You are on page 1of 17

The Emerald Research Register for this journal is available at http://www.emeraldinsight.

com/researchregister

The current issue and full text archive of this journal is available at http://www.emeraldinsight.com/0144-3577.htm

Service process design exibility and customer waiting time


Chwen Sheu and Roger McHaney
Department of Management, College of Business Administration, Kansas State University, Manhattan, Kansas, USA, and

Process design exibility

901

Sunil Babbar
Department of Information Technology and Operations Management, College of Business, Florida Atlantic University, Boca Raton, Florida, USA
Keywords Service operations, Waiting lists, Simulation, Service control, Service industries Abstract Customer waiting is regarded as one of the most critical aspects of service quality. Research has suggested various approaches to reduce the negative impact of waiting. This article investigates the waiting time performance of alternative service process designs that consist of two operations, order taking and order preparation. The research premise is that no single service process design is the best in all operating conditions. Managers should build exibility into service process design by using alternative designs in combination. Several break-even models are developed to examine the contingent nature of the performance of alternative designs. The results point to the need for building exibility into service process designs by demonstrating that waiting time performance can only be optimized if design strategies are altered in response to ongoing changes in service system input parameters.

Introduction Consumers today are more constrained by time than ever before. In an intensely competitive world the pressure, expectation and need to accomplish more in less time is unlikely to diminish. Service providers understand the premium that consumers place on time they view as wasted while waiting for the delivery of services. A customer waiting in line for service is potentially a lost customer. Studies show that up to 27 percent of customers who can not get through on the telephone will likely take their business elsewhere (Fitzsimmons and Fitzsimmons, 2000). As such, managers of service operations constantly strive to shorten customer waiting time during service delivery (Durrande-Moreau, 1999; Jones and Peppiatt, 1996). Firms across a variety of industries have introduced numerous peripheral service elements to the service package experience of their customers, in an attempt to shorten customer waiting times. Such efforts are best illustrated by retail stores that have check-out registers which automatically print the date, amount and name of
The authors wish to thank Dr Brett DePaola (Kansas State University, USA) for his valuable suggestions in preparing this paper.

International Journal of Operations & Production Management Vol. 23 No. 8, 2003 pp. 901-917 q MCB UP Limited 0144-3577 DOI 10.1108/01443570310486347

IJOPM 23,8

902

payee on the cheques customers use when making payment. Other examples include hotels that slip the printed bill under the customers door on the morning of departure, and car rental companies that expeditiously send returning cars on their way by closing the rental transaction at the car as it pulls up in the airport parking area. More recently, new technology offers even more opportunities to improve service process and thus customer service in various industries. For instance, the practice of e-ticketing in the airline business has denitely made a huge impact on ticket purchasing as well as airport check-in processes. Other examples include electronic check-in and check-out systems in the hotel industry, automatic toll booths in transportation, collating copy machines, electronic funds transfer in nancial services, wireless order from waiters to the kitchen in restaurants, optical checkout scanners and self-service checkout in supermarkets and telephone switching systems in communication, etc. (Collier, 1994; Haksever et al., 2000). This article studies the effect of service process design on customer waiting time. Service process design refers to the arrangement of service facilities where the service is provided and the processes through which the service operations are structured and delivered (Ramaswamy, 1996). Lovelock classied service process by the extent of demand uctuations and the extent to which supply capacity is constrained (Lovelock, 1983). When demand is highly uctuated and peak demand regularly exceeds capacity, managers must consider altering either demand pattern or supply capacity so that service can be delivered without incurring long customer waiting time. In case demand pattern cannot be altered, managers could consider operations-oriented strategies to control the level of service supply, such as scheduling part-time workers and cross training service personnel (Fitzsimmons and Fitzsimmons, 2000). Similar to those strategies, improving service process design is another approach to increase the exibility of service system to respond more effectively to demand variations, so that customer waiting can be minimized (Bennett, 1990; Collier, 1994; Kolesar and Green, 1998; Fitzsimmons and Fitzsimmons, 2000). The way in which the service process is designed determines, to a large extent, the wait that customers experience. Any reductions in customer waiting time by better management of process design can certainly help lower both customer dissatisfaction and defection (Davis and Heineke, 1998; Taylor, 1994). Often, managers have a number of alternatives to choose from when shaping the design of their service delivery process. Previous research demonstrates ways by which alternative service process designs can potentially inuence the service quality by reducing average waiting time for customers (Kolesar and Green, 1998; Sheu and Babbar, 1996). Nevertheless, the possible ways in which these alternatives can inuence the average waiting time for customers and the magnitude of such inuence, have yet to be fully explored especially when

considering changes in operating conditions resulting from factors such as demand variations and the application of new service process technology. The investigation of factors that can result in a particular service design outperforming others on waiting time and the determination of break-evens between alternative designs, can be particularly useful to managers as they manage the quality of their service. The primary aim of this study is to investigate the waiting time performance of alternative service process designs. It establishes the extent of benets that can be derived by building in exibility for design change in view of the performance contingencies of alternative designs under different operating conditions. Flexibility can be managed by using alternative designs in combination. The research premise is that no single service process design is the best under all operating conditions. Therefore, we hypothesize that managers can reduce the average waiting time by systematically switching back and forth between designs as operating conditions change even in the short run. This article is organized as follows. First, the alternative service process designs examined in this study are described. Next, the comparative performance of these designs is reviewed. In the meantime we develop several research propositions based on the potential benet from switching between designs as operating conditions change. The simulation experiment that veries this benet is outlined and the results of the experiment provided. The nal section presents the managerial implications. Alternative service process designs We base our experiment on four common service process design alternatives (1, 2, 3 and 4). The service process involves two sequential stages of ordering (the receiving of an order from the customer by the server) and preparation (the subsequent tasks that must be performed to fulll or deliver on the order). A service process such as this, where customers enter the service system, form the waiting lines, place their order and wait for its fulllment, is typical of the fast-food industry (Fitzsimmons and Fitzsimmons, 2000). In a facilities-based service customers must visit the service facility in order for the service to be rendered. Figure 1 presents four alternative service process designs that have received much attention in the research literature (Haksever et al., 2000; Bonini et al., 1997; Sheu and Babbar, 1996; Wolff, 1988). Each of designs 1 through 4 offers a unique design structure that differs in terms of how the waiting lines are formed and the manner in which the service process is designed. Design 1 is a multi-stage, single-queue, single-server system in which the two stages of ordering and preparation are separated, and each stage is assigned a single dedicated server. A customer must enter the system, wait for the server to be available, and go through the two sequential stages to have the service completed. Design 2 is a single-stage, multiple-queue,

Process design exibility

903

IJOPM 23,8

904

Figure 1. Four alternative service process

multiple-server system in which separate waiting lines are formed in front of each of the two servers. Here each server is responsible for both taking an order and preparing (the same server performs both stages of a particular customer order). Design 3 is a single-stage, single-queue, multiple-server system. It is similar to design 2 in that it also has two servers, each responsible for both operations. The only difference lies in the fact that customers form a single

queue in the case of design 3, whereas they form multiple queues in design 2. Design 4 is a single-stage, single-queue, single-team-server system in which customers form a single waiting line and are served by a service team consisting of two servers. Here, one server takes the customers order and after the order is completely received, both servers work together as a team, sharing the work in preparing the order. We found an example of this server-team design in a local coffee shop, where typical customer orders include pastry and a cup of coffee. One of the two servers would take the customer order and fetch the order for pastry, while the other server is dedicated to preparing for the coffee. Given the ever-increasing infusion of total quality management (TQM), with its emphasis on teamwork as a way of doing business, the process designs that involve teams are gaining in popularity. The ratio of ordering time to preparation time constitutes an important determinant of performance in the case of design 4, while this ratio has no impact to the other three designs. This is because the work involved in preparation is shared by the two servers. As an example, in a service process that takes a total of 12 time units for ordering and preparation, if the ratio of ordering to preparation time is on average 3:9, then completion of the entire service would take an average of 7.5 (or 3 + 9/2) time units. If instead, the ratio of ordering to preparation time is 9:3, then completion of the entire service would take an average of 10.5 (or 9 + 3/2) time units. In essence, in the case of design 4, a larger ratio results in a longer waiting and service completion time. Among the four alternative designs discussed, literature shows that a single-queue, multiple-server system will always outperform a multiple-queue, multiple-server system in regard to the average total waiting time (the time spent in both queue and service) (Sheu and Babbar, 1996). The superiority of pooling effect in design 3 is shown to be a very general result, independent of the nature of the arrival process and the distribution of service. In other words, a single-queue, multiple-server system (design 3) will always outperform a multiple-queue, multiple-server system (design 2) or a single-queue, single-server queuing system (design 1). There is no dominance between designs 3 and 4, since the relative performance of these designs depends on the ratio of ordering time to preparation time. Sheu and Babbar (1996) developed a break-even model illustrating the interaction effect between the ratio of ordering time to preparation time and the comparative waiting time performance of designs 3 and 4 (Figure 2). When the ratio is less than the break-even point, design 4 is preferred for its shorter waiting time, resulting from more sharing of preparation time. When the ratio is greater than the break-even, less sharing occurs and design 3 provides better performance. Managers could use the break-even ratio to choose the optimal design to reduce customer waiting time. Based on the break-even model this article develops three research propositions

Process design exibility

905

IJOPM 23,8

906

Figure 2. Hypothesized break-even model and choice of service design

pertaining to benets derived from the exibility of switching between designs 3 and 4. Research propositions The break-even model consists of only two variables, customer arrival rate (or inter-arrival time) and service (ordering and preparation) rate. Considering the dynamic nature of operating conditions in reality, we could expect constant uctuations of arrival and service rates. The implication is that the choice of best service design varies with changes of operating conditions, shaped by demand and service levels. No single service design is optimal at all times. This paper extends previous research by examining two possible changes that affect arrival and service rates and thus break-even ratio and choice of service design. These two changes are: customer demand (measured by inter-arrival time) and technological innovation in the service process. The former change affects customer arrival rate, and the latter one affects service rate. Intuitively, higher customer demand for the service requires a larger proportion of work sharing (in preparation) to maintain the same service level measured as waiting time. In other words, the break-even ratio in Figure 1 decreases (with the need for increasing of preparation time) as customer demand increases. As such, demand level variations could change the relative waiting time performance between designs 3 and 4. Therefore, the rst research proposition is: P1. Server-team design (design 4) outperforms multiple-server design (design 3) at lower demand level. Managers must recognize the shift of the break-even ratio resulting from demand variations in order to choose the optimal design to reduce customer waiting time. In this proposition demand level variations are referred to as the permanent increase or decrease of overall demand level. One possible extension to this scenario is the systematic pattern of temporary demand uctuations, such as

an increase in customer arrivals during meal times in the fast-food industry. Such temporary demand uctuations lend themselves to the potential benets from switching between designs 3 and 4, back and forth. We will refer to the service design with the exibility of switching between designs 3 and 4 as design 5. Essentially, design 5 is design 4 with a shift to design 3 during the high demand period and back to design 4 when demand drops to its original level. Our research proposition two is: P2. There is potential benet that managers can derive from the exible service design (design 5) as demand uctuates from one time period to another. This proposition is developed based on P1 that design 4 performs better at low demand levels and at lower ratios, whereas design 3 performs better at higher demand levels and higher ratios. While demand uctuations affect customer arrival rate, technological innovations can improve service speed. Firms constantly apply technological innovations to service delivery processes in attempts to improve both the efciency and the performance of their service systems (Haksever et al., 2000). Applications of new technology are likely to affect ordering and/or preparation processes, which can alter both the service rate and the ratio of ordering to preparation time. For instance, there are now numerous service processes designed to allow customers to take their own orders (e.g. by pushing buttons on a screen or a keyboard). Another example is the practice of e-ticketing, which speeds up the check-in process in the airline industry. The application of new technology reduces the amount of time spent in ordering and/or preparation processes, which in turn changes the ratio of ordering to preparation time. With a change in the ratio, the design that may traditionally have been considered the better design might now, in fact, be the inferior one. The corresponding research proposition is stated as follows: P3. Managers must recognize the shift of the break-even ratio resulting from technological innovations in order to choose the optimal design to reduce customer waiting time. The simulation experiment in the following section veries these three research propositions. Simulation experiment This study used simulation to explore the contingent nature of the performance of individual designs described in the three research propositions. Methodologically, simulation offers many advantages over mathematical modeling, the most important being the exibility in considering the range of changes in parameters and in capturing real world behaviors. We developed a computer simulation capable of evaluating hypothesized service designs and

Process design exibility

907

IJOPM 23,8

908

deriving approximate break-even points. The constructed model needed to represent changes in the service system over time. It also needed to simulate a variety of behaviors based on conditions within the system. Various probability distributions and related logic were used to fulll this need. Finally, the model had to react to events within the system as they occurred. Overall, the goal of this modeling effort was to capture data representative of three hypothesized service designs, design 3 (multiple servers), 4 (server-team) and 5 (exible design). Each design was represented in terms of resources (servers), entities (customers) and the respective interaction. Design 3 was modeled with a single queue feeding two lines, each having a single server responsible for both ordering and preparation. Design 4 also employed a single queue, but the two server resources cooperated in team fashion to serve a single customer. Design 5 operated as a combination of the rst two designs. Following the conceptual design, the model was coded. A discrete event computer simulation was constructed using the GPSS/H simulation language (Schriber, 1991). In general GPSS/H was selected because of its speed, power, exibility and report generation capability. In addition, GPSS is a premier simulation language with a long, well-tested history. Prior to developing production runs the model was validated by entering appropriate input parameters and comparing the results to earlier mathematically derived outputs in the literature (Sheu and Babbar, 1996). No statistically signicant deviation from expected performance was detected. Assumptions of the break-even model and simulation analysis The following outlines all relevant model assumptions: (1) There are two servers in the service delivery process and they are identical in their service capability. The service delivery process is divided into two sequential operations: ordering and preparation. Each server can work individually or together for a single operation or both operations. (2) Depending on the way servers are assigned their tasks, and how customers form the queue(s), the four alternative service process designs are classied into different queuing systems as follows: . design 1 multi-stage, single-queue, single-server queuing system; . design 2 single-stage, multiple-queue, multiple-server queuing system; . design 3 single-stage, single-queue, multiple-server queuing system; and . design 4 single-stage, single-queue, single-team-server queuing system. (3) Arrival rate follows a Poisson process, and service time follows the general distribution. The mean service rate is independent of the number

(4) (5) (6)

(7)

(8)

of customers in the system. Also, the mean service rate is greater than the mean arrival rate. The queue discipline is rst-come, rst-served. All queuing systems have innite queue and innite calling population. Balking behavior occurs when an entering customer detects more than six customers queued for service. No reneging or jockeying is allowed. Process design is the primary factor that determines the waiting time performance. Some qualitative factors such as facility design/layout of the preparation area, server exibility and the effects of learning through task specialization are not included in the queuing analysis. The day runs from 7 a.m. to 5 p.m. Lunch time runs from 11 a.m. to 1 p.m. Arrival rate is one per 16 time units during the regular hours, increasing to one per ten time units in lunch time. Total service time is 12 units (without new technology) and decreases to ten units with new process technology application.

Process design exibility

909

After, the model validation simulation was used to conduct the following four experiments: (1) Design 3 used throughout each day vs design 4 used throughout each day (inter-arrival time = 16 time units). (2) Experiment (1) with various demand levels (various levels of inter-arrival time). (3) Design 5, which involves switching from design 4 to design 3 only during the higher demand period each day. (4) Experiment (1) with new technology application to preparation process. In design 5 the utilization is 38 percent during the regular hours and increases to 60 percent owing to higher customer demand. Regular arrival rates lasted from 7 a.m. to 11 a.m. and again from 1 p.m. to 5 p.m. Peak rates occurred during the two-hour lunch period from 11 a.m. to 1 p.m. A statistic for total time in system was tabulated as a means of understanding the impact of each design on customer waiting time. Modeling of the service times followed the assumptions outlined previously. The expected total service time was 12 time units. This service time was divided according to a ratio of ordering time to preparation time. Each portion was modeled as a general distribution with the standard deviation equal to 20 percent of the mean. Different levels of standard deviation were examined, but no signicant difference was detected, so 20 percent of the mean was used in production runs. In runs for design 5 the servers were allowed to complete tasks started prior to the beginning of the peak time period. After completely serving a customer who started prior to a peak time period, all customers, both those in queue and new arrivals were serviced under the new service design

IJOPM 23,8

910

scenario. The ratio between ordering and preparation was increased from 0 to 1.0 in 0.1 increments. In order to more closely simulate customer behavior in real-world service systems, balking behavior was introduced and examined in our analysis. In all instances balking occurred when an entering customer detected more than six customers queued for service. A different number of customers was used to examine the balking behavior, but no signicant difference was detected. To ensure statistical validity, a non-terminating approach was selected for model analysis (Law, 2000). Independent random number streams were used to model each service and arrival time. Transient start-up periods were identied and excluded from output data collection. Statistics were gathered only after the system reached a steady state. The subinterval method was used to divide the runs into samples and calculate mean values for key outputs. A start-up period of one million time units was used, and each run lasted for an additional million time units, which were subdivided into ten-hour shifts. This provided approximately 1667 observations for each mean output value and represented nearly ve years of single shift operation. Simulation results Performance of individual designs with constant demand The rst four columns in Table I contain the average waiting time performances of designs 1, 2, 3 and 4, with demand held constant at an inter-arrival rate of 16 time units. While the focus of this study is on the performance of designs 3 and 4 and the contingent nature of their comparative performance, the performance of designs 1 and 2 is also included in Table I to establish and conrm the dominance of design 3. First, the waiting time performance of design 3 is always better than that of designs 1 and 2 over the entire range of the ratio of ordering time to preparation
(1) Design 1 (1/l 16) (2) Design 2 (1/l 16) (3) Design 3 (1/l 16) (4) Design 4 (1/l 16) (5) Design 3 (1/l 10) 15.600 15.524 15.496 15.508 15.528 15.480 15.549 15.546 15.521 15.569 15.451 (6) Design 4 (1/l 10) 10.630 12.623 14.613 16.608 18.711 20.556 22.524 23.959 25.599 27.483 29.026

Ratio

Table I. Waiting time performance of individual designs

0.00 30.576 0.10 24.696 0.20 20.297 0.30 17.994 0.40 16.716 0.50 15.996 0.60 15.271 0.70 14.961 0.80 14.648 0.90 14.386 1.00 14.204 Notes: Service time 12

15.663 13.086 7.968 15.772 13.009 8.991 15.554 13.136 9.896 15.807 13.124 10.339 15.733 12.988 11.059 15.546 13.063 11.898 15.611 13.122 12.989 15.724 12.976 13.147 15.550 13.066 13.925 15.810 13.044 14.815 15.525 12.998 15.020 units; number of servers 2.

time. For ratios of 0.00 through 0.60, design 4 performs better than design 3. For ratios of 0.70 through 1.00, where the proportion of preparation time becomes smaller and the benets of team work begin to diminish, design 3 yields superior results. Therefore, break-even between designs 3 and 4 occurs somewhere between the ratios of 0.60 and 0.70. These simulation results verify previous theoretical development (Sheu and Babbar, 1996). Contingent performance of designs with changes in demand Columns (5) and (6) of Table I contain the performance of designs 3 and 4 with a higher demand level (inter-arrival time ten time units or 60 percent of utilization). Notice that the break-even ratio between designs 3 and 4 is now between 0.20 and 0.30, decreasing from the ratio between 0.60 and 0.70. In other words, design 4 is better when the break-even ratio is less than or equal to 0.20, and design 3 performs better once the break-even ratio is greater than or equal to 0.30. Figure 3 shows the shift of break-even ratio owing to the changes in demand. As demand increases (or service facilities utilization increases), more work sharing (greater preparation time) must occur to justify the choice of server-team design. The waiting time performance of service designs is contingent on both the ratio of ordering time and preparation time, as well as demand level. Managers should monitor demand level changes and check for the possibility of switching between designs 3 and 4 for better service quality. As an example, with the ratio of 0.50, design 4 (server-team) provides lower waiting time (11.898 vs 13.063). However, as demand level increases, as shown in Figure 3 and columns (5) and (6) in Table I, design 3 offers better performance (15.480 vs 20.556). Overall, this observation supports P1. Managers should choose the server-team design (design 4) when the demand is low and switch to two-server design (design 3) when the demand is high. In practice, we could develop a demand break-even model for any given ratio of ordering and preparation time. Using the ratio of 0.50, this model was

Process design exibility

911

Figure 3. Break-even analysis with demand changes from 1/l= 16 to 1l= 10 units

IJOPM 23,8

912

developed by repeating the simulation in Figure 3 with different customer arrival rates and recording the corresponding waiting time. Figure 4 shows the waiting time performance of designs 3 and 4 at different demand levels ranging from 38 percent to 90 percent. Given the ratio of 0.50, the break-even occurs at slightly higher than the inter-arrival time of 14 units (or approximately 45 percent of utilization level). Such information would then serve as a guideline for choosing the best service process design as demand level changes. Managers can reduce customer waiting times by developing the break-even demand level based on the current ratio of ordering to preparation time. With the validation of P1, we then investigated a hypothetical situation where managers build exibility into the service process design. The exibility is provided by automatically switching from design 4 to design 3 during lunch time (11 a.m.-1 p.m.) and the returning to design 4 when customer demand slows down. Table II presents the simulation results. Figure 5 illustrates the

Figure 4. Break-even demand model

Ratio

Design 3

Design 4 8.599 9.873 10.921 12.146 13.230 14.163 15.196 16.275 16.692 17.766 18.450

Design 5 10.053 10.758 11.425 12.006 12.630 13.132 13.515 14.050 14.351 14.967 15.202

Table II. Contingent performance of designs 3, 4, and 5 with demand uctuations

0.00 13.718 0.10 13.728 0.20 13.690 0.30 13.710 0.40 13.679 0.50 13.717 0.60 13.714 0.70 13.695 0.80 13.698 0.90 13.754 1.00 13.653 Notes: Regular hours: 1/l= 16; Lunch time: 1/l= 10

Process design exibility

913

Figure 5. Break-even analysis: designs 3, 4 and 5

comparative performance between designs 3, 4 and 5. At ratios of 0.30 through 0.60 the performance of design 5 is better than the performance of either design 3 and 4. When the proportion of ordering time is low, or the ratio is below 0.30, design 4 (team work approach) offers the most benet from sharing preparation activities. On the other hand, when the ratio goes above 0.60, only minimal benets are gained from sharing due to the low proportion of preparation time, and design 3 gives the best performance in waiting time. This nding supports P2. Managers can derive benets from the exible service design (design 5) that switches from design 4 to design 3 during the high demand period. Overall, a exible service design (design 5) is more effective in a service environment with highly uctuating demand, which supports P2. Contingent performance of designs with technological improvement Table III contains the performance of designs 3 and 4 with the application of new technology. Since the new technology is applied to the preparation stage, its application reduces the preparation time as well as the service time. The improvement on service process in turn reduces waiting time. At ratios of 0.00 through 0.70 the performance of design 4 is better than the performance of design 3. For ratios of 0.80 through 1.00 design 3 is the better design. Break-even between designs 3 and 4 occurs somewhere between the ratios of 0.70 and 0.80, which is higher than the one without the new technology applications, 0.60-0.70 in Table I. The implication is that managers must update the ratio with the new process technology and review the choice of process design. As an example, a service process with the ratio of 0.60 (ordering time 4.50 and preparation time 7.50) would choose design 4 prior to new technology application (see Table I). As process improvement decreases preparation time to 5.5 units, the ratio increases to 0.82 (ordering time 4.50 and preparation time 5.50), and Table III indicates that design 3 is now likely

IJOPM 23,8

Ratio

Design 1 (1/l 14)

Design 2 (1/l 14) 12.378 12.412 12.331 12.329 12.312 12.331 12.300 12.344 12.353 12.346 12.324 2 time units

Design 3 (1/l 14) 10.628 10.618 10.619 10.631 10.622 10.596 10.612 10.624 10.629 10.807 10.619

Design 4 (1/l 14) 6.178 7.033 7.787 8.539 9.174 9.758 10.297 10.530 11.320 11.633 12.162

914
Table III. Contingent performance of service designs with new technology application

0.00 18.450 0.10 15.770 0.20 14.316 0.30 13.043 0.40 12.382 0.50 11.897 0.60 11.650 0.70 11.473 0.80 11.522 0.90 11.688 1.00 11.927 Note: Reduction of preparation time by

to offer lower waiting time than design 3. Overall, this observation supports P3. The analysis provides managers with the mechanism to choose the best service design. Conclusions Experts in the service industry have suggested that no aspect of customer service is more important than waiting in line to be served (Bennett, 1990). Previous studies have investigated the effectiveness of various approaches to reduce the negative impact of waiting (Durrande-Moreau, 1999). This paper studies customer waiting time in a service environment where demand is highly uctuated and capacity is limited. The results should be applicable to other services in the similar environment. We investigate the effect of altering service process design on reducing customer waiting due to demand variations. Specically, we use simulation to analyze the waiting time performance of alternative service process designs that consist of two operations, ordering and preparation. This type of queuing conguration can be found in many fast food stores (Sheu and Babbar, 1996; Fitzsimmons and Fitzsimmons, 2000). While demonstrating the potential impact of process design on customer waiting time performance, the simulation study examines the contingent nature of the performance of two specic service designs, multiple-server (design 3) and server-team (design 4). Several break-even models were developed to show how the average waiting time performance can be improved by switching between alternative service designs instead of committing the system to any single dedicated design alternative. The results point to the need for building exibility into process designs by demonstrating that optimality of specic designs can depend on several system parameter values. There is an interaction effect between the ratio of ordering to preparation time and the waiting time performance of alternative service process designs. Further, as

operating conditions change in demand level and/or service rate, so does the relative performance of designs 3 and 4. Another important managerial implication is the effect of technology innovations on customer waiting time. New technology (including process and information technology), has changed the entire service business. Previous research has discussed various issues of technology innovations in service organizations (Haksever et al., 2000). However, the interaction effect of new technology and service process design has never been investigated. In practice, technology innovations can occur in the front ofce, which could affect the ordering process. On the other hand, the innovations can occur in the back ofce, which would impact the preparation process. Regardless of the location of the new technology application, our results show that managers must be aware of the magnitude and the source of changes in service time in order to design a service process minimizing waiting time. Otherwise, managers may not be able to optimize the benets of technology innovations in improving customer service. The objectives of the study and the analysis we undertook go beyond simply showing which design is preferred over the other, and under what conditions. The main contribution of this research lies in bringing about the realization that contingencies do exist for designing service processes on an ongoing basis. As such, managers can gain from any exibility that they build in for changes in process design. The concept of automation the extent to which individual employees can handle a variety of tasks and processes plays an important role in determining the extent of exibility that managers have in tailoring their process design from one time period to another. It follows that the better trained the servers are in being able to perform the various tasks (ordering and preparation) that constitute the delivery of service, the more they facilitate managements ability to derive the benets that changes in process design can deliver. While this study provides valuable managerial insights on the issue of service process design and the contingent nature of their performance, application of the underlying break-even models requires a careful review of the limitations of this simulation analysis. For example, we use the case where only two servers are involved in the service delivery process. Nevertheless, the simulation model can be easily extended, and the general framework applied in assessing service process designs that may involve a larger number of servers and more operation phases. Another issue that the simulation experiment could not address is the unique competitive advantage of team design (design #4). Our eld studies of some local coffee shops indicate that the team design could provide managers with some exibility in improving customer satisfaction. For instance, as two servers are working as a team there is only one server who would have the actual contact with customers. Therefore, managers have the exibility to

Process design exibility

915

IJOPM 23,8

916

assign the server who is more personable and experienced to interact with customers, while the other server could concentrate on behind-the-scene operations. Alternatively, both servers can back up each other to interact with customers. In practice, the exibility provided by server-team design could certainly improve customer perception of service quality based on the experience of a local coffee shop we visited. Managers should consider this exibility as they compare the performance of design 3 and 4. We also note that some intervening factors that may inuence design 4 (server team) performance are not considered in the break-even analysis. For example, the tasks constituting the preparation stage of the service process may not be divisible to the extent that the work load can be equally distributed between the two servers. Additionally, a greater amount of communication is required between the servers when they work as a team. If the servers are not dedicated to specic tasks, then tasks must be divided and assigned as each customer order is received. The nature of this communication and its level of effectiveness can inuence the efciency with which the team works to prepare the order. Finally, facility design can also inuence the teams efciency. The layout of the facility might increase the efciency with which the servers are able to share the workload and work as a team in preparing the order, or it might impede their working effectively as a team by, for example, having them frequently run into one another. Future research should conduct eld studies and evaluate the signicance of those intervening factors to the comparative performance of alternative service process designs.
References Bennett, A. (1990), Their business is on the line, The Wall Street Journal, Vol. 7, December 7, p. B1. Bonini, C.P., Hausman, W.H. and Bierman, H. (1997), Quantitative Analysis for Management, McGraw-Hill, New York, NY. Collier, D.A. (1994), The Service Quality Solution, ASQC, Milwaukee, WI. Davis, M.M. and Heineke, J. (1998), How disconrmation, perception and actual waiting times impact customer satisfaction, International Journal of Service Industry Management, Vol. 9 No. 1, pp. 64-73. Durrande-Moreau, A. (1999), Waiting for service: then years of empirical research, International Journal of Service Industry Management, Vol. 10 No. 2, pp. 171-89. Fitzsimmons, J.A. and Fitzsimmons, M.J. (2000), Service Management: Operations, Strategy, and Information Technology, McGraw-Hill, New York, NY. Haksever, C., Render, B., Russell, R.S. and Murdick, R.G. (2000), Service Management and Operations, Prentice-Hall, Upper Saddle River, NJ. Jones, P. and Peppiatt, E. (1996), Managing perceptions of waiting times in service queues, International Journal of Service Industry Management, Vol. 7 No. 5, pp. 47-61. Kolesar, P.J. and Green, L.Y. (1998), Insights on service system design from a normal approximation to Erlangs delay formula, Production and Operations Management, Vol. 7 No. 3, pp. 282-93.

Law, A.M. (2000), Simulation Modeling and Analysis, 2nd ed., McGraw-Hill, New York, NY. Lovelock, C.H. (1983), Classifying services to gain strategic marketing insights, Journal of Marketing, Vol. 47, Summer, pp. 9-20. Ramaswamy, R. (1996), Design and Management of Service Processes, Addison Wesley, Reading, MA. Schriber, T.J. (1991), Introduction to Simulation Using GPSS/H, John Wiley & Sons, New York, NY. Sheu, C. and Babbar, S. (1996), A managerial assessment of the waiting time performance for alternative service process designs, Omega, Vol. 24 No. 6, pp. 689-703. Taylor, S. (1994), Waiting for service: the relationship between delays and evaluations of services, Journal of Marketing, Vol. 58, April, pp. 56-69. Wolff, R.W. (1988), Stochastic Modeling and the Theory of Queues, Prentice-Hall, Englewood Cliffs, NJ.

Process design exibility

917

You might also like