Professional Documents
Culture Documents
bbd4 PDF
bbd4 PDF
consequence of a long process of trial and every restaurant. In 1988, Taco Bell in-
error, based occasionally on adapting what vested in an in-store computer network,
other quick-service companies were doing which allowed it to develop more
to the Taco Bell environment. advanced-information solutions. The result
Taco Bell labor costs represent approxi- was the field-automated-staffing program,
mately 30 percent of every sales dollar and FAST. FAST, like the traditional approach,
are among the largest controllable costs. was a top-down control system that allo-
They are also among the most difficult to cated a percentage of sales to labor costs
manage because of the direct link that ex- to meet prescribed financial targets. It still
ists between sales capacity and labor. Be- told managers what to achieve, but not
cause the product must be fresh when how to do it, and managers still had to
sold, it is not possible to produce large rely on experience to do it well. If the
amounts during periods of low demand to amount of labor allocated to stores ap-
be warehoused and sold during periods of pears insufficient to meet customer service
high demand, as is typical in most manu-
facturing enterprises. Instead, the product Labor costs represent
must be prepared virtually when it is or- approximately 30 percent of
dered. And since demand is highly vari- every sales dollar.
able and is concentrated during the meal
periods (52 percent of daily sales occur goals and maintain sales, intense debates
during the three-hour lunch period from occur between financial executives and
11:00 am to 2:00 pm), determining how operations executives. Financial executives
many employees should be scheduled to typically resist increasing the labor-cost al-
perform what functions in the store at any location, and operations executives typi-
given time is a complex and vexing cally argue that sales are being lost be-
problem. cause they cannot assign enough labor to
Because of the complexity of the labor- the restaurants to get everything done.
management problem described above, the In 1988, Taco Bell management decided
traditional approach to managing labor to move the company to a new plateau in
costs is to establish a policy decision that terms of market share, per-unit sales, and
indicates what percentage of sales the operating profits. To do this it developed
company is willing to allocate to labor cost and implemented aggressive new strate-
in order to meet prescribed financial objec- gies and tactics. One of the earliest and
tives. It is then up to those who manage most visible of the new strategies was
store operations to develop rules and pro- “value meals.” Subsequent strategies af-
cedures that seek to maximize the level of fected virtually all aspects of Taco Bell, in-
service and quality that can be delivered cluding new store design and layout,
with the given amount of labor assigned products, equipment, and customer ser-
to the store. Over time, these rules and vice with and without self-service bever-
procedures evolved into a body of “tribal ages. However, before management could
knowledge” that was applied manually at make any decisions to implement change,
INTERFACES 28:1 76
TACO BELL
it had to know what impact this change porate headquarters. This system, together
would have on labor cost and customer with a number of software modules to au-
service. The fast-service food industry had tomate management paperwork, collect
long ago learned the lesson that America and transmit transactional data to the
Online recently endured: Increasing de- mainframes for financial control, and con-
mand without anticipating its effects on trol aspects of store operations, was
operations can make for a very unpleasant dubbed TACO (total automation of com-
experience for all. pany operations). By replacing FAST with
During the planning phase that pre- a new, comprehensive, integrated labor-
ceded the adoption of the new “value management system as an integral part of
strategy” (value meals and associated TACO, the company could take advantage
changes to the system’s stores), Taco Bell’s of and accelerate the deployment of the
managers realized they needed answers to new hardware and software to the field.
questions they had never before asked. Furthermore, with the new labor manage-
They needed to know what impact opera- ment system as a standard component of a
tional decisions would have on labor costs comprehensive restaurant-management
and speed of service. The existing labor system, Taco Bell and its franchisees and
system, FAST, was not designed to answer licensees could manage labor in a consis-
such questions. It was designed to re- tent manner across the entire system. The
spond to questions whose answers could results would be reflected in greater con-
be inferred from the past. It could not be sistency of service and, hence, greater cus-
adapted to yield the information needed tomer satisfaction and brand equity. It
to support the many decisions that had to could also serve as one more tool to make
be made in the new environment. Also, as investing in a Taco Bell store more attrac-
the company institutionalized new poli- tive to prospective franchisees.
cies, FAST would become inadequate be- Project Genesis
cause the experience base upon which In Taco Bell’s search for new ways to
operations management had been success- answer the many questions that accompa-
ful in the past would no longer reflect the nied proposed operational decisions, one
company. As a result, Taco Bell began to individual emerged as a champion for the
look for new ways to answer the many use of operations research as a means to
questions that accompanied each of the support decision making in the new envi-
operational decisions that managers had ronment. This individual, Howard Frantz,
to make to ensure that the value strategy manager of Taco Bell’s newly established
would be successful. industrial engineering function, convinced
Once it understood and accepted the management to fund the development of a
need for better and more timely informa- new labor-management system (LMS). The
tion, management made a commitment to study was to be conducted under his lead-
introduce IBM-compatible personal com- ership assisted by a newly formed labor
puters in every Taco Bell store linked to a steering committee. This steering commit-
centralized mainframe system at the cor- tee consisted of Taco Bell’s six zone vice
January–February 1998 77
HUETER, SWART
presidents and represented all the com- —Provide timely feedback for controlling
pany’s operations management. This com- labor cost at all levels; and
mittee would (1) provide input and guid- —Have the inherent flexibility to evolve as
ance regarding labor issues, (2) provide Taco Bell evolves.
access to internal resources, and (3) review After appropriate deliberations among
and evaluate results. The first directive of all stakeholders in the study and a search
the committee was that Howard Frantz of the literature, we confirmed our earlier
seek the advice and counsel of the leading conclusion that these capabilities could be
authority on the use of operations research provided only by a model-based labor-
in the food service industry. This led to the management system.
hiring of Dr. William Swart as a strategic The Restaurant as a System
consultant for the development of the new As a first step, we developed a concep-
LMS. Dr. Swart had pioneered the use of tual system representation of a quick-
operations research in the industry and re- service restaurant (Figure 1). The system is
ceived an Edelman finalist award in 1981 made up of three interacting segments: the
for his accomplishments. To complete the customer, the delivery, and the production
LMS development team, Howard Frantz subsystems. We do not consider a fourth
hired one full-time industrial engineer, en- segment, the consumption subsystem (or
listed a job shopper, contracted for indus- dining area) because we focused primarily
trial engineering consulting services, and on the other three.
enlisted the services of data-entry and pro- The restaurant system and more tradi-
gramming personnel. Howard Frantz and tional manufacturing concerns are similar
William Swart held the project’s kick-off in many ways:
meeting with the steering committee on —Each product on the menu requires a
January 11, 1991. well-defined sequence of assembly and
In its deliberations with the LMS devel- processing steps, including packaging.
opment team, the steering committee iden- —Multiple products compete for access to
tified the following key objectives of Taco scarce resources (for example, fryers).
Bell’s new labor-management system: —The product line is a mix of standard
—Be responsive and economical; products (for example, tacos) and custom-
—Be able to predict labor requirements ized products (for example, “hold the hot
that minimize labor cost and meet all cor- sauce”).
porate standards for hospitality, quality, —Raw material, in process, and finished
service, and cleanliness for any existing or product inventories must be maintained
planned restaurant configuration; and managed.
—Serve as an effective and efficient in- —Distribution occurs through multiple
store labor-management tool to help the channels (walk-in, drive-through, home
store manager to plan and schedule; delivery).
—Serve as an effective and efficient tool —An order consists of a variety of prod-
for providing the labor required to achieve ucts that have to be assembled and pack-
Taco Bell’s financial targets; aged together.
INTERFACES 28:1 78
TACO BELL
—Employee productivity and training are —Ordering, production, and delivery take
highly correlated. place under the same roof.
However, they also show some very sig- —Employee and management turnover is
nificant differences: high.
—The acceptable lead time between plac- Model Conceptualization
ing an order and order delivery (speed of The total labor requirements in a restau-
service) is only a few minutes (customers rant can be categorized into three types:
cannot wait a long time, in particular dur- fixed labor, variable food-preparation la-
ing lunch time in the work week, and they bor, and variable customer-service labor.
expect fast service). Fixed labor is that which is required and is
—Finished product in inventory perishes largely independent of the restaurant’s
in minutes (customers all like their food daily sales volume. Such labor includes,
freshly prepared). for example, that which is required for
—Demand can easily fluctuate by 1,000 cleaning between the closing of the restau-
percent in a 30-minute interval (for exam- rant at night and its opening the next
ple, between 11:30 am and noon in busi- morning. Variable food-preparation labor
ness areas, stores can go from empty to is that which is required to meet the res-
crowded). taurant’s daily operating needs but is not
January–February 1998 79
HUETER, SWART
INTERFACES 28:1 80
TACO BELL
we [Swart and Donno 1981] have followed utes would be a significant improvement.
and defined the relationship in terms of the Furthermore, they saw the resulting low
number of customers that are likely to balk risk of balking as acceptable. We agreed to
(leave the line and the store) as a function review the risks and consequences again
of speed of service. A study we performed later.
assessed when customers would be likely Addressing Store Management
to leave a waiting line because they per- Requirements
ceived waiting time as excessive. The re- To satisfy one of the key objectives, the
sults indicated that customers in line per- new labor-management system had to
ceive that waiting time is “a couple of serve as an efficient and effective in-store
minutes” although actual waiting time can labor-management tool. It had to help the
reach five minutes. After actual waiting manager plan shifts. That is, it had to de-
time exceeds five minutes, customer per- termine who should begin work when and
ception of waiting time increases exponen- for how long each day in such a manner
tially (Figure 2). The higher the perceived that payroll is minimized. An integral part
waiting time is, the more likely a customer of that decision is planning when employ-
is to “balk.” Further studies quantified the ees should start and finish their allocated
variability of the waiting-time distribution. shifts. To avoid the confusion of having
Using this probabilistic information, we de- employees begin or end their shifts at any
termined that a three-minute average time time, it is common to start and end shifts
in the queue (from arrival at the restaurant at discrete points in time, such as on the
until food is delivered) ensures that only hour, on the half hour, or on the quarter
2.5 percent of customers that wait are likely hour. The selection of start and end points
to balk. We discussed this finding with se- is critical. For example, if they can start
nior managers and they decided that, at only on the hour, then during lunch em-
this stage of the process, achieving a ser- ployees could either start at 11:00 am and
vice level of consistently delivering food to have little to do until noon or they could
customers within an average of three min- start at noon amidst the confusion of the
busiest time of the day. Consequently, it is
desirable for employees to begin and end
their shifts in the smallest practical time
increments (usually at 15-minute intervals,
but no longer than 30 minutes) so that la-
bor scheduled in the store can most closely
adhere to predicted demand. Scheduled
labor and labor demanded are not usually
the same because employees, once brought
to the store to work, must be guaranteed a
minimum shift, the length of which de-
Figure 2: The relationship of perceived versus
actual time spent in line by quick-service es- pends on state law and local labor
tablishment customers. conditions.
January–February 1998 81
HUETER, SWART
To minimize payroll and deliver desired ber of independent time series, each repre-
customer service, store managers require a senting the customer-transaction history
tool that will help them to determine corresponding to a specific 15-minute in-
which employee should begin work and terval during a specific day of the week.
which employee should go home at every For example, the customer-transaction his-
time interval of the day for every day of tory at a particular store for all Fridays
the week. At the beginning of this study, from 9:00 to 9:15 am constituted the time
this interval was 30 minutes because of re- series to be used to forecast customer
strictions imposed by the in-store data- transactions at that store for future Fridays
storage devices. Today, the company is from 9:00 to 9:15 am.
overcoming this limitation and imple- To produce the data on which to base a
menting a 15-minute time increment sys- weekly staffing schedule, the forecasting
temwide. In addition to the number of em- procedure must forecast next week’s cus-
ployees required, such a tool must indicate tomer transactions for each 15-minute in-
where they must be positioned and how terval of every day of the week that the
they are to share the required duties. And restaurant is open. It does this by forecast-
before it can determine that, it must know ing the transaction time series associated
how many customer transactions will take with each 15-minute interval of the restau-
place at the store during every 15-minute rant’s daily open time during the week
interval of every day of the week. one period ahead. In an unpublished
This means that the new labor manage- study conducted by one of us (Swart), the
ment system needed three major compo- complete array of forecasting methods in-
nents: (1) a forecasting model for predict- cluded in Makridakis and Wheelwright’s
ing customer transactions; (2) a simulation [1978] Sybill/Runner forecasting system
model to translate customer transactions was used to predict one-period-ahead
to labor requirements; and (3) an integer sales (sales and transactions are directly
programming model to schedule employ- correlated) for 30-minute interval data for
ees to satisfy labor requirements and mini- all company-owned Burger King restau-
mize payroll. These models must operate rants. The forecasting procedure that mini-
in concert while permitting the manager to mized the average square deviation be-
intervene and make appropriate revisions tween actual and predicted data was the
before implementing the results (Figure 3). naı̈ve method (the projection is the same
The Forecasting Model as the last data point). With the under-
Customer transactions during a 15-min- standing provided by that experience, we
ute interval are subject to many sources of set forth to determine the best forecasting
variability, including but not limited to procedure to use for Taco Bell.
time of the day, day of the week, week of Although some argue that there is some
the month, and month of the year. To evidence that trend and seasonal patterns
eliminate as many sources of variability as might exist in Taco Bell data, we could not
possible, we separated all customer- extract a reliable database of multi-year,
transaction data for all stores into a num- 15-minute-interval customer transactions
INTERFACES 28:1 82
TACO BELL
by day from the corporate databases be- justed for quadratic trend. Because the
cause the corporation had not collected time series consisted of only six data
such data in the past. In cooperation with points, we did not consider using expo-
Taco Bell’s information technology group, nential smoothing. The results indicated
we identified 200 restaurants that had re- that a six-week moving average was best.
cently upgraded their telephone lines to Taco Bell incorporated this forecasting
take part in an effort to collect 15-minute- methodology in each store’s computer sys-
interval transaction data. We developed tem. The computations are transparent to
and implemented a rolling database con- the restaurant manager and do not tax in-
taining six weeks of in-store and drive- store computer resources. The model com-
through transaction data, the maximum pares weekly activity projections obtained
that restaurant in-store computers could from the forecasting model to statistical
store. We then conducted tests to deter- forecasting control limits that are continu-
mine which applicable forecasting meth- ously updated. The model also reoptim-
ods would yield the least mean square de- izes the length of the moving-average-
viation. The methods tested were (1) period when it declares the forecasts out
moving averages (from one through six of control, but this is also transparent to
periods); (2) moving averages adjusted for the restaurant manager. The system pro-
linear trend; and (3) moving averages ad- vides the manager with forecasts obtained
January–February 1998 83
HUETER, SWART
Figure 4: Flow chart depicting the process used to transform a statistical forecast into a final
forecast at Taco Bell.
INTERFACES 28:1 84
TACO BELL
that there was no reason to reject the hy- register to the pickup area, and of the em-
pothesis that they followed an exponential ployee from the preparation area to the
distribution, regardless of the sales volume finished product area, varies. Thus the
of the particular store(s) from which we hours devoted to collecting the data
obtained the data. greatly exceeded those devoted to
In some cases, determining the nature of modeling.
the data to use to populate the model be- Once we had verified and validated the
came the subject of debate. For example, model, we obtained the minimum labor
should one measure the time required to requirements to meet any level of activity
prepare a breakfast taco (or any other by staffing the model with the minimum
product or process) by observation and in- acceptable store complement (a policy de-
clude observed statistical variability, or cision at Taco Bell that takes into account
should one obtain the time via a predeter-
mined time standard method, such as This system was dubbed
MOST [Mishra 1982], which yields deter- TACO (total automation of
ministic results? Taco Bell elected to collect company operations).
data for the variable customer-service la-
bor by taking slow-motion videotapes of safety as well as operational factors). We
trained employees working at a normal then increased simulated store-activity lev-
pace using acceptable procedures in actual els by increments until it became impossi-
restaurants. We then analyzed these vid- ble to keep average customer-waiting time
eos and expressed the resulting standards below three minutes. At that time, we in-
as an empirical frequency distribution to creased the staffing level by one, explored
reflect the stochastic nature of the real the logical assignment of duties, and
world. adopted that assignment of duties that al-
Collecting all the data required to popu- lowed for the maximum increase in store
late the simulation model became the ma- activity while maintaining service stan-
jor task we had in developing the LMS in dards. Thus the model also determined
terms of time and effort. For example, we the optimal deployment for each staffing
had to conduct time studies and data anal- level to maximize throughput and there-
yses for every task that is part of prepar- fore service. We again increased the staff-
ing every product on the menu board. We ing level by one and incrementally in-
had to time every step of the process of in- creased the store-activity level, repeating
teracting with the customer, including the the same process until additional staffing
time to place an order, the time to make could not accommodate the increased ac-
change, and the time a customer took to tivity level while maintaining speed-of-
walk from the cash register to the food- service standards, or until it became physi-
pickup area. We had to repeat this effort to cally impossible to add additional employ-
account for each type of restaurant in the ees (Figure 5).
system, because the travel distance, and We captured the end result of the simu-
hence time, of the customer from the cash lation process in a set of files containing
January–February 1998 85
HUETER, SWART
Figure 5: Flow chart depicting the iterative process used by Taco Bell to develop its staffing
requirements.
the minimum staffing levels and duty as- customer-service labor. We use simulation
signments required to satisfy customer- models to determine the amount of vari-
service standards at each possible activity able customer-service labor that is re-
level. We tailored these files to each res- quired for any level of store activity. We
taurant type and adjusted these to reflect use traditional time-study methods to cal-
the natural variability that exists in menu culate fixed labor requirements and ex-
mix and check averages at each restaurant. press them as labor hours per day per to-
We use simulation to make these adjust- tal store-activity level (typically as a step
ments when warranted. For many situa- function, for example, x labor hours for
tions, however, we have developed adjust- sales to $y, x ` a hours for sales between
ment factors by using various decision $y and $y ` b, and so forth). The store
rules and heuristic processes. These files manager decides when this labor will be
are used to generate data for the integer performed—generally before the restau-
programming models. rant opens or after it closes.
The Integer Programming Model We also use traditional time-study meth-
In a Taco Bell restaurant, the total labor ods to calculate the variable food-
requirements consist of fixed labor, vari- preparation time and also express it as a
able food-preparation labor, and variable step function of store-activity level. How-
INTERFACES 28:1 86
TACO BELL
ever, unlike fixed labor, this labor must be scheduling period and is obtained by put-
performed at various periods during the ting the appropriate wage plus fringes in
day. Ingredients must be restocked on the front of each integer variable.
production line periodically (more often Godward and Swart’s model proved to
during peak hours than during slack be satisfactory, but it did not account for
hours). Based upon projected store-activity the fact that many stores have a limited
level and such factors as the size of stor- number of employees on the payroll. Con-
age bins, the manager will know how sequently, we augmented the model to in-
many times they are likely to need re- clude a constraint on the total number of
stocking during the lunch period. Conse- employees that could be scheduled (for ex-
quently, the manager can predict when ample, the sum of all variables had to be
this type of labor must be performed. less than or equal to the number of em-
In the first phase of developing a model ployees on the payroll and available to
to schedule employees, we calculate the work during that scheduling period). Un-
total labor requirements of a store for each fortunately, this one constraint destroyed
15-minute interval during the day as the the characteristic that allowed us to repre-
sum of fixed labor, variable food- sent the model as a flow network. Conse-
preparation labor, and variable customer- quently, we would have to use more tradi-
service labor. The demand for total labor tional integer-programming algorithms to
at any time interval must be satisfied by schedule employees. But the overhead and
the number of employees in the store dur- time required by traditional solution algo-
ing that interval, regardless of when their rithms made them infeasible to use within
shifts started. Thus, the solution to the the Taco Bell environment. As an alterna-
employee-scheduling problem, as defined tive approach, we decided to use general-
for Taco Bell, follows that described in de- ized Lagrangean multipliers [Everett 1963]
tail by Godward and Swart [1994] and is to eliminate this new constraint from the
an integer program with the decision vari- constraint set. The out-of-kilter algorithm
ables representing the number of employ- provided the computational efficiency to
ees scheduled to begin their shifts at time t solve the reformulated problem efficiently
and the length of shifts s. The value of s is and within the store’s computational
constrained by the minimum and maxi- resources.
mum permissible shift lengths. The con- The solution to the above-formulated
straints ensure that the total number of employee-scheduling problem provides
employees working at time t is equal to or the manager with the set of shifts that will
greater than the number required. This minimize payroll while satisfying all con-
formulation can be transformed into an straints. Although Taco Bell has the capa-
equivalent network formulation and Ford bility to automatically assign names to the
and Fulkerson’s [1962] out of-kilter algo- shifts based on a number of criteria, store
rithm is the basis for the solution algo- managers have preferred to do this manu-
rithm. The objective function to be mini- ally because they can and often do assign
mized represents the total payroll for the desirable shifts as an internal reward or
January–February 1998 87
HUETER, SWART
incentive system for store employees. The compared the labor use obtained by that
manager enters employee names manually process to that reported as actually used in
on the working screen created by the LMS the restaurants. We aggregated the results
after reviewing the model outputs and, if and reported the difference as actual labor
warranted, modifying them to reflect spe- savings. We used this process during the
cial and unanticipated store, market, or five-store and the 10-store operations test
personnel circumstances. The outputs of described above. After companywide im-
the LMS also provide the managers with plementation, we reversed the process. We
information useful in supporting their reported the labor actually used under the
profit-and-loss responsibilities. new LMS system. We then applied the old
Model Implementation FAST system to actual data. We then re-
We developed the individual modules corded the difference in labor used by the
and integrated them into a cohesive labor- simulated application of the old FAST sys-
management system between 1990 and tem and the new LMS system. Since 1993,
1992, implementing them to all company the LMS system has shown a consistent
stores during 1993. Prior to implementing difference of one less hour per day of la-
it systemwide, we prototyped the inte- bor than the FAST system.
grated system and tested it in a sample of Implementation Cost
five restaurants. Once we achieved satis- From January 1991 until the LMS was
factory performance in the five restau- fully implemented, Taco Bell budgeted
rants, we added an additional five restau- $287,500 to support the labor consultant,
rants to the test group. After a five-month the job shopper, the industrial engineering
test period, Taco Bell decided to imple- consultants, and the programmers work-
ment the new LMS in all company restau- ing on the project. Approximately $90,000
rants. At the same time, we developed a was budgeted to support the simulation
measurement and validation process to modeling effort, $80,000 for the data col-
quantify and document the benefits of the lection and development effort, and
new system. $117,500 for the combined tasks of devel-
To measure and validate the LMS, we oping the forecasting model and develop-
collected information on the use of labor ing the integer-programming model. In
from restaurants using the old FAST labor- addition, the industrial-engineering func-
management system (part of the manage- tion grew into the restaurant-systems-
ment routine in restaurants). At the end of engineering department. Today, this de-
each week, members of Taco Bell’s partment has four individuals with train-
operations-services group, assisted by ing in operations research or industrial en-
members of the information-systems- gineering who devote approximately half
technology group and the restaurant- of their time to using the LMS to improve
systems-engineering group, retrieved the the efficiency and effectiveness of the Taco
previous week’s sales data, and applied Bell system.
the new LMS process to those data just as Model Benefits
a restaurant manager would. We then Between 1993 and 1996, the number of
INTERFACES 28:1 88
TACO BELL
Taco Bell company and franchised restau- the store. Before, with the FAST system,
rants using the new LMS system increased the manager was told how many hours of
from 1,298 to 3,954. Labor costs saved labor were allocated to the store. It was
through the use of the LMS system went then his or her responsibility to decide
from $3.51 million in 1993 to $16.4 million how to most effectively deploy this labor.
in 1996 (Table 1). Licensed stores are out- This resulted in variability in the level of
side the scope of the LMS. Taco Bell li- service, hospitality, quality, and cleanliness
censes Taco Bell express stores, which de- among stores. Such variability is one of
liver fast food beyond the traditional “four the most significant sources of customer
walls of a restaurant.” They operate in dissatisfaction in the quick-service food
such nontraditional locations as dispens- industry. Eliminating variability increases
ing carts, kiosks, and mobile units and do customers’ satisfaction, which leads to
not use Taco Bell’s in-store processing more frequent visits and hence to higher
systems. sales.
The total savings in labor costs Taco Bell Another major benefit from the new
has achieved by using the LMS instead of LMS system is that it functions as a tool
the old FAST system over the four-year managers can use to schedule labor. A re-
period, 1993–1996, are over $40.34 million. view of labor-scheduling practices by one
These cost savings will continue to accrue of the authors revealed that, in a number
each and every year in relation to the of quick-food companies, managers
growth of company and franchised units tended to schedule too little labor to meet
in the Taco Bell system. customer service goals during the peak
Improving the Quality and Consistency hours, and too much during nonpeak
of Customer Service hours of the day. This was attributed to
The documented savings in labor costs both the difficulties in scheduling itself
are impressive. Other benefits may be and the manager’s perception that every-
even more significant. As a result of this body tended to work faster during peak
study, Taco Bell now has a new, flexible, hours and the manager could always per-
and adaptable methodology for determin- sonally staff a position to resolve a bottle-
ing how much labor it actually needs at neck in production. Failing to deliver ex-
each store and how to deploy that labor in cellent service during peak periods can
lower the sales curve for the entire day
1993 1994 1995 1996 and vice versa, or as the well-known say-
ing goes, “If you can’t meet your sales,
Company stores they will decrease until you can.” The new
using LMS 1298 2763 2785 2550 LMS system has diminished this problem
Franchise stores
in the Taco Bell system.
using LMS 0 412 809 1404
Labor cost savings
Enhancing Productivity
(millions) $3.51 $8.54 $11.89 $16.40 Taco Bell also avoided costs by using the
Table 1: Growth in stores using the LMS and LMS to evaluate the impact on labor of a
the corresponding yearly savings over FAST. continuous stream of ideas for improving
January–February 1998 89
HUETER, SWART
INTERFACES 28:1 90
TACO BELL
January–February 1998 91