats), and the potentialrevenue would increase to $9,600, anincrease of 20 percent. The question of howcustomers would react to such changes,however, causes restaurant operators toapproach time decreases with caution.

If customers think that dinner takes toolong (because the service is lax), they may

6 See: Gary M. Thompson, “Dedicated or Combinable? A

Simulation to Determine Optimal Restaurant Table Configura-

tion,” CHR Reports, Cornell University Center for Hospitality

Research, 2003 (www.chr.cornell.edu). 7 Kimes, op. cit.



choose not to return to a restaurant. Simi-larly, if customers believe that dinner is tooshort, they may feel shortchanged or rushedand also may not return. The expecteddining duration is affected by a number ofvariables, including the type of restaurant,the reason for dining (e.g., special occasion,entertainment, routine), and the characteris-tics of the diners (e.g., nationality, age,income, frequency of dining out, andamount of free time). For example, considera couple who decide to dine out for theiranniversary. They select what they perceiveto be the nicest restaurant in town andexpect to make an evening of the meal. Ifthey go to the restaurant and they are hustledthrough dinner in only an hour, they mayfeel shortchanged.

There’s no question that diners have aspecific idea of how long their meal shouldlast. In that regard, my associates and Iconducted a survey to find out how longcustomers think a dinner with friends at acasual restaurant should take.8 On average,customers expected the meal to last forabout one hour. Interestingly, the expecta-tions varied by nationality: Asian respon-dents gave the shortest expected dining time,at 57 minutes; North Americans’ expecta-tions averaged 59 minutes; and Europeansthought the meal should last about 77minutes.

The same research team also askedquestions about the length of time consid-ered to be short, too short, long, and toolong. The expected dining time of 60minutes was much higher than the timeconsidered to be short (30 minutes) or tooshort (23 minutes). The average timeconsidered to be long was 82 minutes, whilethat considered to be too long was 93minutes. This indicates that casual restau-

rants have considerable latitude in adjustingmeal duration without upsetting customers.

As with arrival time, restaurant opera-tors can exert control over meal duration.Internal approaches in this case revolvearound setting up systems and training tomake the meal length shorter and moreconsistent, while external approaches involve

encouraging guests to give up their table evenif they are not really ready to leave andchoose to linger elsewhere in the restaurant.

By reducing time variability, managerswill be better able to give accurate estimatesof waiting time for those in the queue anddetermine whether and for what timereservations should be accepted. A restaura-teur can manage duration by concentratingon menu design, service-delivery design,labor scheduling, and communication tools.

Menu designMenu designMenu designMenu designMenu design. Some restaurants haveredesigned or established their menuaccording to the preparation and consump-tion time for each menu item. Menu itemsthat exceed the established target for prepa-ration or consumption are either recon-figured or eliminated from the menu. Like-wise, menu items that cause customers tolinger can be eliminated if those items do notcontribute to an increase in RevPASH.

Improved service processesImproved service processesImproved service processesImproved service processesImproved service processes. The keyto improving service processes is to carefullyobserve your current front-of-the-houseprocedures and target specific areas forimprovement. The meal experience can bebroken into three segments: pre-process, in-process, and post-process. The pre-processsegment includes the time from when thecustomer is seated until the first course is

Reducing meal duration offers great potentialfor increasing restaurant revenue, especially

during busy periods.

8 S.E. Kimes, J. Wirtz, B.M. Noone, “How Long Should

Dinner Take? Measuring Expected Meal Duration for Restaurant

Revenue Management,” Journal of Revenue and Pricing

Management, Vol. 1, No. 3 (2002), pp. 220–233.



delivered; the in-process segment starts whenthe first course is delivered and concludeswhen the check is requested; and the post-process segment includes the time from thewhen check is requested until the customerdeparts.

The largest opportunities for improve-ment generally come during the pre-processand post-process stages. Luckily, these arealso the parts of the meal that most guestsprefer to move along, and so the operatordoes not have to worry too much aboutguests’ feeling rushed in those segments. AsI’ve mentioned, care must be taken whentrying to speed up the in-process segment.

The pre-process segment consists ofseating the guest, greeting the guest, takingthe drink order, delivering drinks, taking thefood order, and delivering the first course.

Problem areas in this segment generallyinclude delays in greeting the guest, delays indelivering the drinks, and delays in takingthe food order. If an operator can figure outways to reduce delays of that kind, mealduration will drop and customer satisfactionwill most likely improve.

The concerns regarding the in-processsegment are primarily to keep focus andmaintain the pace. The operator mustensure that food is delivered in a timelyfashion, that the timing of the courses isappropriate, and that the guest does not feelrushed. Pre-bussing can and should occur,but guests should not feel as if they are beingpushed out the door.

The post-process segment consists ofthe check request, the check delivery, the

payment process, the check return, andguests’ departure. Generally, once guestsrequest the check, they are ready to leave therestaurant. Anything that the restaurant cando to speed the check-processing time willenhance guest satisfaction and reduce diningduration (again, without unduly rushing theguest).

StaffingStaffingStaffingStaffingStaffing. Redesigning the menu andprocedures, in conjunction with improvedforecasts of customer arrivals, shouldimprove labor scheduling, which is a keyelement in controlling meal duration.Restaurateurs’ common desire to minimizelabor costs may backfire if reduced staffingleads to slower table turnovers and longermeal times. The increased revenue resultingfrom faster table changeovers made possibleby extra bussers or servers may more thancompensate for the increased labor costs.Implementing a revenue-managementstrategy would help a restaurant operatordetermine appropriate staffing levels.

CommunicationsCommunicationsCommunicationsCommunicationsCommunications. Some restaurantshave improved communication systemsamong employees and have increasedcontrol by tracking the connection betweenfood preparation and food delivery. Bysetting up appropriate communicationmechanisms, kitchens can notify servers thata course is ready for pick up and servers cannotify bussers that a table is ready to becleared, thereby speeding the meal service(usually to the guests’ delight) and making itpossible to improve RevPASH. To assistwith employee communication, somerestaurants have used information technol-ogy such as headsets and table-managementsystems.

External ApproachesPart of duration management involvesfinding ways to signal to guests that it is timefor them to relinquish their table. Customerswho unexpectedly linger after their meal mayinterfere with seating the next party. A

The greatest chances for tightening the diningprocess are before and after the meal—which

most guests also would like to move along.



restaurant can use both implicit and explicitsignaling devices to remind guests that themeal is over. Many restaurants use subtleimplicit approaches such as bussing thetable, dropping off the check, or offeringvalet service. In a few restaurants, customersare asked to specify how long they plan tostay, but that is rare. Instead, the restaurantmanager must rely on timing the courses andother implicit signals to remind the customerthat the meal has ended.

Explicit approaches risk customer ire.A manager obviously cannot directly askcustomers to leave, but the restaurant couldattempt other, less offensive methods ofturning the table. Some restaurants in thetheater district of New York City, forinstance, place an hourglass on each party’stable. When the sand in the hourglass isgone, patrons have a visual cue to finishdinner and leave so that they will not be lateto the theater (and, not incidentally, releasethe table).

Reducing Changeover TimeReducing the amount of time betweencustomers (changeover time) increasescapacity and revenue. This tactic will notoffend a departing customer and shouldplease the customers who are waiting to beseated. As an example, reducing changeovertime has become a common strategy in theairline industry. Southwest Airlines andRyanAir both boast 20-minute aircraftturnarounds and have thereby been able toincrease plane use. Cabin employees atSouthwest, for one, actually enlist passen-gers’ assistance in clearing the cabin byasking them to hand over discarded newspa-pers and other trash that usually would beleft in the seatback pouch. A quick-turnstrategy suits the restaurant industry. Themanager of a fine-dining restaurant in LasVegas with a RevPASH of $60 knows, forinstance, that each seat in his restaurantgenerates $1/minute. If customers are

waiting and a 4-top has not been bussed andsits vacant for five minutes, it costs thatrestaurant $20. When viewed this way, thecost of an additional busser to promote aquick turn seems minor. (Bear in mind,however, that the $20 in question is revenue,and not necessarily contribution, a fact thatshould be considered in looking at the valueof the busser.)

Managers can do a number of things toensure that changeover time is minimized.First, they should insist that bussers andservers work as teams and do a good job ofpre-bussing the table. The fewer items left onthe table when the guests depart, the easierand faster it will be to clear and reset thetable. In addition, as I indicated, pre-bussing,if not overly aggressive, can send a reason-ably subtle signal to customers that it is timeto relinquish the table.

Clear communication among bussers,servers, and hosts helps bussers and serversknow when a table is ready to be cleared andreset, and those employees should, in turn,notify the host so that he or she can find thenext party to be seated at that table. Somerestaurants use technology such as table-management systems to facilitate thiscommunication process, but as long as aclear line of communication is established,technology is not always necessary.

Management should also analyze andstreamline the process of clearing andresetting tables to minimize changeover time.Again, it helps to think of an idle table aspotential revenue. Anything that can be doneto shrink this time, whether it is having pre-rolled flatware and napkins, easily accessibletablecloths, or well-placed glassware, canhelp the restaurant increase revenue.

When customers are waiting at busyrestaurants, one of the major stumblingblocks to reducing the changeover time islocating the next party to be seated. Somerestaurants use technology (such as loud-speakers or hand-held signaling devices) to



quickly notify guests, but this technology isnot appropriate for all restaurants, sincecustomers don’t always respond well to suchapproaches. If hosts know which tables willbe available and know where to find the nextparty, they can find the party ahead of timeand have the customers ready to be seated assoon as the table comes up. Otherwise, ifhosts wait for notification that a table isready, it may take some time to find theparty, some time for the party to settle ortransfer its account in the bar (if that’s wherethe customers are waiting), and some time towalk the party to the table. Again, it helps inthis context to think of an idle table aspotential revenue being squandered. It maybe worthwhile to hire an additional seater tohandle this process.

PricingThe key to any successful revenue-manage-ment strategy is to offer multiple prices to avariety of market segments, as appropriate.For example, movie theaters often offerlower prices to seniors and children atcertain times of the day or on certain days ofthe week. Similarly, the airline industryoffers a wide variety of prices on the sameroute depending on the time, method, anditinerary for the booking. That means, forexample, that economy passengers flyingfrom Los Angeles to Singapore may paynothing (by using frequent-flyer miles) toover $1,500 for the same seat.

To apply this multiple-price approachto the restaurant industry, managers mustanswer two questions: what prices should becharged, and who should pay which price?The answers to those two questions areaffected by the answer to a third question—how will customers react to variable prices?

Setting the PriceRevenue management is often associatedwith manipulating prices according todemand characteristics. On the surface,

many restaurants seem to do so by offeringvariable prices, usually based on the time ofday (e.g., the early bird special) or the day ofthe week (e.g., the Friday fish fry). Suchvariable pricing, however, cannot necessarilybe said to constitute revenue management inthe absence of a customer-focused strategy.The three chief methods for setting pricesare cost based, demand based, and competi-tively based. Revenue management typicallyinvolves demand-based pricing, but somecompanies practicing revenue managementuse other types of pricing strategies, as well.

Cost-based pricingCost-based pricingCost-based pricingCost-based pricingCost-based pricing. Restaurants havetraditionally used cost-based pricing, inwhich the cost of the menu item’s ingredi-ents is first calculated and then multiplied bysome constant (usually about 3) so that therestaurant can maintain a certain food-costpercentage (usually about 30 percent). Whileit is certainly important to track costs, a cost-based pricing approach can lead to sub-optimal results, particularly in situations inwhich customers are willing to pay morethan the cost-based price. In addition, it maybe worthwhile in some situations to offer alower price in an attempt to stimulatedemand (of course, assuming that all costswould still be covered).

Demand-based pricingDemand-based pricingDemand-based pricingDemand-based pricingDemand-based pricing. Demand-basedpricing is based on the notion of respondingto guests’ demand characteristics, in particu-lar their response (or lack thereof) tochanges in prices. As an example, a resort inMalaysia catered to three major marketsegments: Europeans, Japanese, and groupsfrom other parts of Asia. The segmentsvaried in price sensitivity: the Europeans andJapanese were not price sensitive, while theAsian groups were extremely price sensitive.This hotel operated three restaurants: abuffet restaurant, a sit-down Asian-stylerestaurant, and a sit-down steak and seafoodrestaurant. The average check per personwas about $12 in the buffet restaurant, $15 inthe Asian restaurant, and $30 in the steak



and seafood restaurant. The price-sensitiveAsian groups preferred to eat in the buffetrestaurant, and the non-price-sensitiveEuropeans preferred the sit-down Asianrestaurant, in which they could get “safe”local food. Upon careful reflection on theprice sensitivity of the resort’s differentmarkets, the manager decided to increasethe prices at the Asian restaurant by 30percent. The decision was well taken, asthere was no decrease in that restaurant’svolume, so revenue and profit increased by30 percent.

One version of demand-based pricingis charging a relatively high price during high-demand times and a lower price during low-demand times. As an example of thatapproach, a restaurant in Singapore experi-enced extremely high demand on weekendsand much lower demand during the week.The managers decided to develop a specialweekend menu that featured prices that were25-percent higher than during the week.Once again, the restaurant suffered nochange in volume, so weekend revenue andprofit increased by 25 percent.

Both examples illustrate uses ofdemand-based pricing. The hotel in Malay-sia used information on the price insensitiv-ity of its European customers to changeprices in the restaurant considered mostattractive by those guests, and theSingaporean restaurant realized that it wasunderpricing its high-demand periods.9

Menu engineering supports anotherform of demand-based pricing.10 With

menu engineering, the contribution margin(selling price less the cost) and the numberof units sold of each menu item are calcu-lated and plotted on a graph. The menuitems are then divided into the following fourcategories: (1)(1)(1)(1)(1) Stars: above-average contribu-tion margin and above-average volume;(2) (2) (2) (2) (2) Cash cows: below-average contributionmargin and above-average volume;(3)(3)(3)(3)(3) Question marks: above-average contribu-tion margin and below-average volume; and(4)(4)(4)(4)(4) Dogs: below-average contribution marginand below-average volume. Stars and cashcows are good candidates for potential priceincreases since they both have high demand,

while question marks may be possible itemsfor price decreases.

Restaurants that use menu engineeringgenerally review their results each monthand make necessary price adjustments. Ofcourse, this necessitates the printing of newmenus, but that cost is generally minimal andshould be more than covered by the associ-ated revenue increase.

Competitive pricingCompetitive pricingCompetitive pricingCompetitive pricingCompetitive pricing. Some restaurantsset their prices according to competitors’prices. For example, if they offer a grilledfish, they make sure that their price is similarto that of their competitors. If their grilledfish is slightly better or if the atmosphere oftheir restaurant is more upscale, they maycharge a slight premium over the competi-tion. On the other hand, if their grilled fish isnot quite as good or if the restaurant is lessupscale, they might offer a slightly lowerprice. Competitive pricing, which is preva-lent in the hotel and airline industries, can besuccessful, but companies that use competi-

The key to any successful revenue-managementstrategy is to offer multiple prices to a variety of

market segments, as appropriate.

9 A cautionary note: Increasing prices, as in the cases cited

here, must be handled carefully. At no time should a restaurant (or

any other hospitality operation) be seen as price gouging or taking

unfair advantage of customers by raising prices in times of stressful

or emergency situations.10 For a discussion of menu engineering, see: Lee M. Kreul,

“Magic Numbers: Psychological Aspects of Menu Pricing,” CornellHotel and Restaurant Administration Quarterly, Vol. 23, No. 2

(August 1982), pp. 70–75; David K. Hayes and Lynn Huffman,

“Menu Analysis: A Better Way,” Cornell Hotel and RestaurantAdministration Quarterly, Vol. 25, No. 4 (February 1985), pp. 64–

70; and David V. Pavesic, “Prime Numbers: Finding Your Menu’s

Strengths,” Cornell Hotel and Restaurant Administration Quarterly,

Vol. 26, No. 3 (November 1985), pp. 70–77.



tive pricing seem to assume that theircompetitors have set their prices correctly. Ifthis assumption is untrue, the use of com-petitive pricing could be dangerous.

Need for experimentationNeed for experimentationNeed for experimentationNeed for experimentationNeed for experimentation. Successfulprice increases are often associated withexperimentation. The cost of printing a newmenu is relatively low, and a new menu withnew prices can easily be tested. If customerreaction is negative, the new menu can bewithdrawn, and if customer reaction isneutral or positive, the menu can be contin-ued. Offering nightly specials allows arestaurant to experiment with different pricesand menu items in a non-threatening way.

Who Pays Which Price?In differential-pricing systems, the questionof which customers pay which price isusually addressed through the use of ratefences.11 Hotels and airlines use rate fencesto offer discounts on inventory that mightotherwise not be sold at all to customers whomight otherwise not purchase that inven-tory—while at the same time preventingcustomers who were going to buy anywayfrom taking advantage of a discount that theydid not actively seek. For example, thefamiliar airline or hotel rate fences includerequiring customers to make their reserva-tion in advance, prepay for their reservation,or stay over a Saturday night. The fences cancomprise almost any set of rules as long asthey somehow make sense to the customer.

While early bird specials and the likeconstitute rudimentary rate fences, managersmust think beyond happy hours and two-for-one specials, which do not really discrimi-nate the price-sensitive customers from theirfree-spending friends. Instead, restaurateursmust develop stratagems for offering differ-ential prices that make sense for the demand

level for a given time. As a rule, restaurantsoffer the same menu prices regardless of thecustomer’s demand characteristics. Perhapsthe question for restaurateurs is whether theycould implement some kind of pricingdifferential for busy times (e.g., Saturdaynights) and slack times. Early bird specialsare a step in this direction, as are specialprices for affinity groups and frequent-dinerclubs. The next step is to create an overalldemand-management program based in parton demand-based pricing.

Unfair? Unfair? Unfair? Unfair? Unfair? Restaurant operators are oftenreluctant to use demand-based pricingbecause of the potential customer backlashstemming from perceptions of unfairconduct. If increased prices cannot bejustified in some way (say, by menu itemsthat obviously have high production valuesor by offering other desirable conditions),customers may view demand-based-pricingpolicies as unfair. The issue of fairness hasbeen studied extensively in a variety ofindustries. In general, it has been found thatfair behavior on the part of operators isinstrumental to the maximization of theirlong-term profits.

Consumers may perceive demand-based pricing as being unfair for at least tworeasons. First, they may view charging highprices during high-demand times as exceed-ing their reference price, or their referenceprice may already have been shifted downbecause of low prices charged during low-demand periods. In either event, the “new,”higher regular prices may be perceived asless fair than before. Second, the restaurantmay not be seen as providing more value forthe higher price. I will first discuss the effectthat demand-based pricing can have onconsumer reference prices, followed by anexamination of the potential effect on theperceived fairness of demand-based pricing.

Reference pricesReference pricesReference pricesReference pricesReference prices. I have used the terms“reference transaction” and “referenceprice,” which are often used when discussing

11 R.B. Hanks, R.P. Noland, and R.G. Cross, “Discounting in

the Hotel Industry, A New Approach,” Cornell Hotel andRestaurant Administration Quarterly, Vol. 33, No. 3 (June 1992),

pp. 40–45; and R.J. Dolan and H. Simon, Power Pricing (New

York: The Free Press, 1996).



fairness.12 A reference transaction is howcustomers think a transaction should beconducted, and a reference price is howmuch customers think a service (or product)should cost. Reference prices can comefrom the price last paid (especially at yourrestaurant), the price most frequently paid,and what other customers say they paid forsimilar offerings, as well as from marketprices and posted prices. For example,customers may know that they generally payabout $25 for dinner at a particular restau-rant, and so their reference price for dinnerat that restaurant is $25.

To assess a transaction’s fairness,customers often rely on reference prices inrelation to what they are paying for thecurrent transaction. For consumers toperceive the price increases inherent indemand-based pricing as being fair, guests’reference prices would have to shift in linewith the restaurant’s variable-pricing sched-ule. This may be difficult for operators toachieve for two reasons. First, as I justindicated, the low price used during low-demand periods may become the customer’sreference price and may make futurepurchases at the regular or peak rate seemunfair. Second, consumers may believe thatthe restaurant is charging them higher pricesto reap higher profits without having in-creased the customer’s value.

Rate fences.Rate fences.Rate fences.Rate fences.Rate fences. Rate fences are designedto allow customers to segment themselvesbased on their willingness to pay, theirbehavior, and their needs.13 One chiefpurpose of a rate fence is to create customersegments and justify why different peoplepay different prices. To be perceived as fair,fences need to be logical, transparent, up-front, and fixed, so that they cannot becircumvented.

The two overarching categories of ratefences are physical and non-physical.Physical rate fences for a hotel might includeroom location, furnishings, and the presenceof amenities or a view. Non-physical ratefences include time of consumption, transac-tion characteristics, buyer characteristics, andcontrolled availability. In a restaurantcontext, physical rate fences include tablelocation (e.g., a better table commands ahigher price), view (e.g., tables with a scenicview cost more), and amenities (e.g., tables ina private room with fresh-cut flowers costmore). Non-physical rate fences mightinclude time (e.g., a weekend dinner mightcost more, or meals consumed before 6:00PM might cost less), transaction characteristics(e.g., customers who make a reservation overa month ahead of time might pay less), buyer

characteristics (e.g., frequent customersmight pay less or get free-of-charge extras),and controlled availability (e.g., customerswith coupons will pay less).

Research on rate fencesResearch on rate fencesResearch on rate fencesResearch on rate fencesResearch on rate fences. My colleaguesand I studied the perceived fairness of fivepotential rate fences for restaurants—specifi-cally, several time-based fences (i.e., differen-tial pricing for lunch and dinner, weekdaysand weekends, and specific times of the day),one physical fence (i.e., table location), andone controlled-availability fence (i.e., twomeals for the price of one).14 We also

12 D. Kahneman, J.L. Knetsch, and R.H. Thaler, “Fairness

and the Assumption of Economics,” Journal of Business, Vol. 59

(October 1986), pp. S285–S300.13 Hanks et al., op cit.; Kahneman et al., op cit.

14 S.E. Kimes and J. Wirtz, “Perceived Fairness of Demand-

Based Pricing for Restaurants,” Cornell Hotel and RestaurantAdministration Quarterly, Vol. 43, No. 1 (2002), pp. 31–38; S.E.

Kimes and J. Wirtz. “Has Revenue Management Become

Acceptable? Findings from an International Study on the

Perceiving Fairness of Rate Fences,” Journal of Service Research.

Vol. 7, No. 2 (2003), pp. 125–135.

The question of which customers paywhich price is usually addressed through

the use of rate fences.



investigated how customers react to the wayprice differences are presented. Specifically,we wanted to know whether a fence framedas a price decrease would be evaluated morefavorably than the same fence presented as aprice increase. Research has shown thatpresentations that emphasize customer gainsare preferable to economically equivalentframes that emphasize customer losses.

In that regard, consider a restaurantwith a static menu that decides to establishtwo sets of menu prices. The restaurant canpresent the price differences in two ways: itcan either present the lunch prices as being20-percent lower than the dinner prices(framed as a gain from the diner’s perspec-tive), or it can present the dinner prices asbeing 20-percent higher than the lunchprices (framed as a loss). The situations areeconomically equivalent, but research hasshown that customers will view the “re-duced” prices more favorably, and that therestaurant should frame the price differenceaccordingly.

ResultsResultsResultsResultsResults. We found that restaurantpatrons consider demand-based pricing inthe form of coupons (i.e., two for the priceof one), time-of-day pricing, and lunch-versus-dinner pricing to be fair. Variableweekday-versus-weekend pricing was per-ceived as neutral to slightly unfair. Table-location pricing was seen as somewhatunfair, with potential negative consumerreaction to that practice. With regard toframing demand-based pricing as discountsor surcharges, we found that demand-basedpricing presented as discounts made thedifferential prices seem fairer in the consum-ers’ eyes, and therefore less likely to have anegative effect on consumers’ perceptionsand reactions.

Our findings provide restaurantoperators with some useful guidelines, butthose findings do not guarantee that allguests will willingly accept demand-based-pricing practices. Therefore, when develop-

ing demand-based pricing using fences,restaurant operators must make sure that therate fences are easy to explain and adminis-ter, and that customers can understand thereasoning behind them. This will make iteasier for front-line employees to pacifyunhappy or confused customers. Moreover,employees must understand that demand-based pricing is a win–win situation. It needsto be emphasized that variable pricing allowspatrons to choose prices that suit theirneeds, and, by having tight fences, a restau-rant can ensure that patrons who see highvalue in a good view, a desirable table, oreating dinner during peak times are muchmore likely to get the dining experience theyseek. Accordingly, increased profitability viademand-based pricing does not have tocome at the expense of customer satisfactionand loyalty.

Developing aRevenue-management Program

When developing a revenue-managementprogram, a restaurant operator must firstunderstand current conditions and perfor-mance.15 Following this, the operator mustevaluate the possible drivers of that perfor-mance. This understanding will help manag-ers determine how to improve RevPASHstatistics. Finally, the manager must monitorthe effects of changes on revenue perfor-mance. I describe each of these steps below.

(1)(1)(1)(1)(1)Establish the baseline. Most manag-ers know their average check and their labor-and food-cost percentages, but few canaccurately gauge their restaurants’ seatoccupancy or RevPASH. To develop arevenue-management program, operatorsmust collect detailed information on arrival,seat occupancy, and RevPASH patterns;party-size mix; meal times; and customerpreferences. This information can becollected from a variety of sources, including

15 Much of this section comes from: Kimes, op. cit.



the POS system, guest checks, and methodi-cal observation. Once collected, the datamust be analyzed to determine the mean andvariation of dining time and daily and hourlyseat occupancy and RevPASH patterns.

(2)(2)(2)(2)(2)Understand the drivers. Once thebaseline data have been collected, managersshould analyze the factors that affect mealduration and RevPASH performance.Simple tools such as process analysis, serviceblueprints, and fishbone diagrams can beused to better delineate the possible reasonsfor why meals last as long as they do and tohelp identify the most important problems incontrolling meal duration.16

(3)(3)(3)(3)(3)Develop a strategy. After identifyingthe causes of the most important problemsaffecting the service cycle, managers shoulddevelop detailed recommendations on howto correct those problems. Some solutionsmay deal with reducing the overall mealduration, while others may deal with reduc-ing variability in particular service steps (e.g.,order-taking, bussing), and still othersinvolve table mix or customer-arrival man-agement. The manager should analyze thepotential return on investment for eachrecommendation to ensure prudent decisionmaking.

(4)(4)(4)(4)(4)Implement the changes. Forrevenue management to be successful, res-taurant operators must ensure that managers,servers, bussers, and other employees clearlycomprehend the purpose and practice ofrevenue management. This requires aposition-specific training program that helpsemployees understand their role in revenuemanagement and how revenue managementwill benefit both the restaurant and theemployees themselves. Additionally, opera-tors should align any employee-incentiveprograms to coincide with the objectives ofrevenue management.

(5)(5)(5)(5)(5)Monitor outcomes. As with anybusiness practice, the success of revenuemanagement cannot be assessed withoutmeasuring changes. After establishing thebaseline and implementing revenue manage-ment, operators must develop a system tomeasure financial, operational, and cus-tomer-satisfaction performance.

An Application:Chevys Freshmex Restaurants

Chevys Freshmex Restaurants, a chain ofover 200 mid-scale Mexican restaurantslocated primarily in the southwestern UnitedStates, was interested in increasing restaurantprofitability by implementing a revenue-management strategy.17 Chevys prides itselfon its fresh ingredients, its lively atmosphere,and its friendly staff. The chain has donewell, but management noticed that waitswere long, particularly on weekend nights,and guests sometimes complained aboutthe length of time it took to dine at therestaurant.

The Chevys Arrowhead restaurant islocated in the Arrowhead Shopping Mall ina suburb of Phoenix, Arizona. Given that therestaurant is surrounded by a host of com-peting restaurants, many potential customerswould put their names on the wait list forseveral of these restaurants and patronize theone that gave them a table first. Managersreasoned that if Chevys could reduce its waittime, it should be able to attract more ofthese customers and increase revenue.

The restaurant had been open for overnine years, and the general manager hadbeen in place since the restaurant opened.She was considered to be one of the topmanagers in the Chevys system and had builta strong and loyal team.

This particular restaurant had 230 seatsin the main dining room (MDR) and 50

16 For a discussion of process-analysis tools, see: D. Daryl

Wyckoff, ”New Tools for Achieving Service Quality,” Cornell

Hotel and Restaurant Administration Quarterly, Vol. 25, No. 3

(November 1984), pp. 78–91.

17 Much of this section comes from: S.E. Kimes, “Restaurant

Revenue Management: Implementation at Chevys Arrowhead,”

Cornell Hotel and Restaurant Administration Quarterly, Vol.45,

No. 1 (February 2004).



more seats in the bar. In addition, patioseating was available from March throughNovember. Most tables (53) in the MDRwere 4-tops, and the remaining three tableswere 6-tops. Open from 11:00 AM to 11:00PM on weekdays and 11:00 AM to midnighton weekends, the restaurant draws a variety

of customers, including shoppers, couples,and families. As is typical in many restau-rants, this outlet is busy for weekend dinnersand lunches, when customers often wait for atable for over an hour. On the other hand,the restaurant is particularly slow on week-days before noon and between 2:00 and 5:00PM, and after 8:00 PM on Sundays throughThursdays.

The Revenue-managementApproach

The five-step process outlined above wasused to develop a revenue managementstrategy for Chevys Arrowhead.

Step #1: Establishing the Baseline

Two four-week periods of POS data anddetailed time studies were used to developthe baseline for the Arrowhead restaurant.The average check, the revenue per availableseat-hour (RevPASH), the seat occupancy,the meal duration (both from the POS dataand from the time studies), and the party-sizemix were analyzed, as follows.

Average checkAverage checkAverage checkAverage checkAverage check. The average check perperson for the 230-seat MDR was calculatedby day of week and hour of day. The averagecheck was $12.70, with hourly ranges from$9.02 at 11:00 AM on Fridays to $14.47 at

8:00 PM on Thursdays. The highest averagechecks occurred on Thursday, Friday, andSaturday evenings, while the lowest occurredfor weekday lunches.

RevPASHRevPASHRevPASHRevPASHRevPASH. The POS data were used toderive hourly RevPASH figures. RevPASHwas calculated by first determining the totalhourly revenue for each day of the week andthen dividing the hourly revenue by the 230seats in the MDR.

RevPASH ranged from $0.43 onTuesdays at 2:00 PM to $7.03 on Fridays at6:00 PM. The highest RevPASH occurred onFridays from 5:00–9:00 PM, on Saturdaysfrom noon–2:00 PM and 6:00–8:00 PM, andon Sundays from 1:00–2:00 PM. The lowestRevPASH was experienced after 9:00 PM

every day, and before noon and between2:00 and 5:00 PM on all weekdays.

Duration. Duration. Duration. Duration. Duration. The check-opening and-closing times from the POS data were usedto calculate the mean and variation of diningduration. The duration figures may beslightly inaccurate because the opening ofthe check did not necessarily correspond towhen the customers were seated at the tableand the closing of the check may not reflectwhen the guests actually left the table.

The average meal duration for dinner(after 4:00 PM) was 50 minutes with a stan-dard deviation of 20 minutes, while theaverage meal duration for lunch (before4:00) was 44 minutes with a standarddeviation of 16 minutes. Other than thedivision between lunch and dinner, theaverages and standard deviations did not varymuch by day of week or by hour of the day.

Seat occupancy.Seat occupancy.Seat occupancy.Seat occupancy.Seat occupancy. RevPASH is definedas seat occupancy multiplied by averagecheck and divided by average meal duration,so seat occupancy was calculated by dividingthe average check by the RevPASH andmultiplying by the average meal duration forthat time period. Seat occupancy figures forthe main dining room were calculated by dayof week and hour.

A revenue-management-based analysiswould highlight the portions of the service

process that could improve revenue peravailable seat-hour.



Seat occupancies are constrained bythe table mix, as I explain below, with thehighest seat occupancy achieved being only55 percent (between 6:00 and 7:00 PM onFridays). More generally, seat occupancyover 50 percent was achieved on Fridays andSaturdays from 6:00 to 8:00 PM, on Satur-days from noon to 2:00, and on Sundaysfrom 1:00 to 2:00 PM. Extremely low seatoccupancy (that is, under 15 percent)occurred on weekdays before noon, between2:00 and 5:00 PM, and after 9:00 PM on mostdays.

Time-study dataTime-study dataTime-study dataTime-study dataTime-study data: Since the POS dataincluded only information on total mealduration and did not have information onthe timing of the meal, detailed time studieswere conducted for weekend dinners. Astudent observer timed 100 parties overseveral weeks. The following ten categorieswere timed: seated, greeted, drinks deliv-ered, order taken, entrée delivered, checkrequested, check delivered, departure, tablebussed, and table reseated. The averagemealtime of 53:15 (with a standard deviationof 22:46) was a bit longer than the estimateobtained from the POS data since the time-study data clocked the time when the partywas actually seated, while the POS datacould only show when the check was openedand closed in the POS system.

The mean, standard deviation, andcoefficient of variation (defined as thestandard deviation divided by the mean) ofeach category were calculated. Categorieswith a high average time represent areas inwhich time savings and revenue gains can beachieved. Categories with a high coefficientof variation represent areas in which varia-tion can be reduced to increase revenue.

Examining the dining experience.Examining the dining experience.Examining the dining experience.Examining the dining experience.Examining the dining experience. Thedining experience represents the time fromwhen the customers are seated until the timethey depart.

(1)(1)(1)(1)(1)Seating to greeting (mean = 1:48,standard deviation =1:44). Greeting was

quick, but had high variation. Most partieswere greeted within two minutes.

(2)(2)(2)(2)(2)Greeting to drinks delivered (mean= 2:40, standard deviation =1:57). Drinkswere delivered fairly quickly but someinconsistency was noted.

(3)(3)(3)(3)(3)Drinks delivered to order taken(mean = 3:28, standard deviation = 3:20).Again, orders were taken fairly promptly, butthe variation was high.

(4)(4)(4)(4)(4)Order taken to entrée delivered(mean = 11:28, standard deviation = 6:12).Entrées took about eleven-and-a-half min-utes to deliver, and the standard deviationwas fairly high. This could be because of thevariety of entrées offered or because ofservers’ not knowing when orders were readyto be delivered.

(5)(5)(5)(5)(5)Entrée delivered to check dropped(mean = 20:12, standard deviation = 7:19).On average, customers took slightly over 20minutes between when their order wasdelivered and when the check was deliveredto the table (generally after the guest re-quested the check). There was a reasonablyhigh variation in this category.

(6)(6)(6)(6)(6)Check dropped to change returned(mean = 5:40, standard deviation = 3:48). Ittook over five minutes from when customersrequested the check until it was delivered.This represents a process ripe for tightening.The variation was fairly high, which indi-cates that training might help improve thissituation.

(7)(7)(7)(7)(7)Change back to departure (mean =4:20, standard deviation = 5:07). Customerstook a bit over four minutes to leave aftertheir change had been delivered. Thevariation was fairly high, which again indi-cates a potential area for improvement.

(8)(8)(8)(8)(8)Bussing and reseating. Bussing took2:44 (standard deviation = 2:11), and oncethe table was bussed, reseating only took0:52 (standard deviation = 0:31).

Party compositionParty compositionParty compositionParty compositionParty composition. Data on party sizewere obtained from the POS system. The



majority of parties (approximately 65 to 70percent) comprised one or two people.About 20 to 25 percent of parties had threeor four guests, 5 to 6 percent of parties werefive to eight guests, and only about 1 percentof parties had nine or more guests. (Partysize was not entered for 3 to 5 percent of allparties.)

The findings on party size are an issuefor the Arrowhead restaurant, because nearlyall of its tables are 4-tops. If the table mixcould be reconfigured while still maintainingthe same number of seats, revenue potentialduring peak demand periods would almostcertainly increase.

Step #2: Understanding the Causes

The baseline analysis led to several ques-tions: (1)(1)(1)(1)(1) Why was seat occupancy so low

even though customers were waiting fortables?, (2)(2)(2)(2)(2) Why was the meal itself taking solong?, (3) (3) (3) (3) (3) Why was there so much variationin meal duration?, and (4)(4)(4)(4)(4) Why were certainparts of the meal (mostly at the beginningand end of the meal) taking so long?

The analysis highlighted two areas thatwere interfering with the flow of service:namely, the table mix and the service-delivery process.

Table mix.Table mix.Table mix.Table mix.Table mix. Even if dining time couldbe reduced, customer flow would notincrease as much as desired because of thetable mix. The party-size calculations showedthat 52 to 56 percent of the customers werein parties of one or two, but, as I mentionedabove, all of the tables were 4-tops or 6-tops.It was no wonder that seat occupancy

struggled to exceed 50 percent at the sametime that customers were waiting for tables.What became clear is that all of the tableswere occupied, but often had empty seatsbecause of the large number of small parties.

Service delivery. Service delivery. Service delivery. Service delivery. Service delivery. Three major prob-lems were identified with the service-deliveryprocess. The high variability in dining timeseemed troublesome, as did the delays incompleting payment and bussing. Each wasanalyzed to help determine the possiblecauses for the delays.

VariabilityVariabilityVariabilityVariabilityVariability. The standard deviation ofdining duration was relatively high (nearly 23minutes), and the same was true of thestandard deviations of many of the mealsegments. This high standard deviation hadseveral consequences, but the chief out-comes were that (1) (1) (1) (1) (1) the restaurant was moredifficult to manage, and (2) (2) (2) (2) (2) customers mightview the service as inconsistent.

The sources for such high variationwere the customers themselves (e.g., theywanted to linger after their meal, or theytook a long time to order), the employees(e.g., insufficient staffing, poor training, poorcommunications), the menu and the kitchen(e.g., too many menu items, inefficientkitchen design), service processes (e.g., poorservice delivery, inconsistency), and themanagers (e.g., not visible or insufficientlyproactive).

After much discussion, training andmanagement were deemed to be the majorcontrollable causes of the high variability indining duration. Even though Chevys hadcorporate standards for the duration of eachmeal segment, all employees learned thosestandards in training, and a manager wasalways on duty, somehow enforcement ofthese standards was lax.

Payment processPayment processPayment processPayment processPayment process. Analysis of the timestudy data indicated substantial delays at thebeginning and end of the meal. Chevyschose to focus its attention chiefly on the

Because the table mix did not match thecustomer mix, the restaurant struggled toexceed 50-percent occupancy, even at its

busiest times.

Continued on page 28



Making Revenue Management Work for You

By analyzing its service processes and table mix, Chevys Arrowhead wasable to increase revenue by approximately 5 percentage points more thanthe other two Chevys that we examined. This performance boost came fromits improved table mix, changes in the service delivery, and improvedtraining. Seat occupancy and RevPASH increased, dining duration andvariation in the duration decreased, and revenue and profitability increased.If you want to improve your restaurant’s revenue in a manner similar toChevys Arrowhead, you should first establish your restaurant’s baselineperformance. Collect at least a month of detailed POS data and analyze yourseat occupancy, average check, RevPASH, party mix, and dining duration byday of week and hour of day. In addition, hire someone to conduct timestudies of your restaurant during your busy periods (this is an ideal part-timejob for a student).

After you have established your baseline performance, sit down with yourmanagement team and staff members to make sense of what you havediscovered. Discuss what might be driving your performance and pinpointspecific areas in need of improvement.

When developing a strategy, focus on your busy periods, establish specificperformance goals, and determine feasible ways of meeting these goals. Inaddition, be sure to assess the financial effects of any given strategy. Properimplementation of the strategies is probably the most difficult part of theprocess, but it also is the most important. Implementation requires a strongmanagement team, good training, and a willingness to try new things (suchas changing the table mix or hiring more employees).

Finally, when you have made all of the changes, reevaluate yourperformance after about two months by comparing your current results toyour baseline statistics. Gather additional POS data and conduct additionaltime studies to see whether your efforts have paid off.

For more information on developing a revenue-management program, youcan read the articles mentioned in this report, take the courses in restaurantrevenue-management from e-Cornell (www.ecornell.com), or enroll in arestaurant-revenue-management course offered through the Cornell HotelSchool’s Professional Development Program (PDP).—S.E.K.



end-of-meal delays. Payment delays could becaused by a number of problems, includingtechnology (e.g., credit card authorization,insufficient number of POS terminals),personnel (e.g., servers were inattentive orworking on other tasks, servers were notnoticing when guests wanted to pay, serverswere not picking up the completed checkfolder before guests left), and customers(e.g., customers were not ready to pay,customers did not ask for the check, custom-ers were uncomfortable leaving the com-pleted check folder on an unattended table).

Managers presented the staff with theresults of the baseline analysis and discussedthe payment process. Many servers said thatcredit card processing was slow, particularlyduring busy shopping periods, and thatservers often grew impatient with waiting forthe card approval, returned to the restaurantfloor to serve other customers, and some-times forgot about the bill that was beingprocessed. In addition, many servers saidthat they felt uncomfortable with picking upthe completed paid folio because theythought that the guests would think they weretoo eager to see their tip. Interestingly,several guests had complained or com-mented that the servers had not picked upthe completed check folder quickly, and thatthe customers felt uncomfortable leavingtheir credit card number or cash on theempty table, where other people couldconceivably have access to it.

BussingBussingBussingBussingBussing. Bussing was also a problem. Ittook nearly three minutes to clear and resetthe table. If this time could be reduced,

particularly when the restaurant was on await, more customers could be served,perceived service quality would increase(since seated guests would not have to belooking at a dirty table, and waiting guestswould not have to see empty tables), andrevenue would increase. The slow bussingcould be attributed to one or more of thefollowing factors: communication (e.g., poorcommunication between servers andbussers), processes (e.g., poor pre-bussing,not enough space for dirty dishes in kitchen),and staffing (e.g., insufficient bussers, lack ofsense of urgency).

During the staff discussion, bothbussers and servers pointed out that pre-bussing was inconsistent. In addition, manyservers were neglecting to pre-bus theirtables, feeling that it was the bussers’ respon-sibility, while the bussers felt that theirresponsibilities were to provide chips andsalsa and to clear and reset the table oncethe guests had left. In addition, many serverswere not “tipping out” the bussers, sobussers felt little incentive to help out withpre-bussing. Finally, the bussers pointed outthat the recent changes in the bussingprocess (previously they had used buckets toclear the table, but currently they had to usetrays) as being a potential contributor to theproblem.

Step #3: Developing a Revenue-managementStrategy

The two major goals were to reduce diningduration by five minutes and to increase seatoccupancy by 10 percent. An ancillary goalwas to reduce the standard deviation of totaldining time by 30 percent. Managers esti-mated that these changes would increaserevenue by 5 percent during the nine busyhours per week.

After determining the goals, themanagers started their implementation byspecifying the busy (hot) and slow (cold)

Reducing the variability in mealsegments was as important as trimmingthe actual duration of those segments.



periods by day of week and hour of day.“Hot” periods were defined as times whencustomers were waiting to dine, and all otherperiods were designated as “cold.” Therestaurant’s nine hot hours were Friday, 5:00to 9:00 PM; Saturday, noon till 2:00 and 6:00to 8:00; and Sunday afternoon from 1:00 to2:00). These were the busy hours that werethe focus of the revenue managementprogram’s goal of a 5-percent revenueincrease.

The goal of increased seat occupancycould be achieved by attracting morecustomers, providing a better table mix somore customers could be accommodated, orreducing the dining duration so morecustomers could be served. There was nopoint in attempting to attract more custom-ers, because restaurant already had morecustomers than it could currently serveduring its hot periods (by definition).Moreover, since customers could easilydefect to another restaurant even after theyput their name on the waitlist, attractingmore customers made no sense. Not onlythat, but the restaurant’s existing table mixand dining duration would not allow therestaurant to serve additional customers.Thus, the managers’ focus was on improvingthe table mix and reducing dining duration.

Of the two strategic levers available torestaurants for revenue management (i.e.,duration control and price), the Chevysmanagers chose to focus on duration, orprocess control. The team concentrated onways to better manage arrivals, tighten mealduration, and reduce the amount of timebetween customers.

Arrival Management

As mentioned earlier, customer arrivals canbe managed either internally (with policiesthat do not directly involve customers) orexternally (with methods that involve cus-tomers). Since it did not take reservations,the restaurant could not use methods

relating to reservations (including over-booking and better forecasting). Instead, itstactics had to focus on an internal policy(namely, a better table mix) and an externalprocedure (i.e., improved hosting). Althoughthe management team considered revisinghosting procedures, it first took on the tablemix—a tactic that would be least visible tocustomers and perhaps offer the bestrevenue rewards. The team did reviewpotential hosting changes, including bettertraining, technology alternatives (such as atable-management system or buzzers tonotify customers that their table was ready),and assignment of management personnel tothis position.

It seemed to all involved that animproved table mix would have the mosteffect on revenues. The optimal table mixwould be one that maximized revenue forthe restaurant’s existing party mix. A table-mix simulator developed by Gary Thomp-son was used to develop the optimal tablemix for the restaurant as well as a set of“near-optimal” mixes (near optimal wasdefined as within 1.5 percent of optimal).18

The optimal table mix maintained thesame number of seats, but changed the tablemix from fifty-three 4-tops and three 6-topsto forty 2-tops, twenty-four 4-tops, five 6-tops, and three 8-tops. Chevys hired arestaurant designer, informed her of theoptimal table mix, and set her to work withthe following two stipulations: maintain thesame number of seats, and make the designattractive and useable. The implementationof the table mix will be discussed below.

Duration Management

The management team then turned toprocess-control issues. Like arrivals, durationcan also be managed internally or externally.Internal methods include changes in theservice-delivery process, changes in the

18 See: Thompson, CHR Reports, loc.cit.



menu, and changes in the restaurant ambi-ence (e.g., color, sound, and comfort).External methods include providing cues tocustomers that it is time to leave and setting apre-specified time that customers can use atable. Again, the restaurant focused oninternal methods to avoid the possibility ofoffending customers with external tactics.

The time-study results and the analysisof potential causes led to a focus on thepayment process. Specifically, the goal was toreduce the mean payment time by one-and-one-half minutes and to reduce the variationaround that mean.

The team also wanted to reduce theamount of time between parties by 30seconds. The team judged that bussing tooktoo long and focused on that as the step tobe tightened up.

Finally, the team wanted to reduceoverall meal duration variability. Based onthe analysis of the causes, the focus was onimproving training and strengtheningmanagerial oversight.

Potential Revenue Effects

Before assessing the revenue improvementsstemming from increased occupancy anddecreased dining duration, the annualrevenue during hot periods was calculated.To review, the 230-seat main dining roomhad nine hot hours per week. Beforerevenue-management interventions, anaverage “hot” seat occupancy of 50 percent,an average “hot” check of $12.73, and anaverage dining time of 53 minutes. Annualsales for the restaurant in 2001 were$2,358,874.

Under baseline conditions, the restau-rant made $775,617 per year during its ninehot hours. Thus, about one-third of itsannual revenues were made during its nineweekly hot hours.

Then, the potential impact of a 20-percent increase in seat occupancy and a 5-minute decrease in dining duration was

assessed. If hot seat occupancy increasedfrom 50 percent to 60 percent and diningduration remained the same, the annualrevenue potential would increase by$155,123 [((.6 - .5)/.5) * $775,617]. On theother hand, if hot dining duration deceasedfrom 53 minutes to 48 minutes and seatoccupancy remained the same, the annualrevenue potential would increase by $80,793[((53 - 48)/48) * $775,617]. Finally, if seatoccupancy increased from 50 percent to 60percent and dining duration dropped to 48minutes, the annual revenue potential wouldincrease by $252,076 [(.6/.5 * 53/48 -1) *$775,617]. Even if only half of that potentialrevenue ($126,000) could be achieved, therestaurant would experience a 5-percentincrease in annual revenue.

Step #4: Implementation

Implementation largely revolved aroundtraining staff and implementing the new tablemix. The management team began theprocess by giving all staff members a briefingon revenue management, showing themsome of the results from the baselineanalysis (particularly the time study results),and splitting them into function-based teams(i.e., all servers, all bussers, all kitchen staff)for discussion of possible approaches toresolving observed problems. This team-building approach served the restaurant well.The staff members bought into the revenue-management program, and managers wereable to obtain many useful ideas in theensuing discussions.

Based on the discussion sessions,management decided to increase the empha-sis on pre-bussing in training, to reinforce theneed for pre-bussing for existing employees,and to improve managerial oversight of thisprocess. In addition, management decidedto train all servers on the importance ofquickly processing the check and on pickingup the completed check folders from thetable.



Since part of the problem with thepayment process seemed to relate to tech-nology, the chief information officer of theChevys chain decided to investigate othertechnological solutions to the slow process-ing times. A particular problem was therestaurant’s mall location, which meant thatit competed for credit-card-processing timewith the mall’s many stores. If Chevys couldobtain a dedicated payment server, it couldimprove check-processing time.

Table Mix

Because the restaurant had been open formore than nine years with no updating, thecorporate staff decided that the redesignwould not only involve reconfiguring thetable mix, but also repainting the restaurant,relocating the host and service stations, andseveral other aesthetic improvements.

Management wanted an attractivedesign that allowed ample space for move-ment but did not make customers feelcrowded. Constrained somewhat by therestaurant’s footprint and the structurallimitations of the building (several supportpillars had to be maintained and the hoststand couldn’t be moved too far because ofelectrical considerations), the designer’sinitial plan fulfilled the requirements ofmaintaining the same number of seats andusing the optimal table mix. To accomplishthis the designer employed a variety ofdesign approaches, such as half-walls,banquettes, high-top tables, and well-placedbooths.

When the management team, thecorporate team, and the designer evaluatedthese initial plans on site, they made severalchanges to accommodate the building’sspecific attributes and certain customercharacteristics (for example, young familieswith babies had large carriages that neededample space). The management teamdecided to adopt one of the near-optimaltable mixes (thirty-nine 2-tops, thirty-five

4-tops, and two 6-tops) to achieve its goals ofattractive design with attention to customerneeds—an increase of 20 tables (but notmore seats). The simulation results showedthat this table mix would provide resultswithin 1.3 percent of the optimal mix.19

Construction was set to begin theMonday after Valentine’s Day 2002. Themanager had wanted to start constructionbefore Valentine’s Day so that she couldhave the benefit of the added deuces on thatoccasion, but the corporate staff decided that

the revised table mix coupled with the highValentine’s Day demand might provedifficult to manage. All construction wasscheduled to occur during the hours that therestaurant was closed so as not to disruptnormal operations—which continued duringthe construction period. All half-walls andbanquettes were constructed off-site so thatthey could be quickly installed.

The increase from 56 tables to 76tables meant that more customers could beseated and wait time would be reduced.Offsetting those good outcomes, the newtable mix increased the load on the kitchen,increased the number of servers required,and required the purchase of additional tablesupplies.

The reduction in wait time trimmedthe buffer time that the kitchen had formerlyenjoyed. To offset that problem, therestaurant’s managers decided to add anexpeditor who worked the middle lineduring hot periods (four shifts per week).

Changes in the restaurant’s table mix requiredadjustments in staffing, including new

positions to accelerate operations.

19 Kimes and Thompson, working paper 04-23-03.



This person facilitated communicationbetween the broiler line and the enchiladaline to help speed delivery to guests. Twopeople were trained for this position at a rateof $9 per hour for the four shifts of threehours each, for a weekly cost of $108.

The additional 20 tables and newcorporate rules on station size required theaddition of approximately five server posi-tions during hot periods, as the serverstations were reconfigured into stations offour to five tables. The cost of the additionalservers was $127.80 per week ($2.13 perhour for four three-hour shifts).

The additional tables also requiredadditional tablecloths, plates, glasses, chipbaskets, salsa dishes, Texas holders, andcutlery. The cost of the supplies (not includ-ing tablecloths) was $1,395.

Service-delivery Changes

As discussed above, the service-deliveryimprovement strategies focused on the endof the meal—specifically, the payment andbussing processes. Changes involved im-proved training, more diligent management,and certain additional staff.

Entrée delivery to check requested.Entrée delivery to check requested.Entrée delivery to check requested.Entrée delivery to check requested.Entrée delivery to check requested.The baseline time between the entréedelivery and the check delivery was approxi-mately 23 minutes. The team did not wish torush guests, but believed that better pre-bussing would move the meal along a bitmore quickly. For this reason, both serversand bussers were trained to do a better jobof pre-bussing. In addition, the store man-agement hired a stacker, who was respon-sible for scraping and stacking all used dishesfor the dishwashers. This position’s pay was$5.15 per hour for the four hot shifts ofthree hours each, for a weekly cost of$61.80.

Check requested to change returnedCheck requested to change returnedCheck requested to change returnedCheck requested to change returnedCheck requested to change returned.In the test it took nearly six minutes from thetime the check was requested to the time thechange was returned. In an effort to trim that

time, servers were trained either to drop thecheck upon clearing the entrée or to be alertfor signs that the guest would like to leave. Inaddition, as I explained above, the corporateoffice sought to speed up the slow creditcard authorizations.

Change returned to customer depar-Change returned to customer depar-Change returned to customer depar-Change returned to customer depar-Change returned to customer depar-turetureturetureture. It originally took about four-and-a-halfminutes from the time the change wasreturned until when customers left. Themanagement team found that many custom-ers did not want to leave their credit cardreceipt on an empty table and stayed at thetable until a server came to pick up thatreceipt. Consequently, servers were trainedto pick up the completed check folder assoon as the customer was done with it. Inaddition, the team hoped that the enhancedpre-bussing would signal the customer that itwas time to leave.

Customer departure to completion ofCustomer departure to completion ofCustomer departure to completion ofCustomer departure to completion ofCustomer departure to completion ofbussingbussingbussingbussingbussing. In the test run, it took nearly threeminutes to clear and reset a table. The majorcause of this delay was the lack of pre-bussing by servers and the small spaceavailable in the kitchen. Improved pre-bussing and the stacker position would helpreduce this time.

Step #5: Evaluation

About a month after the restaurant had beenoperating with the new table mix and revisedprocedures, its performance was reevaluated.Specifically, updated POS data were col-lected so that seat occupancy, RevPASH,and meal duration could be recalculated. Inaddition, a new series of time studies wasconducted. Finally, a financial analysis wasperformed.

Additional POS data were collected todetermine the change in seat occupancy.During the pre-test period, seat occupancyduring the hot periods averaged 50 percentwith a peak occupancy of 55 percent, whileduring the post-test period, average seatoccupancy during the hot periods increased



to 59 percent with a peak seat occupancy of82 percent.

RevPASH showed a similar improve-ment. Average RevPASH for the hot periodduring the pre-test period was $5.85. Itincreased to $6.32 during the post-testperiod.

Another series of time studies (82observations) was conducted to analyze theeffects of the duration-management strategiesthat had been adopted. The mean meal timedropped from 53:15 during the pre-testperiod to 50:56 for the post-test period. Thestandard deviation of total meal durationdropped from 22:46 to 15:09. The goal of afive-minute drop in dining duration was notachieved, but the variation was decreasedconsiderably.

Most duration-management strategieshad focused on the end of the meal. Theteam noticed considerable time reductions inthat segment of the meal, although all mealparts were tightened. Decreases were notedin the following areas (numbers in parenthe-ses refer to the variation, or standard devia-tion, in the mean test times):

(1)(1)(1)(1)(1)Seat to greet: pre-test: 2:20 (2:01),post-test: 1:38 (1:30). The restaurantdropped this time by 42 seconds andreduced the standard deviation by 25percent. This could be attributed to theimproved training and awareness.

(2)(2)(2)(2)(2)Entrée to check dropped: pre-test:22:41 (12:19), post-test: 21:47 (6:44). Thetotal time dropped by about a minute, butthe standard deviation went down by over 40percent. This can be attributed to theimproved pre-bussing.

(3)(3)(3)(3)(3)Check dropped to change returned:pre-test: 5:40 (2:55), post-test: 2:55 (1:56).The restaurant was able to cut this timenearly in half and was able to reduce thestandard deviation by over 60 percent. Thiscan be attributed to training and the en-hanced awareness of the servers of theimportance of this process.

(4)(4)(4)(4)(4)Change returned to departure: pre-test: 4:27 (6:38), post-test: 4:09 (5:09). Therestaurant did not have complete control ofthis time, since customers choose when theywant to leave. The time decreased slightly, asdid the standard deviation, which mayindicate that the emphasis on picking up thecheck folder had worked.

(5)(5)(5)(5)(5)Order to entrée: pre-test: 11:31(5:06), post-test: 11:25 (4:06). While this isnot a substantial decrease in time, the factthat the kitchen was able to maintain thesame time to prepare and deliver entréeseven with the increased number of custom-ers from the new table mix was impressive.Restaurant managers attributed this to thestrong kitchen staff and its willingness toexperiment with various approaches tohandling the increased demand.

The focus had been on tightening theend of the meal, and that focus showed inthe results. In contrast to the other mealsegments, two of the pre-dining times (greetto drinks, drinks to order) increased—

representing areas for future improvement.Also, the restaurant had been doing anexcellent job of reseating tables once theywere cleared. This time increased by 10seconds from pre-test to post test.

Financial Analysis

Although the estimated target was a 5-percent increase in revenue, it was necessaryto test the actual effects of the process-control tactics. Since RevPASH had in-creased, restaurant revenue had increased,but it was unclear whether that increasestemmed from the revenue-managementproject or because of other, general marketconditions. To control for this possibility,

The restaurant’s new table mix increasedpeak seat occupancy by nearly 50 percent.



financial performance for both the Arrow-head restaurant and two other Chevysrestaurants in the same market was analyzed.For the seven weeks before implementation,the Arrowhead outlet had experienced a 5.7-percent drop in revenue (comparing 2001 to2002), while the other two restaurants hadsuffered a 10.6-percent drop.

A similar RevPASH analysis wasconducted after the new table mix and othermeasures had been implemented. TheArrowhead restaurant had experienced a2.0-percent increase in revenue from 2001 to2002, while the 2002 sales for the controlrestaurants were 8.0-percent less than thesame period in 2001. Thus, the Arrowheadoutlet had realized a 7.7-percent increase inrevenue from pre-test to post-test, while thecomparable restaurants had increased salesby 2.6 percent. The difference (5.1 percent-age points) was attributed to the revenue-management interventions.

The new table mix and customer-volume changes were not without cost. Theremodeling cost approximately $49,000,additional small wares cost approximately$1,400, and labor costs increased by $15,000per year.

The Arrowhead restaurant hadapproximately $2.4 million in sales in 2001.Based on the projected increase of 5.1-

percent calculated above, there was anestimated annual sales increase of approxi-mately $122,000. Chevys had a 45.5-percentEBITDA flow-through, which meant thatapproximately $55,000 of this amount wouldgo to the bottom line. The total cost of theproject was around $50,000. The restaurantcalculated its cash-on-cash return at 107.7percent with an 11.1-month payback.

Summary and ConclusionBy implementing revenue-managementtactics, Chevys Arrowhead has been able toincrease revenue by approximately 5 per-cent. The improved table mix, the changesin the service delivery, and the improvedtraining led to the improvement in therestaurant’s performance. Seat occupancyand RevPASH increased, dining durationand variation decreased, and revenue (and,hence profitability) increased.

Other restaurant operators couldrealize similar results by carefully analyzingtheir current performance, determining thecauses of that performance, and developingappropriate strategies to improve it. Changesin table mix and problematic service-deliveryprocesses hold particular promise, but onlywith proper implementation that emphasizestraining, employee buy-in, and enhancedmanagement. ■

Sheryl E. Kimes, Ph.D., is a professor at the Cornell University School of Hotel

Administration ([email protected]). She is the author of numerous articles on yield

management theory and applications for hotels, golf courses, and conference centers, in

addition to the restaurant industry.

About the Author



The Center for Hospitality ResearchA T CO R N E L L UN I V E R S I T Y

CHR REPORTS

FOR THE

RESTAURANT INDUSTRY

www.chr.cornell.edu

Dedicated or Combinable? A Simulation to Determine Optimal

Restaurant Table Configuration, by Gary M. Thompson, Ph.D.

Multiunit Restaurant-productivity Assessment: A Test of Data-

envelopment Analysis, by Dennis Reynolds, Ph.D., and Gary M.

Thompson, Ph.D.

Strengthening the Purchaser–Supplier Relationship: Factors that Make a

Difference, by Judi Brownell, Ph.D., and Dennis Reynolds, Ph.D.

Key Issues of Concern for Food-service Managers,

by Cathy A. Enz, Ph.D.

Also of interest:

Tools for the Hospitality Industry #1: Turnover Cost Evaluator, by Tim

Hinkin, Ph.D., and Bruce Tracey, Ph.D.

Tools for the Hospitality Industry #2: MegaTips: Scientifically Tested Ways

to Increase Your Tips, by Michael Lynn, Ph.D.

“As both a contributor and reader, I have found the Cornell Quarterlyto be a source of powerful insights that are also of great practicalvalue.”

Peter Yesawich, Ph.D. President and CEO, Yesawich, Pepperdine, Brown & Russell

““

“““The Cornell Quarterly is a leading source of new lodging and food-service topics and ideas that are both conceptual and practical for planning and operations.”

Bjorn Hanson, Ph.D.Global Hospitality Industry Managing Partner, PricewaterhouseCoopers

2455 Teller Road, Thousand Oaks, CA 91320 U.S.A.Telephone: (800-818-7243 (U.S.) • (805)-499-9774 (Outside of the U.S.)

FAX: (800-583-2665 (U.S.) • (805)-499-0871 (Outside of the U.S.)Email: [email protected] or visit www.sagepub.com

1032094/S01204

KKeep ahead of the world's fastest-growing industry with Hospitality'sPremier Management Publication! Subscribe to the Cornell Hotel andRestaurant Administration Quarterly today!

Cornell Hotel and RestaurantAdministration QuarterlyHospitality’s Premier Management Publication!

Just some of the vital topics in recent issues:• ADR Rule of Thumb: Validity and Suggestions for its Application• The Effects of Smoke Free Regulations on Employment and

Sales• Best Practices in the U.S. Lodging Industry • Hotel-Room Inventory Management• The Impact of Internet Discounting on the Hotel Industry• Strategic Management Model for Assessing a Firm's Strategic

Position• Advertising Strategies for Hospitality Services• Actions and Behaviors Needed for Successful Leaders• Changes in the Legal Relationship between Hotel Owners and

their Management Companies and Franchisors • The Costs of Turnover• Restaurant Revenue Management Strategy

Contact:

Whether you are a CEO, an

operations manager, an executive

trainee, or an undergraduate,

there is something for you in every

informative issue of

CQCQCQCQCQCornell

Hotel and RestaurantAdministration

Quarterly!

Kimes rest.revenue mgt.1.26.04

Documents

Transcript of Kimes rest.revenue mgt.1.26.04