© 2004 CORNELL UNIVERSITY
DOI: 10.1177/0010880403260107
Volume 45, Issue 1 52-67
1A0.1177/00108R80403260107TICLE
Restaurant Revenue
Management
Implementation at Chevys Arrowhead
by SHERYL E. KIMES
Revenue-management tools can be used by restau- Revenue management can be described as sellingrant managers to analyze the effects of process-con- the right inventory unit (in this case, a restauranttrol changes. A dinner house seeking to shift demand seat) to the right customer for the right price and
and to achieve greater facility utilization during busy at the right time. The determination of right in that def-
times analyzed the factors that caused delays in the inition—and, indeed, the essential element of revenue
service process—and thus increased the guest management—entails achieving the greatest possible
queue. Although the restaurant was able to hasten the
actual dining time, much of the slack was found in the revenue contribution for the company while delivering
processes that occurred before and after the actual the greatest value or use to the customer. Companies
dining period. Moreover, the restaurant managers using revenue management have reported increases in
were able to analyze customer-arrival and market-mix revenue ranging from 2 to 5 percent.
1
data in relation to the restaurant’s table mix. Seat Prevalent in airline, cruise line, hotel, and rental-car
occupancy was improved by matching the table operations,
2 revenue management can also be applied
arrangement to the customer mix, and table turns to restaurants and golf courses.3 Just as hotels offer
were increased by improving the kitchen operations room rates that change according to demand or in rela-
so that front-of-the-house functions could be tight- tion to customers’ length of stay, restaurants can offer
ened up. In particular, end-of-meal steps were different menu prices based on customers’willingness
speeded up. As a result of its process improvements, to dine (or make reservations) during slack times. The
the restaurant enjoyed revenue growth greater than classic, though not scientific, examples of restaurants
that of comparable restaurants. offering price incentives to build demand during slow
periods are happy hour, early-bird specials, and
Keywords: restaurant revenue management;
Chevys Arrowhead restricted-use coupons. In addition, some restaurants
52 Cornell Hotel and Restaurant Administration Quarterly FEBRUARY 2004
RESTAURANT REVENUE MANAGEMENT REVENUE
take advantage of high-demand periods cludes with an evaluation of the Chevys
(e.g., Easter or Mother’s Day brunch) by revenue-management strategy and recom-
offering only special (premium price) mendations for how other restaurateurs
meals. Those examples, however, do not can implement revenue management.
constitute actual revenue management,
because they are broad-brush promotions Revenue Management
unrelated to an analysis of actual demand. Restaurant operators have two main stra-
Revenue management generally in- tegic levers that they can use to manage
volves a combination of demand-based revenue: price and meal duration.5 Price is
pricing and control over the timing of cus- a fairly obvious target for manipulation,
tomers’ use of a service. Focusing on the and many operators already offer the
timing aspect of revenue management, price-related menu promotions that I just
restaurants have experimented with con- mentioned to augment or shift peak-
trolling customers’dining time by increas- period demand. More sophisticated mani-
ing the efficiency of service delivery to pulations of price include daypart pricing,
shorten the amount of time customers day-of-week pricing, and price premiums
spend at the table.4 That latter aspect of or discounts for different party sizes,
revenue management—duration control— tables, and customer types.6
is at the heart of this article. Managing meal duration can be more
Given that restaurants should theoreti- complicated than adjusting prices since
cally be able to apply revenue manage- most restaurants do not explicitly sell time
ment, I wondered how well that might periods. Add to that restaurateurs con-
work in practice. In this article, I discuss cerns about rushing customers, and con-
how Chevys Freshmex Restaurant devel- trolling meal duration can be a sensi-
oped, implemented, and evaluated a reve- tive matter. However, as I explain here,
nue-management program involving pro- duration control has great potential in a
cess analysis and duration control at one revenue-management strategy.
of its restaurants. The purpose of this To control duration, managers can use
article is to illustrate how to use revenue- either internal means (i.e., those that do
management tools effectively and to pro- not involve customers) or external means
vide managers with an easy-to-follow (that do involve customers).7 The chief
implementation process for their own internal duration-control methods involve
restaurants. regulating and redesigning service pro-
I start with a brief introduction to reve- cesses (including speeding up service to
nue management, followed by a descrip- promote customer turnover and providing
tion of the Chevys restaurant that provided an optimal table mix), forecasting cus-
data for this study. In so doing, I analyze tomer arrivals (i.e., forecasting the timing
the restaurant’s baseline performance, and party-size mix of arriving customers),
including seat occupancy, revenue per and implementing inventory controls
available seat hour (RevPASH), party- (usually through overbooking, if a restau-
size mix, and dining duration. I also ana- rant takes reservations). External methods
lyze and discuss the possible causes of include booking fees or guarantees (e.g.,
performance. After reviewing the revenue- having guests guarantee reservations on
management strategies for duration a credit card) and such behavioral ap-
control, I discuss how managers imple- proaches as restricting the length of time
mented those strategies. The article con- that customers can use the table. Not sur-
FEBRUARY 2004 Cornell Hotel and Restaurant Administration Quarterly 53
REVENUE RESTAURANT REVENUE MANAGEMENT
prisingly, most firms have chosen to man- ends, draws a variety of customers, includ-
age duration internally, so as not to risk ing shoppers, couples, and families. This
dissatisfied customers. restaurant has 230 seats in the main dining
room, an additional 50 seats in the bar, and
Step by step. Restaurant managers patio seating that is available from March
who want to implement revenue manage- through November. Fifty-three tables in
ment should first establish their restau- the main dining room were 4-tops, and the
rant’s baseline performance so that they remaining three tables were 6-tops. As is
have a measure of comparison for any typical in many restaurants, Chevys
duration-control interventions. Baseline Arrowhead is busy for weekend dinners
performance includes information on and lunches, when customers often wait
average check, seat occupancy, RevPASH, over an hour for a table.
meal duration, and party-size mix. Sec-
ond, managers should study the causes of The Five-step Revenue-
speedy and slow service performance so management Approach
that they can develop a strategy to help The managers and I used the five-step pro-
employees overcome factors that slow cess explained here to develop a revenue-
performance. Third, managers should management strategy for the restaurant.
develop a revenue-management strategy, Rather than attempt price-related promo-
which can involve a mix of duration con- tions, we focused on internal revenue
trol and demand-based pricing. Fourth, management, specifically related to the
they should implement the strategy and,
8 duration of the dining experience. Al-finally, monitor its outcome. though the data presented here are specific
to Chevys Arrowhead, the process and
Problem Setting analyses described can be applied to any
Chevys Freshmex Restaurants, a U.S.- restaurant.
based chain of over one hundred midscale
Mexican-style restaurants, was interested Step 1: Establish a Baseline
in increasing its units’ profitability by
implementing revenue management. The first step in the process was to estab-
Chevys prides itself on its use of fresh lish the restaurant’s baseline performance.
ingredients, its lively atmosphere, and its Baseline statistics were drawn from two
friendly staff members. The chain has per- four-week periods of point-of-sale (POS)
formed well, but its managers noticed that data and detailed time studies. Using these
customers had to wait a long time on data, we analyzed average check per per-
weekend nights in many restaurants, and son, RevPASH, seat occupancy, meal
guests sometimes complained about the duration (from both the POS data and the
length of time it took to get through a time studies), and the party-size mix by
meal. day of week and hour of day. We excluded
Chain executives chose to test revenue- data recorded after 10:00 p.m. because of
management applications at the Chevys the limited number of observations from
Arrowhead, located in a busy shopping those late hours.
mall in Glendale, Arizona, a suburb of The POS data showed that the average
Phoenix. The restaurant, which is open check per person for the 230-seat main
from 11:00 a.m. to 11:00 p.m. on week- dining room was $12.70 (see Exhibit 1).
days and 11:00 a.m. to midnight on week- Calculated by day of week and hour of
54 Cornell Hotel and Restaurant Administration Quarterly FEBRUARY 2004
RESTAURANT REVENUE MANAGEMENT REVENUE
Exhibit 1:
Average Check by Day of Week and Hour of Day
Hour of Day
Day of Week 11:00 noon 1:00 2:00 3:00 4:00 5:00 6:00 7:00 8:00 9:00
Sunday $12.26 $11.72 $11.73 $12.04 $10.74 $11.68 $12.24 $11.73 $11.38 $10.25
Monday $10.08 $10.04 $10.24 $11.59 $11.22 $11.06 $11.82 $10.70 $12.55 $10.37 $11.47
Tuesday $10.35 $9.39 $9.61 $10.14 $9.76 $11.77 $11.46 $11.80 $12.54 $11.74 $11.42
Wednesday $10.11 $10.01 $9.35 $9.86 $8.72 $11.32 $12.02 $12.65 $12.79 $11.32 $10.65
Thursday $9.95 $9.03 $9.78 $10.48 $9.16 $9.75 $11.75 $13.33 $13.66 $14.47 $12.82
Friday $9.02 $10.14 $9.86 $9.84 $8.59 $12.01 $12.85 $12.70 $13.84 $13.61 $13.03
Saturday $11.65 $10.54 $12.40 $12.19 $10.78 $12.41 $12.42 $13.55 $14.22 $12.57 $11.16
Note: Numbers are based on eight weeks of point-of-sale data from fall 2001.
day, average check ranged from $8.59 at most restaurants) does not track seat occu-
3:00 on Fridays to $14.47 at 8:00 on pancy, we had to derive that statistic.
Thursdays. The highest check averages Because RevPASH is defined as seat occu-
occurred on Thursday, Friday, and Satur- pancy multiplied by average check, we
day evenings, while the lowest checks could calculate seat occupancy by divid-
occurred for weekday lunches. ing the average check by the RevPASH
RevPASH provides a good estimate of and then multiplying the result by the
seat occupancy combined with the aver- average meal duration (in hours) for each
age check.9 That statistic is useful in two time period, with the results shown in
ways, the first being the important matter Exhibit 3.
of how much revenue the restaurant is The hour with the highest seat occu-
realizing in each time period. RevPASH pancy was from 6:00 to 7:00 on Fridays,
was calculated by first determining the and that was a mere 61 percent of available
total hourly revenue from the main dining seats. Overall, the highest seat occupan-
room for each day of the week and then cies occurred in dayparts with the highest
dividing the hourly revenue by the 230 RevPASH, namely, Fridays from 5:00 to
seats in the main dining room, as shown in 9:00, Saturdays from 6:00 to 8:00, Satur-
Exhibit 2. RevPASH ranged from $0.27 days from noon to 2:00, and Sundays from
on Mondays at 9:00 to $7.03 on Fridays 1:00 to 2:00. Lowest seat occupancies
at 6:00. The highest RevPASH was (under 15 percent) occurred on weekdays
recorded on Fridays from 5:00 to 9:00, on before noon, between 2:00 and 5:00, and
Saturdays from noon to 2:00 and 6:00 to after 9:00 p.m. on most days.
8:00, and on Sundays from 1:00 to 2:00
p.m. The lowest RevPASH was experi- Dining duration. Check opening and
enced each day after 9:00 p.m., before closing times allowed us to calculate the
noon, and between 2:00 and 5:00 on all mean and standard deviation of dining
weekdays. duration by day of week and hour of day.
We knew that duration figures thus calcu-
Occupancy. The second application of lated could be slightly inaccurate because
RevPASH was in the derivation of seat use the opening of the check did not necessar-
or occupancy. Since this restaurant (like ily correspond to when the customers
FEBRUARY 2004 Cornell Hotel and Restaurant Administration Quarterly 55
REVENUE RESTAURANT REVENUE MANAGEMENT
Exhibit 2:
RevPASH by Day of Week and Hour of Day
Hour of Day
Day of Week 11:00 noon 1:00 2:00 3:00 4:00 5:00 6:00 7:00 8:00 9:00
Sunday $2.19 $4.29 $5.17 $3.95 $2.50 $2.87 $3.74 $3.44 $1.87 $0.54
Monday $1.87 $2.42 $2.45 $1.78 $1.09 $1.03 $2.81 $3.75 $1.96 $1.00 $0.27
Tuesday $1.11 $1.73 $1.45 $0.43 $0.28 $0.89 $1.98 $2.87 $2.19 $0.94 $0.44
Wednesday $1.32 $2.36 $1.87 $0.64 $0.48 $1.07 $2.11 $4.12 $3.44 $1.22 $0.35
Thursday $1.32 $2.05 $1.68 $0.69 $0.64 $1.05 $2.50 $4.67 $3.55 $1.19 $0.46
Friday $1.00 $2.89 $2.05 $0.89 $0.94 $1.62 $5.03 $7.03 $6.50 $5.22 $1.81
Saturday $2.63 $5.20 $5.58 $4.28 $3.11 $3.98 $4.88 $6.72 $6.25 $3.46 $0.98
Note: Numbers are based on eight weeks of point-of-sale data from fall 2001.
Exhibit 3:
Seat Occupancy of the Main Dining Room by Day of Week and Hour of Day
Hour of Day
Day of Week 11:00 noon 1:00 2:00 3:00 4:00 5:00 6:00 7:00 8:00 9:00
Sunday 20% 39% 49% 36% 26% 27% 34% 32% 18% 6% 20%
Monday 20% 27% 26% 17% 11% 10% 26% 39% 17% 11% 20%
Tuesday 12% 20% 17% 5% 3% 8% 19% 27% 19% 9% 12%
Wednesday 14% 26% 22% 7% 6% 10% 19% 36% 30% 12% 14%
Thursday 15% 25% 19% 7% 8% 12% 24% 39% 29% 9% 15%
Friday 12% 31% 23% 10% 12% 15% 43% 61% 52% 42% 12%
Saturday 25% 54% 50% 39% 32% 35% 43% 55% 49% 30% 25%
Note: Numbers are based on eight weeks of point-of-sale data from fall 2001.
were seated at the table and the closing of mation on course timing, detailed time
the check did not always reflect when the studies were conducted for busy weekend
guests actually left the table. That said, the dinner periods. A student observer timed
average meal duration for dinner (after one hundred parties over a several-week
4:00) was fifty minutes, with a standard period. The following ten different cate-
deviation of twenty minutes, while the gories were timed: when the party was
average meal duration for lunch (before seated, greeted by the server (and drink
4:00) was forty-four minutes, with a stan- orders taken), drinks delivered, order
dard deviation of sixteen minutes. The taken, entrée delivered, check requested,
averages and standard deviations did not check delivered and settled, departure,
vary much by day of week or by hour. table bussed, and table reoccupied by the
next party (see Exhibit 4). The average
Time study. Since the POS data meal time recorded in this manner (fifty-
included information only on total meal three minutes, fifteen seconds with a stan-
duration and did not have detailed infor- dard deviation of twenty-two minutes,
56 Cornell Hotel and Restaurant Administration Quarterly FEBRUARY 2004
RESTAURANT REVENUE MANAGEMENT REVENUE
Exhibit 4:
Time Study of Dining Experience (mean, standard deviation, and coefficient of variation between
events)
Mean Standard Deviation Coefficient of Variation
Preprocess
Seat to greet 2:20 2:01
Greet to drinks 3:52 4:24 0.67
Drinks to order 2:34 3:23
Production
Order to entrée served 11:31 5:06 0.44
In-process
Entrée served to check dropped 22:41 12:19 0.54
Postprocess
Check dropped to change returned
(final settlement) 5:40 4:54 0.82
Change returned to departure 4:27 6:38
Between customers
Departure to bussed 2:48 2:36 0.75
Bussed to reseated 0:56 1:04
Total dining duration 53:15 22:46 0.43
Note: Numbers are based on a time study conducted in fall 2001.
forty-six seconds) was a bit longer than subsegments with high standard devia-
the figure obtained from the POS data for tions and coefficients of variation, where
the same time periods, showing the dis- variation can be reduced. Generally
crepancies introduced by using check speaking, areas with a coefficient of varia-
opening and closing times as against tion over 0.5 should be targeted for
actual seating times. potential improvement.
Meal duration was broken into four While the in-process segment (meal
segments, those being preprocess, pro- consumption) was the longest (at nearly
duction, in-process, and postprocess. In twenty-three minutes), it had a marginally
addition, the between-customer processes weak coefficient of variation of 0.54. In
(i.e., bussing and reseating) were ana- contrast, the preprocess segment took just
lyzed. Each segment had one or more nine minutes but had a coefficient of varia-
subsegments (e.g., the preprocess step tion of around 0.67. Running a mean of
included seating, greeting, drink delivery, eleven and a half minutes, the production
and order taking). segment had a modest coefficient of varia-
We calculated the average duration and tion of less than 0.5. The postprocess seg-
standard deviation of each segment (along ment took around ten minutes, with a coef-
with the coefficient of variation, defined ficient of variation of 0.82. Finally, the
as the standard deviation divided by the between-customer processes took about
average, see Exhibit 4). Segments and four minutes but had a large coefficient of
subsegments with high average times rep- variation, nearly 0.75.
resent areas in which time savings (and
possibly revenue gains) can be achieved, Party composition. The majority of
and the same is true of segments and parties (about 70 percent) were just one or
FEBRUARY 2004 Cornell Hotel and Restaurant Administration Quarterly 57
REVENUE RESTAURANT REVENUE MANAGEMENT
Exhibit 5:
Party-size Composition
Party Size
1-2 3-4 5-8 9+
Sunday 63.9% 27.8% 7.2% 1.0%
Monday 71.1% 22.7% 5.2% 1.0%
Tuesday 72.2% 21.6% 4.1% 1.0%
Wednesday 72.2% 22.7% 5.2% 0.0%
Thursday 72.2% 20.6% 5.2% 1.0%
Friday 68.0% 24.7% 5.2% 1.0%
Saturday 62.9% 27.8% 5.2% 1.0%
Note: Numbers are based on eight weeks of point-of-sale data from fall 2001.
two people (shown in Exhibit 5). About 25 long, and why was there so much varia-
percent of the parties comprised three or tion? (3) Why were certain parts of the
four guests; just over 5 percent of parties meal experience (mostly at the beginning
included five to eight guests; and barely 1 and end of the meal) taking so long? Using
percent of parties had nine or more guests. fishbone analysis, we divided the possible
By analyzing the baseline data, we causes into five categories, as shown as a
found that Chevys Arrowhead exceeded table (not a fishbone) in Exhibit 6: equip-
50 percent seat occupancy for a mere five ment, methods, personnel, customers and
hours per week. This statistic surprised materials, and management and staff.10
everyone since we all were well aware of
customers’ waiting lines on weekends. Low seat occupancy. It was not hard to
More critically, even though the typical see that the major factor limiting seat
party size was small, no tables were 2- occupancy was the inappropriate table
tops, almost guaranteeing empty seats mix. Even though the majority of parties
even while customers waited. Looking at comprised one or two people, all the tables
the average dining time of fifty-three min- were intended for parties larger than that.
utes and its standard deviation of nearly This meant that even when all tables were
twenty-three minutes, one sees that the occupied, the restaurant would have many
greatest source of that variation occurred empty seats. It was no wonder that the seat
during the end of the meal and while the occupancy rarely exceeded 50 percent,
table was being reset between parties. even when customers were waiting for a
After developing the baseline, we wanted table.
to determine the possible causes of the Several other factors may have also
performance that we observed. contributed to the low seat occupancy. The
bussing process took nearly three min-
Step 2: Understanding the Causes utes. Thus, a reduction in the bussing time
(discussed below) could improve seat
The baseline analysis led to the following occupancy. The reseating process was
questions. (1) Why was seat occupancy so relatively efficient, and so a time reduc-
low even though customers were waiting tion there would have little effect on
for tables? (2) Why was dinner taking so occupancy.
58 Cornell Hotel and Restaurant Administration Quarterly FEBRUARY 2004
RESTAURANT REVENUE MANAGEMENT REVENUE
Exhibit 6:
Possible Causes of Poor Performance
Possible Cause Equipment Methods Personnel Customers Materials
Low seat occupancy Table mix Bussing Training Hard to find Wait list
Hosting Number Reneging
Communication Commitment
Compensation
Management
Meal duration Point-of-sale terminals Hosting Training Choose to linger Trays
and variation Credit-card Seating Number Unsure how to
authorization behave
Service stations Greeting Commitment Party size
Restaurant layout Food and Compensation
beverage delivery
Cooking Management
Check processing
Prebussing
Payment and Credit-card Check drop Training Choose to linger Check
folders departure authorization folders
Point-of-sale terminal Check pick-up Number Unsure how
Check processing Commitment to behave
Folder drop Compensation Uncomfortable
Folder pick-up Management
Bussing Stacking space Prebussing Training Bucket,
trays
Service stations Communication Number Cleaning
supplies
Hosting Commitment New place
settings
Compensation
Management
Meal duration and variability. While taurants’ managers were not consistently
meal consumption (the in-process stage) enforcing those standards. Even though
took the longest time of any step, we employee turnover was low, job commit-
focused instead on the segments with the ment varied, and compensation (espe-
highest coefficients of variation. Although cially for nontipped positions) was some-
we could possibly have worked to reduce times an issue. We reasoned that if training
the time required for meal consumption, and managerial oversight were improved,
its coefficient of variation was not alarm- the variation in meal duration would
ing. Moreover, the managers and I did not decrease. To reduce overall meal duration,
want guests to feel rushed while eating. though, it was necessary to study some of
Of the many possible factors shown in the end-of-meal processes in detail.
Exhibit 6, we decided that the primary
cause of the long and variable meal dura- Payment and departure. Several fac-
tion was related to personnel. Although tors slowed payment and departure. Serv-
Chevys has corporate time standards for ers were sometimes inattentive and slow
each step in the dining experience, the res- in processing and delivering the check and
FEBRUARY 2004 Cornell Hotel and Restaurant Administration Quarterly 59
REVENUE RESTAURANT REVENUE MANAGEMENT
in picking up the completed check folder. restaurant had ten hot hours per week,
That last factor was a sticking point which became the focus of the revenue-
because of the restaurants location in a management program.11
mall, with its high foot traffic. A notice- The two major goals were to reduce
able number of guests stayed until the dining duration by five minutes and to
server removed the signed credit-card slip. increase seat occupancy by 10 percent
We surmised that those diners felt uncom- during the hot periods. An ancillary goal
fortable leaving their completed check was to reduce the standard deviation of
folder on a vacated table. Although total dining time by 30 percent. We
another problem with credit-card authori- expected these changes to increase reve-
zation was that the actual machine trans- nue by at least 5 percent during the ten hot
action was sluggish, the primary cause of hours, as I explain in a moment.
delay in this subsegment seemed to be The goal of increased seat occupancy
related to personnel. We decided to focus could be achieved by attracting more cus-
on raising employees awareness of the tomers, providing a better table mix so
importance of efficient check processing, more customers could be accommodated,
improving training on how to process a and reducing the dining duration so more
check, and emphasizing the need to pick customers could be served. The restaurant
up the completed check folder in a timely already had excess demand on weekend
fashion. In addition, the Chevys chief nights (as indicated by the waiting lines).
information officer investigated possible More worrisome, however, the restaurant
technological improvement to the authori- operates in a competitive location where
zation equipment. customers put their names on the wait list
at several restaurants and patronize the
Bussing. We determined that the pri- first one that can make a table available.
mary reason that the mean bussing time Because the restaurant’s current table mix
was relatively long was the lack of and dining duration would not allow the
prebussing (another personnel-related restaurant to serve additional customers,
cause). We decided to emphasize to both the manager’s focus was on improving the
servers and bussers the need to clear dirty table mix and reducing dining duration.
dishes from tables throughout the meal.
That would leave less work when it came The 5 percent solution. To assess the
time to clear the table. In addition, there revenue effects of increased occupancy
was inadequate space in the kitchen to and decreased dining duration, we first
stack used dishes, so we recommended calculated the annual revenue for the hot
that a dedicated dish scraper and stacker periods. To review, during the ten hot
be on duty during busy periods. hours each week, the main dining room
had an average seat occupancy of 50 per-
Step 3: Developing a cent, an average check of $12.73, and an
Revenue-management Strategy average dining time of fifty-three minutes.
Annual sales for the restaurant in 2001
We first identified the busy (hot) and slow totaled $2,358,874. The restaurant took in
(cold) periods by day of week and hour of $861,797 per year (or about one-third of
day. Hot periods were defined as times its annual revenue) during its ten hot
when guests were waiting to be seated, hours. If hot seat occupancy increased
and the remaining periods were cold. The from 50 percent to 60 percent, even if din-
60 Cornell Hotel and Restaurant Administration Quarterly FEBRUARY 2004
RESTAURANT REVENUE MANAGEMENT REVENUE
ing duration remained the same, annual a restaurant designer whose task was to
revenue would potentially increase by achieve or approach the optimal table
$172,359, or 7.3 percent, as shown in the mix—provided that the restaurant ended
calculation in Exhibit 7. Beyond that, if up with the same number of seats as
dining duration could be decreased from before and the design was attractive and
fifty-three minutes to forty-eight minutes, functional. Because the restaurant had
even if seat occupancy remained the same, been open for almost ten years without
the annual revenue potential would being remodeled, the corporate managers
increase by $89,771, or 3.8 percent. If decided not only to reconfigure the
both factors could be changed (i.e., seat tables but also to repaint the restaurant, to
occupancy increased and dining duration relocate the host and service stations,
decreased at the same time), the annual and to provide several other aesthetic
revenue potential would increase by improvements.
$280,084, or 11.9 percent. Even if only The restaurant’s managers wanted an
half of the revenue potential could be attractive design that allowed ample space
achieved, the restaurant could neverthe- for movement but did not make customers
less achieve better than a 5 percent feel crowded. Despite the restaurants foot-
increase in annual revenue. print and structural limitations (support
pillars could not be moved and the host
Step 4: Implementation stand had to stay near electrical connec-
tions), the designer fulfilled her mandate
Once the strategy was developed, the hard of creating a plan that retained the original
work of implementation began. In keep- number of seats and creating the optimal
ing with the strategy, implementa- table mix. To achieve those goals, she used
tion focused on improved table mix and a variety of design approaches, such as
on improving the efficiency of service half walls, banquettes, high-top tables,
delivery. and well-located booths.
The management team, the corporate
Table mix. An optimal table mix, one team, and the designer then reevaluated
that matches party-size mix as closely as the initial design plans and modified that
possible, would allow this restaurant to design to accommodate both the build-
serve an increased number of customers ing’s site-specific issues and certain cus-
with no increase in the number of seats, tomer characteristics (e.g., families with
thereby boosting seat occupancy during babies needed ample space for their stroll-
busy periods. We used the table-mix simu- ers and carriages). With those additional
lator developed by Gary Thompson, of considerations, the managers decided to
Cornell University, to develop the optimal adopt one of the near-optimal table mixes
table mix for the restaurant and also to (i.e., thirty-nine 2-tops, thirty-five 4-tops,
generate a set of near optimal mixes (near and two 6-tops). The simulation results
optimal was defined as within 1.5 percent showed that this table mix would provide
of optimal).12 results within 1.3 percent of the optimal
The optimal table mix maintained the mix’s expected revenue.13
same number of seats but changed the The restaurant continued to operate
table mix from fifty-three 4-tops and three during the reconstruction, with the bulk of
6-tops to forty deuces, twenty-four 4-tops, work occurring during the hours when the
five 6-tops, and three 8-tops. Chevys hired restaurant was closed. To facilitate that
FEBRUARY 2004 Cornell Hotel and Restaurant Administration Quarterly 61
REVENUE RESTAURANT REVENUE MANAGEMENT
strategy, half walls and banquettes were The new rules reconfigured server stations
fabricated off-site so that they could be to comprise four or five tables. The cost of
quickly installed. the additional servers was $127.80 per
week ($2.13 per hour for the four three-
hour shifts).
The additional tables also required
Restaurants can apply the principles additional tablecloths, plates, glasses,
of revenue management to exert sub- chip baskets, salsa dishes, Texas holders,and cutlery. The entire cost of the extra
tle control over the duration of cus- supplies (not including tablecloths) was
$1,395.
tomers’ visits. Here is an example of
Improving service. As I indicated
how that might happen. above, strategies for improving service fo-
cused on the end of the meal—specifically,
the payment and bussing processes.
Changes involved improved training,
The restaurant’s reconfiguration opened more diligent management, and additional
other issues, notably in the back of the staff.
house. Expanding from fifty-six tables to
seventy-six tables would increase the load Tightening the meal time. The time
on the kitchen (chiefly because seat occu- between the entrée delivery and the check
pancy would be improved), increase the delivery averaged about twenty-three
labor requirements (e.g., cooks and serv- minutes in the baseline study. The team
ers), and (less critically) require the pur- did not wish to rush guests’ meals, but
chase of additional table supplies. managers believed that better prebussing
Although reducing customers wait time would cause some guests to leave a bit
was a desirable goal, the kitchen might more quickly and would almost certainly
suffer from the reduction in the buffer time reduce the time needed for final bussing.
that came during table bussing and reseat- For this reason, both servers and bussers
ing. Moreover, increased seat occupancy were trained to do a better job of pre-
would require the preparation of more bussing. In addition, the restaurant’s man-
meals. To ease the kitchen’s workload, the agers augmented the back of the house
restaurant’s managers decided to add an during the hot times by hiring a stacker,
expeditor who worked the middle line who was responsible for scraping all
during hot periods (four shifts per week). dishes and stacking them for the dish-
This person facilitated communication washers. The pay was $5.15 per hour for
between the broiler line and the enchilada the four hot shifts of three hours each
line and helped speed delivery. Two peo- (weekly cost of $61.80).
ple were trained for this position at an
hourly rate of $9 for four shifts of three Check settlement. The baseline study
hours each (a weekly cost of $108). found that it took an average of nearly six
The additional twenty tables and new minutes from the time the guest requested
corporate rules on station size required an the check to the time the check was settled
additional five servers during busy periods and change (or a receipt) was returned.
(again, four shifts of three hours each). Rather than wait for the guest to ask for the
62 Cornell Hotel and Restaurant Administration Quarterly FEBRUARY 2004
RESTAURANT REVENUE MANAGEMENT REVENUE
Exhibit 7:
Annual Incremental Revenue Potential
Seat Occupancy
Dining Duration (minutes) 50% 52% 54% 56% 58% 60%
53 $0 $34,472 $68,944 $103,416 $137,888 $172,359
52 $16,573 $51,708 $86,843 $121,977 $157,112 $192,247
51 $33,796 $69,620 $105,443 $141,267 $177,091 $212,915
50 $51,708 $88,248 $124,788 $161,328 $197,869 $234,409
49 $70,351 $107,637 $144,923 $182,209 $219,494 $256,780
48 $89,771 $127,833 $165,896 $203,959 $242,021 $280,084
47 $110,017 $148,889 $187,762 $226,634 $265,507 $304,379
check, servers were trained either to drop by servers and the small space available in
the check upon clearing the entrée dishes the kitchen. Our team anticipated that
or, if that was inopportune, to be alert for improved prebussing and the addition of
signs that the guests would like to leave. In the stacker position would help reduce this
addition, the corporate office kept its com- time.
mitment to study information-technology Construction and training were com-
solutions to the problem of slow credit- pleted by mid-March 2002. As expected,
card authorization. the first few weeks of revamped operation
were hectic as management and staff
Departure. Guests in the baseline members adjusted to the new layout and
study lingered an average of another four increased customer volume. When I
and one-half minutes from the time the returned in May 2002 to evaluate the
change or credit-card slip was returned reconfigured restaurant, it seemed that the
until their actual departure. As I men- restaurant’s revenue performance had
tioned, the managers determined that improved over the baseline. To confirm
guests were (understandably) loath to that notion, I reexamined the subsegments
leave active, signed credit-card slips on of the service process and recalculated
the table. To remedy that situation (and to revenues.
signal to guests that the transaction was,
indeed, complete), servers were trained to Step 5: Evaluation
pick up completed check folders as soon
as customers were done with them. In Updated POS and new time-study data
addition to the prompt conclusion of the were collected so that seat occupancy,
financial transaction, the team hoped that RevPASH, and meal duration could be
enhanced prebussing would signal the recalculated. A financial analysis was per-
customer that it was time to leave. formed to assess the effects of the changes
three months after implementation.
Final bussing. Clearing and reseating Seat occupancy during the hot periods
the table took an average of nearly three increased from 50 percent (with a peak of
minutes in the baseline study. As men- 61 percent) in the baseline study to a mean
tioned above, the major contributor to this of 59 percent, with a peak of 82 percent (as
lengthy process was the lack of prebussing shown in Exhibit 8). RevPASH numbers
FEBRUARY 2004 Cornell Hotel and Restaurant Administration Quarterly 63
REVENUE RESTAURANT REVENUE MANAGEMENT
Exhibit 8:
Posttest Seat Occupancy by Day of Week and Hour of Day
Hour of Day
Day of Week 11:00 noon 1:00 2:00 3:00 4:00 5:00 6:00 7:00 8:00 9:00
Sunday 25% 43% 51% 32% 23% 30% 26% 36% 18% 4% 2%
Monday 13% 27% 14% 8% 8% 12% 42% 40% 36% 13% 4%
Tuesday 12% 23% 14% 5% 8% 13% 25% 42% 30% 4% 3%
Wednesday 18% 24% 18% 7% 8% 14% 28% 39% 31% 9% 4%
Thursday 15% 25% 15% 11% 8% 19% 22% 35% 34% 9% 14%
Friday 21% 31% 25% 11% 10% 21% 49% 82% 69% 43% 9%
Saturday 26% 46% 46% 30% 25% 31% 57% 71% 61% 25% 2%
Note: Numbers are based on eight weeks of point-of-sale data from spring 2001.
showed a similar increase. Average able to reduce the standard deviation by
RevPASH for the hot periods increased over 60 percent. This can be attributed totraining and the enhanced awareness of the
from $5.85 to $6.32. servers of the importance of this process.
Additional time studies conducted dur- 4. Final settlement to departure. The restau-
ing peak dinner times (82 observations) rant did not have complete control of thistime, since customers can choose when they
showed that the service-delivery improve- want to leave. The time decreased slightly,
ments had largely achieved their intended as did the standard deviation, which may
goal, particularly at the end of the meal. indicate that the emphasis on picking up the
check folder had worked.
The mean total meal time dropped from 5. Order to entrée delivery. The goal here was
fifty-three minutes, fifteen seconds to fifty mostly to avoid deterioration in the face of
minutes, fifty-six seconds, and the stan- greater seat occupancy. Indeed, the time for
dard deviation of total meal duration this step did not decrease substantially, butthe impressive fact was that the kitchen was
dropped from twenty-two minutes, forty- able to maintain its pace for preparing and
six seconds to fifteen minutes, nine sec- delivering entrées even with the increased
onds. The specific goal of a five-minute number of customers from the new tablemix. The restaurant’s managers attributed
drop in dining duration was not achieved, this to the strong kitchen staff members who
but the variation was decreased substan- were willing to experiment with various
tially (see Exhibit 9). approaches to handling the increased
demand.
The mean duration of the steps on
which managers had focused all re- Again, the focus of managers inter-
sponded to the interventions, as follows: ventions was improving the end of the
1. Seat to greet. The restaurant reduced this meal, and it showed. In fact, the duration
time by forty-two seconds and trimmed the of two subsegments at the beginning of
standard deviation by 25 percent. This could the meal (greet to drinks, drinks to
be attributed to training and improved order) increased, thus representing areas
awareness.
2. Entrée served to check drop. The total time for future improvement. Also, the res-
diminished by about a minute, but the stan- taurant had been doing an excellent job
dard deviation went down by over 40 per- of reseating tables once they were
cent. This can be attributed to the improved
prebussing. cleared. This time increased by ten sec-
3. Check drop to final settlement (change or onds after the renovations and interven-
receipt returned). The restaurant was able to tions were instituted.
reduce this time by nearly one-half and was
64 Cornell Hotel and Restaurant Administration Quarterly FEBRUARY 2004
RESTAURANT REVENUE MANAGEMENT REVENUE
Exhibit 9:
Pretest and Posttest Time-Study Results
Pretest Posttesta
M SD M SD
Seat to greet 2:20 2:01 1:38 1:30
Greet to drinks 3:52 4:24 4:39 5:49
Drinks to order 2:34 3:23 3:14 4:18
Order to entrée 11:31 5:06 11:28 4:06
Entrée to check dropped 22:41 12:19 21:47 6:44
Check dropped to change returned 5:40 4:54 2:55 1:56
Change returned to departure 4:27 6:38 4:09 5:09
Departure to bussed 2:48 2:36 2:00 1:54
Bussed to reseated 0:56 1:04 1:06 1:02
Total dining duration 53:15 22:46 50:56 15:09
a. Times in bold were shorter during the posttest than they were in the pretest.
Financial Analysis Arrowhead store had experienced a 2.0
While the drop in service times was heart- percent increase in revenue from 2001 to
ening, Chevys managers wanted to ascer- 2002, while the 2002 sales for the control
tain whether the restaurant had actually stores were 8.0 percent less than the same
met its key goal of a 5 percent increase in period sales in 2001. The Arrowhead store
revenue. Since RevPASH had increased, had realized a 7.7 percentage-point
store revenue had increased, but it was increase in revenue from pretest to post-
unclear whether that increase resulted test, while the control stores had increased
from the managers’ interventions or from sales by 2.6 percentage points. The 5.1
other market conditions. To control for percentage-point difference was attrib-
outside factors, the financial performance uted to management interventions.
for the Arrowhead restaurant and for two
other Chevys restaurants in the Phoenix At what cost? The new table mix and
market was analyzed. service changes were not without cost. In
To begin with, the effects of the round figures, the remodeling cost
national recession seemed to show up in $49,000, additional small wares cost
the restaurant’s finances. During the approximately $1,400, and labor costs
seven weeks before revenue-management increased by about $15,000 per year.
14
implementation, the Arrowhead store had The Phoenix Arrowhead restaurant had
experienced a 5.7 percent drop in year-to- approximately $2.4 million in sales in
year revenue (i.e., from the similar period 2001. Based on the projected increase of
2001 to the seven weeks before construc- 5.1 percentage points calculated above,
tion began in 2002), while the comparable there was an estimated annual sales
restaurants had seen combined revenues increase of approximately $120,000 over
fall by 10.6 percent. the amount that general market conditions
A similar seven-week analysis was might have afforded had the restaurant’s
conducted after the new table mix and management done nothing. Given Chevys
other measures were implemented. The 45.5 percent EBITDA flow through,
FEBRUARY 2004 Cornell Hotel and Restaurant Administration Quarterly 65
REVENUE RESTAURANT REVENUE MANAGEMENT
approximately $54,600 of this amount formance goals, and determine feasible
would go to the bottom line. The total cap- ways of meeting these goals. In addition,
ital cost of the project was around be sure to assess the financial effects of
$50,400, so the cash-on-cash return was any given strategy. Implementation of
about 108 percent, with a payback of less the strategies is probably the most diffi-
than one year (11.1 months). The remod- cult, but most important, part of the
eling expenses were a one-time expense, process. Implementation required a
so the return will be even greater in the strong management team, good training,
future. and a willingness to try new things (such
as changing the table mix or hiring more
Making the Chevys Arrowhead employees).Finally, when you have made all of the
Experience Work for You changes, reevaluate your performance
By analyzing its service process and table after about two months. Gather additional
mix, Chevys Arrowhead was able to POS data and conduct additional time
increase revenue by approximately 5 per- studies to see whether your efforts have
centage points more than the two other paid off.
Chevys that we examined. This perfor- For more information on developing a
mance boost came from its improved revenue-management program, you can
table mix, changes in the service delivery, read the articles mentioned in this article,
and improved training. Seat occupancy you can take the courses in restaurant rev-
and RevPASH increased, dining dura- enue management available online from
tion and variation in that duration e-Cornell (www.ecornell.com), or enroll
decreased, and revenue and profitability in a restaurant-revenue-management
increased. course offered through the Cornell hotel
If you want to improve your restau- school’s Professional Development Pro-
rant’s revenue in this way, you should gram (PDP).
first establish your restaurant’s baseline Other restaurants could realize similar
performance. Collect at least a month of results by carefully analyzing their current
detailed POS data and analyze your seat performance, determining the causes of
occupancy, average check, RevPASH, that performance, and developing appro-
party mix, and dining duration by day priate strategies to improve performance.
of week and hour of day. In addition, Changes in table mix and problematic ser-
hire someone (this is an ideal part-time vice delivery hold particular promise but
job for a student) to conduct time stud- only when such changes are combined
ies of your restaurant during your busy with implementation that emphasizes
periods. training, employee buy-in, and enhanced
After you have developed the baseline management.
performance, sit down with your manage-
ment team and staff members to make Endnotes
sense of what you have discovered. Dis-
cuss what might be driving your perfor- 1. See: B. A. Smith, J. F. Leimkuhler, and R. M.
mance and pinpoint specific areas in need Darrow, Yield Management at American Air-
of improvement. lines, Interfaces, vol. 22, no. 1 (1992), pp. 8-
When developing a strategy, focus on 31; and R. B. Hanks, R. P. Noland, and R. G.
Cross, Discounting in the Hotel Industry, A
your busy periods, establish specific per- New Approach, Cornell Hotel and Restaurant
66 Cornell Hotel and Restaurant Administration Quarterly FEBRUARY 2004
RESTAURANT REVENUE MANAGEMENT REVENUE
Administration Quarterly, vol. 33, no. 3 (June 9. For a thorough discussion of RevPASH, see:
1992), pp. 40-45. Ibid.; and Kimes, Barrash, and Alexander, pp.
2. See: Smith et al., pp. 8-31; Hanks et al., pp. 40- 18-30. Hourly revenue is defined as revenue
45; and W. J. Carroll and R. C. Grimes, Evolu- from all checks that are opened during that
tionary Change in Product Management: Ex- hour.
periences in the Car-rental Industry, Inter- 10. Fishbone analysis is a quality-management
faces, vol. 25, no. 5 (1995), pp. 84-104. technique that can be used to help determine the
3. See: S. E. Kimes, R. B. Chase, S. Choi, E. N. possible causes of a problem. For a discussion
Ngonzi, and P. Y. Lee, Restaurant Revenue of fishbone analysis, see: D. D. Wyckoff, New
Management, Cornell Hotel and Restaurant Tools for Achieving Service Quality, Cornell
Administration Quarterly, vol. 40, no. 3 (June Hotel and Restaurant Administration Quar-
1998), pp. 40-45; S. E. Kimes, Implementing terly, vol. 42, no. 4 (August 2001), pp. 25-38.
Restaurant Revenue Management: A Five-step 11. Those times were Friday, 5:00 to 9:00; Satur-
Approach, Cornell Hotel and Restaurant Ad- day, 12:00 to 2:00 and 5:00 to 8:00; and
ministration Quarterly, vol. 40, no. 3 (June Sunday, 1:00 to 2:00.
1999), pp. 16-21; S. E. Kimes, D. I. Barrash, 12. See: G. M. Thompson, Optimizing Restau-
and J. E. Alexander, Developing a Restaurant rant-table Configurations: Specifying Com-
Revenue-management Strategy, Cornell Hotel binable Tables, Cornell Hotel and Restaurant
and Restaurant Administration Quarterly, vol. Administration Quarterly, vol. 44, no. 1 (Feb-
34, no. 5 (October 1999), pp. 18-30; S. E. ruary 2003), pp. 53-60; G. M. Thompson, Op-
Kimes, Revenue Management on the Links: timizing a Restaurants Seating Capacity: Use
Applying Yield Management to the Golf In- Dedicated or Combinable Tables? Cornell Ho-
dustry, Cornell Hotel and Restaurant Adminis- tel and Restaurant Administration Quarterly,
tration Quarterly, vol. 41, no. 1 (February vol. 43, no. 3 (June 2002), pp. 48-57; S. E.
2000), pp. 120-27. Kimes and G. M. Thompson, Restaurant Reve-
4. See: B. Sill, Capacity Management: Making nue Management at Chevys: Determining the
Your Service-delivery System More Produc- Best Table Mix, working paper no. 07-05-02,
tive, Cornell Hotel and Restaurant Adminis- Cornell University School of Hotel Adminis-
tration Quarterly, vol. 33, no. 1 (February tration (2002); and S. E. Kimes and G. M.
1991), pp. 77-87; B. Sill and R. Decker, Apply- Thompson, A Comparison of Techniques for
ing Capacity-management Science: The Case Identifying Optimal and Near-Optimal Res-
of Browns Restaurants, Cornell Hotel and taurant Table Mixes, working paper no. 09-04-
Restaurant Administration Quarterly, vol. 40, 02, Cornell University School of Hotel Ad-
no. 3 (1999), pp. 22-30; and the above-cited ar- ministration (2002).
ticles by Kimes et al. 13. Kimes and Thompson, no. 07-05-02.
5. See: S. E. Kimes and R. B. Chase, The Strate- 14. Ongoing costs included $61.80 per week for an
gic Levers of Yield Management, Journal of additional busser ($5.15 per hour for four
Service Research, vol. 1, no. 2 (1998), pp. 156- shifts of three hours each), $108.00 per week
66; Kimes et al. (1998), pp. 40-45; and S. E. for a kitchen stacker ($9.00 per hour during the
Kimes, J. Wirtz, and B. M. Noone, How Long same four three-
Should Dinner Take? Measuring Expected hour shifts), and
Meal Duration for Restaurant Revenue Man- $127.80 per week
agement, Journal of Revenue and Pricing for five additional
Management, vol. 1, no. 3 (February 2003), pp. servers (at $2.13
220-33. per hour for the
6. S. E. Kimes and J. Wirtz, Perceived Fairness of four shifts). The
Demand-based Pricing for Restaurants, Cor- total additional la-
nell Hotel and Restaurant Administration bor cos t was
Quarterly, vol. 43, no. 1 (February 2002), pp. $297.60 per week.
31-38.
7. See: Kimes and Chase, pp. 156-66; and Kimes
et al. (October 1999), pp. 18-30. Sheryl E. Kimes is a professor and director of
8. For a thorough discussion of the five-step pro- graduate studies at the Cornell University School
cess, see: Kimes, pp. 16-21. of Hotel Administration.
FEBRUARY 2004 Cornell Hotel and Restaurant Administration Quarterly 67