Harrell−Ghosh−Bowden: I. Study Chapters 14.

Modeling Service © The McGraw−Hill

Simulation Using Systems Companies, 2004
ProModel, Second Edition

C H A P T E R

14 MODELING SERVICE
SYSTEMS

“No matter which line you move to, the other line always moves faster.”
—Unknown

14.1 Introduction
A service system is a processing system in which one or more services are pro-
vided to customers. Entities (customers, patients, paperwork) are routed through
a series of processing areas (check-in, order, service, payment) where resources
(service agents, doctors, cashiers) provide some service. Service systems exhibit
unique characteristics that are not found in manufacturing systems. Sasser, Olsen,
and Wyckoff (1978) identify four distinct characteristics of services that distin-
guish them from products that are manufactured:
1. Services are intangible; they are not things.
2. Services are perishable; they cannot be inventoried.
3. Services provide heterogeneous output; output is varied.
4. Services involve simultaneous production and consumption; the service
is produced and used at the same time.
These characteristics pose great challenges for service system design and man-
agement, particularly in the areas of process design and staffing. Having dis-
cussed general modeling procedures common to both manufacturing and service
system simulation in Chapter 7, and specific modeling procedures unique to
manufacturing systems in Chapter 12, in this chapter we discuss design and
operating considerations that are more specific to service systems. A description
is given of major classes of service systems. To provide an idea of how simula-
tion might be performed in a service industry, a call center simulation example is
presented.

357
Harrell−Ghosh−Bowden: I. Study Chapters 14. Modeling Service © The McGraw−Hill
Simulation Using Systems Companies, 2004
ProModel, Second Edition

358 Part I Study Chapters

14.2 Characteristics of Service Systems

Service systems represent a class of processing systems where entities (cus-
tomers, orders, work, and so forth) are routed through a series of service stations
and waiting areas. Although certain characteristics of service systems are similar
to manufacturing systems, service systems have some unique characteristics. The
aspects of service systems that involve work-flow processing (orders, paperwork,
records, and the like) and product delivery are nearly identical to manufacturing
and will not be repeated here. Those aspects of service systems that are most
different from manufacturing systems are those involving customer processing.
Many of the differences stem from the fact that in service systems, often both
the entity being served and the resource performing the service are human.
Humans have much more complex and unpredictable behavior than parts and
machines. These special characteristics and their implications for modeling are
described here:

• Entities are capricious. System conditions cause humans to change their

minds about a particular decision once it has been made. Customer
reactions to dissatisfactory circumstances include balking, jockeying, and
reneging. Balking occurs when a customer attempts to enter a queue, sees
that it is full, and leaves. Jockeying is where a customer moves to another
queue that is shorter in hopes of being served sooner. Reneging, also
called abandonment, is where a customer enters a waiting line or area,
gets tired of waiting, and leaves. Modeling these types of situations can
become complex and requires special modeling constructs or the use of
programming logic to describe the behavior.
• Entity arrivals are random and fluctuate over time. Rather than
scheduling arrivals as in production, customers enter most service systems
randomly according to a Poisson process (interarrival times are
exponentially distributed). Additionally, the rate of arrivals often changes
depending on the time of day or the day of the week. The fluctuating
pattern of arrivals usually repeats itself daily, weekly, or sometimes
monthly. Accurate modeling of these arrival patterns and cycles is
essential to accurate analysis. Having a fluctuating arrival rate makes most
service simulations terminating simulations with focus on peak activities
and other changes of interest as opposed to steady-state behavior.
• Resource decisions are complex. Typically resource allocation and task
selection decisions are made according to some general rule (like first-
come, first-served). In service systems, however, resources are intelligent
and often make decisions based on more state-related criteria. An increase
in the size of a waiting line at a cashier, for example, may prompt a new
checkout lane to open. A change in the state of an entity (a patient in
recovery requiring immediate assistance) may cause a resource (a nurse or
doctor) to interrupt its current task to service the entity. The flexibility to
respond to state changes is made possible because the resources are
Harrell−Ghosh−Bowden: I. Study Chapters 14. Modeling Service © The McGraw−Hill
Simulation Using Systems Companies, 2004
ProModel, Second Edition

Chapter 14 Modeling Service Systems 359

humans and therefore capable of making more complex decisions than

machines. Because service personnel are often cross-trained and activity
boundaries are flexible, resources can fill in for each other when needed.
Modeling the complex behavior of human resources often requires the use
of if–then logic to define decision rules.
• Resource work pace fluctuates. Another characteristic of human resources
is that work pace tends to vary with time of day or work conditions.
A change in the state of a queue (like the number of customers in line)
may cause a resource (a cashier) to work faster, thereby reducing
processing time. A change in the state of the entity (such as the length of
waiting time) may cause a resource to work faster to complete the
professional service. A change in the state of the resource (like fatigue
or learning curve) may change the service time (slower in the case
of fatigue, faster in the case of a learning curve). To model this variable
behavior, tests must be continually made on state variables in the system
so that resource behavior can be linked to the system state.
• Processing times are highly variable. Service processes vary considerably
due to the nature of the process as well as the fact that the entity and the
server are both human. Consequently, processing times tend to be highly
variable. From a modeling standpoint, processing times usually need to
be expressed using some probability distribution such as a normal or beta
distribution.
• Services have both front-room and back-room activities. In front-room
activities, customer service representatives meet with customers to take
orders for a good or service. In the back room, the activities are carried out
for producing the service or good. Once the service or good is produced, it
is either brought to the customer who is waiting in the front room or, if the
customer has gone, it is delivered to a remote site.

14.3 Performance Measures

The ultimate objectives of a service organization might include maximizing prof-
its or maximizing customer satisfaction. However, these measures for success are
considered external performance criteria because they are not completely deter-
mined by any single activity. Simulation modeling and analysis help evaluate
those measures referred to as internal performance criteria—measures tied to a
specific activity. Simulation performance measures are both quantitative and
time-based and measure the efficiency and the effectiveness of a system configu-
ration and operating logic. Examples of internal performance measures are wait-
ing times, hours to process an application, cost per transaction, and percentage of
time spent correcting transaction errors. Collier (1994) uses the term interlinking
to define the process of establishing quantitative, causal relationships such as
these to external performance measures. He argues that interlinking can be a
Harrell−Ghosh−Bowden: I. Study Chapters 14. Modeling Service © The McGraw−Hill
Simulation Using Systems Companies, 2004
ProModel, Second Edition

360 Part I Study Chapters

powerful strategic and competitive weapon. Here are some typical internal per-
formance measures that can be evaluated using simulation:
• Service time.
• Waiting time.
• Queue lengths.
• Resource utilization.
• Service level (the percentage of customers who can be promptly serviced,
without any waiting).
• Abandonment rate (the percentage of impatient customers who leave the
system).

14.4 Use of Simulation in Service Systems

The use of simulation in service industries has been relatively limited in the past,
despite the many areas of application where simulation has proven beneficial:
health care services (hospitals, clinics), food services (restaurants, cafeterias), and
financial services (banks, credit unions) to name a few. In health care services,
Zilm et al. (1983) studied the impact of staffing on utilization and cost; and
Hancock et al. (1978) employed simulation to determine the optimal number of
beds a facility would need to meet patient demand. In food services, Aran and
Kang (1987) designed a model to determine the optimal seating configuration for
a fast-food restaurant, and Kharwat (1991) used simulation to examine restaurant
and delivery operations relative to staffing levels, equipment layout, work flow,
customer service, and capacity. The use of simulation in the service sector contin-
ues to expand.
Even within manufacturing industries there are business or support activities
similar to those found in traditional service industries. Edward J. Kane of IBM
observed (Harrington 1991),

Just taking a customer order, moving it through the plant, distributing these require-
ments out to the manufacturing floor—that activity alone has thirty sub-process steps
to it. Accounts receivable has over twenty process steps. Information processing is a
whole discipline in itself, with many challenging processes integrated into a single
total activity. Obviously, we do manage some very complex processes separate from
the manufacturing floor itself.

This entire realm of support processes presents a major area of potential applica-
tion for simulation. Similar to the problem of dealing with excess inventory in
manufacturing systems, customers, paperwork, and information often sit idle in
service systems while waiting to be processed. In fact, the total waiting time for
entities in service processes often exceeds 90 percent of the total flow time.
The types of questions that simulation helps answer in service systems can be
categorized as being either design related or management related. Here are some
Harrell−Ghosh−Bowden: I. Study Chapters 14. Modeling Service © The McGraw−Hill
Simulation Using Systems Companies, 2004
ProModel, Second Edition

Chapter 14 Modeling Service Systems 361

typical design and management decisions that can be addressed by simulation:

Design Decisions
1. How much capacity should be provided in service and waiting areas?
2. What is the maximum throughput capability of the service system?
3. What are the equipment requirements to meet service demand?
4. How long does it take to service a customer?
5. How long do customers have to wait before being serviced?
6. Where should the service and waiting areas be located?
7. How can work flow and customer flow be streamlined?
8. What effect would information technology have on reducing non-
value-added time?
Management Decisions
1. What is the best way to schedule personnel?
2. What is the best way to schedule appointments for customers?
3. What is the best way to schedule carriers or vehicles in transportation
systems?
4. How should specific jobs be prioritized?
5. What is the best way to deal with emergency situations when a needed
resource is unavailable?
Because service systems are nearly always in a state of transition, going from
one activity level to another during different periods of the day or week, they rarely
reach a steady state. Consequently, we are frequently interested in analyzing the
transient behavior of service systems. Questions such as “How long is the transient
cycle?” or “How many replications should be run?” become very important. Over-
all performance measures may not be as useful as the performance for each partic-
ular period of the transient cycle. An example where this is true is in the analysis of
resource utilization statistics. In the types of service systems where arrival patterns
and staff schedules fluctuate over the activity cycle (such as by day or week), the av-
erage utilization for the entire cycle is almost meaningless. It is much more infor-
mative to look at the resource utilization for each period of the activity cycle.
Most design and management decisions in service systems involve answering
questions based on transient system conditions, so it is important that the results
of the simulation measure transient behavior. Multiple replications should be run,
with statistics gathered and analyzed for different periods of the transient cycle. In
an attempt to simplify a simulation model, sometimes there is a temptation to
model only the peak period, which is often the period of greatest interest. What is
overlooked is the fact that the state of the system prior to each period and the
length of each period significantly impact the performance measures for any
particular period, including the peak period.
ProModel provides a periodic reporting option that allows statistics to be
gathered by periods. This provides a more complete picture of system activity
over the entire simulation.
Harrell−Ghosh−Bowden: I. Study Chapters 14. Modeling Service © The McGraw−Hill
Simulation Using Systems Companies, 2004
ProModel, Second Edition

362 Part I Study Chapters

14.5 Applications of Simulation in Service Industries

Some of the primary applications of simulation in service industries are process
design, method selection, system layout, staff planning, and flow control.

14.5.1 Process Design

Because of the number of options in customer or paper processing as well as the
flexibility of largely human resources, service systems can be quite fluid, allow-
ing them to be quickly and easily reconfigured. Simulation helps identify config-
urations that are more efficient. Consider a process as simple as a car wash service
that provides the following activities:
• Payment.
• Wash.
• Rinse.
• Dry.
• Vacuum.
• Interior cleaning.
How many ways can cars be processed through the car wash? Each activity could
be done at separate places, or any and even all of them could be combined at a sta-
tion. This is because of the flexibility of the resources that perform each operation
and the imparticularity of the space requirements for most of the activities. Many
of the activities could also be performed in any sequence. Payment, vacuuming,
and interior cleaning could be done in almost any order. The only order that
possibly could not change easily is washing, rinsing, and drying. The other con-
sideration with many service processes is that not all entities receive the same
services. A car wash customer, for example, may forgo getting vacuum service or
interior cleaning. Thus it is apparent that the mix of activities in service processes
can vary greatly.
Simulation helps in process design by allowing different processing sequences
and combinations to be tried to find the best process flow. Modeling process flow is
relatively simple in most service industries. It is only when shifts, resource pools,
and preemptive tasks get involved that it starts to become more challenging.

14.5.2 Method Selection

Method selection usually accompanies process design. When determining what
activities will be performed and how the process should be broken up, alternative
methods for performing the activities are usually taken into consideration. Some
of the greatest changes in the methods used in service industries have come about
as a result of advances in information technology. Routine banking is being
replaced by automated teller machines (ATMs). Shopping is being conducted
over the Internet. Electronic document management systems are replacing paper
Harrell−Ghosh−Bowden: I. Study Chapters 14. Modeling Service © The McGraw−Hill
Simulation Using Systems Companies, 2004
ProModel, Second Edition

Chapter 14 Modeling Service Systems 363

document processing. Processing of orders and other documents that previously

took weeks now takes only days or hours.
Automation of service processes presents challenges similar to the automation
of manufacturing processes. If automating a service process speeds up a particular
activity but does not minimize the overall processing time, it is not effective and
may even be creating waste (such as large pools of waiting entities).

14.5.3 System Layout

An important consideration in designing a service system is the layout of the fa-
cility, especially if the building construction is going to be based on the system
layout. For many service operations, however, the layout is easily changed as it
may involve only rearranging desks and changing offices. Human resources and
desktop computers are much easier to move around than an anchored 200-ton
press used in sheet-metal processing. Still, a good layout can help provide a more
streamlined flow and minimize the amount of movement for both customers and
personnel.
The way in which work areas are laid out can have a significant impact on
customer satisfaction and processing efficiency. In some systems, for example,
multiple servers have individual queues for customers to wait before being
served. This can cause customers to be served out of order of arrival and result in
jockeying and customer discontent. Other systems provide a single input queue
that feeds multiple servers (queuing for a bank teller is usually designed this way).
This ensures customers will be serviced in the order of arrival. It may cause some
reneging, however, if grouping all customers in a single queue creates the percep-
tion to an incoming customer that the waiting time is long.

14.5.4 Staff Planning

A major decision in nearly every service operation pertains to the level and type
of staffing to meet customer demand. Understaffing can lead to excessive waiting
times and lost or dissatisfied customers. Overstaffing can result in needless costs
for resources that are inadequately utilized.
Related to staffing levels is the staff type. In service operations, personnel are
frequently cross-trained so that alternative resources may perform any given task.
One objective in this regard is to make the best use of resources by allowing
higher-skilled resources to fill in occasionally when lower-skilled workers are
unavailable.
Modeling staffing requirements is done by defining the pattern of incoming
customers, specifying the servicing policies and procedures, and setting a trial
staffing level. If more than one skill level is capable of performing a task, the list
and priority for using these alternative resources need to be specified.
After running the simulation, waiting times, abandonment counts, and re-
source utilization rates can be evaluated to determine if the optimum conditions
have been achieved. If results are unacceptable, either the incoming pattern, the
Harrell−Ghosh−Bowden: I. Study Chapters 14. Modeling Service © The McGraw−Hill
Simulation Using Systems Companies, 2004
ProModel, Second Edition

364 Part I Study Chapters

service force, or the servicing policies and procedures can be modified to run
additional experiments.

14.5.5 Flow Control

Flow control is to service systems what production control is to manufacturing
systems. Service system operations planners must decide how to allow customers,
documents, and so on to flow through the system. As in manufacturing systems,
customers and information may be pushed or pulled through the system. By lim-
iting queue capacities, a pull system can reduce the total number of customers or
items waiting in the system. It also reduces the average waiting time of customers
or items and results in greater overall efficiency.
Fast-food restaurants practice pull methods when they keep two or three food
items (like burgers and fries) queued in anticipation of upcoming orders. When
the queue limit is reached, further replenishment of that item is suspended. When
a server withdraws an item, that is the pull signal to the cook to replenish it with
another one. The lower the inventory of prepared food items, the more tightly
linked the pull system is to customer demand. Excessive inventory results in de-
teriorating quality (cold food) and waste at the end of the day.

14.6 Types of Service Systems

Service systems cover a much broader range than manufacturing systems. They
range from a pure service system to almost a pure production system. Customers
may be directly involved in the service process or they may not be involved at all.
The classification of service systems presented here is based on operational char-
acteristics and is adapted, in part, from the classification given by Schmenner
(1994):
• Service factory.
• Pure service shop.
• Retail service store.
• Professional service.
• Telephonic service.
• Delivery service.
• Transportation service.
A description of each in terms of its operating characteristics is given next.

14.6.1 Service Factory

Service factories are systems in which customers are provided services using
equipment and facilities requiring low labor involvement. Consequently, labor
costs are low while equipment and facility costs are high. Service factories
Harrell−Ghosh−Bowden: I. Study Chapters 14. Modeling Service © The McGraw−Hill
Simulation Using Systems Companies, 2004
ProModel, Second Edition

Chapter 14 Modeling Service Systems 365

usually have both front-room and back-room activities with total service being
provided in a matter of minutes. Customization is done by selecting from a menu
of options previously defined by the provider. Waiting time and service time are
two primary factors in selecting the provider. Convenience of location is another
important consideration. Customer commitment to the provider is low because
there are usually alternative providers just as conveniently located.
Examples include banks (branch operations), restaurants, copy centers, bar-
bers, check-in counters of airlines, hotels, and car rental agencies.

14.6.2 Pure Service Shop

In a pure service shop, service times are longer than for a service factory. Service
customization is also greater. Customer needs must be identified before service can
be provided. Customers may leave the location and return later for pickup, check-
ing on an order, payment, or additional service. Price is often determined after the
service is provided. Although front-room activity times may be short, back-room
activity times may be long, typically measured in hours or days. The primary con-
sideration is quality of service. Delivery time and price are of secondary impor-
tance. The customer’s ability to describe the symptoms and possible service
requirements are helpful in minimizing service and waiting times.
When customers arrive, they usually all go through some type of check-in
activity. At this time, a record (paperwork or computer file) is generated for the
customer and a sequence of service or care is prescribed. The duration of the ser-
vice or the type of resources required may change during the process of providing
service because of a change in the status of the entity. After the service is pro-
vided, tests may be performed to make sure that the service is acceptable before
releasing the entity from the facility. If the results are acceptable, the customer
and the record are matched and the customer leaves the system.
Examples include hospitals, repair shops (automobiles), equipment rental
shops, banking (loan processing), Department of Motor Vehicles, Social Security
offices, courtrooms, and prisons.

14.6.3 Retail Service Store

In retail services, the size of the facility is large in order to accommodate many
customers at the same time. Customers have many product options from which
to choose. Retail services require a high degree of labor intensity but a low degree
of customization or interaction with the customer. Customers are influenced
by price more than service quality or delivery time. Customers are interested in
convenient location, assistance with finding the products in the store, and quick
checkout. Total service time is usually measured in minutes.
When customers arrive in a retail shop, they often get a cart and use that cart
as a carrier throughout the purchasing process. Customers may need assistance
from customer service representatives during the shopping process. Once the cus-
tomer has obtained the merchandise, he or she must get in line for the checkout
Harrell−Ghosh−Bowden: I. Study Chapters 14. Modeling Service © The McGraw−Hill
Simulation Using Systems Companies, 2004
ProModel, Second Edition

366 Part I Study Chapters

process. For large items such as furniture or appliances, the customer may have to
order and pay for the merchandise first. The delivery of the product may take
place later.
Examples include department stores, grocery stores, hardware stores, and
convenience stores.

14.6.4 Professional Service

Professional services are usually provided by a single person or a small group
of experts in a particular field. The service is highly customized and provided by
expensive resources. Duration of the service is long, extremely variable, and
difficult to predict because customer involvement during the process is highly
variable.
Processing may be performed by a single resource or multiple resources.
When the customer arrives, the first process is diagnostic. Usually an expert re-
source evaluates the service needed by the customer and determines the type of
service, the estimated service time, and the cost. This diagnosis then dictates what
resources are used to process the order. The duration of the service or the type of
resources required may change during the process of providing service. This is
usually a result of the customer’s review of the work. After the service is provided,
a final review with the customer may be done to make sure that the service is ac-
ceptable. If the results are acceptable, the customer and the record are matched
and the customer leaves the system.
Examples include auditing services, tax preparation, legal services, architec-
tural services, construction services, and tailor services.

14.6.5 Telephonic Service

Telephonic services or teleservicing are services provided over the telephone. They
are unique in that the service is provided without face-to-face contact with the
customer. The service may be making reservations, ordering from a catalog, or pro-
viding customer support. In a telephonic service system, there are a number of
issues to address, including
• Overflow calls—the caller receives a busy signal.
• Reneges—the customer gets in but hangs up after a certain amount of
time if no assistance is received.
• Redials—a customer who hangs up or fails to get through calls again.
The most important criterion for measuring effectiveness is delivery time. The cus-
tomer is interested in getting the service or ordering the product as quickly as possi-
ble. The customer’s ability to communicate the need is critical to the service time.
Calls usually arrive in the incoming call queue and are serviced based on the
FIFO rule. Some advanced telephone systems allow routing of calls into multiple
queues for quicker service. Processing of a call is done by a single resource.
Service duration depends on the nature of the service. If the service is an ordering
Harrell−Ghosh−Bowden: I. Study Chapters 14. Modeling Service © The McGraw−Hill
Simulation Using Systems Companies, 2004
ProModel, Second Edition

Chapter 14 Modeling Service Systems 367

process, then the service time is short. If the service is a technical support process,
then the service time may be long or the call may require a callback after some
research.
Examples include technical support services (hotlines) for software or hard-
ware, mail-order services, and airline and hotel reservations.

14.6.6 Delivery Service

Delivery services involve the ordering, shipping, and delivery of goods, raw
materials, or finished products to points of use or sale. Customers may accept
deliveries only within certain time schedules. In practice, there are often other
constraints besides time windows. Certain sequences in deliveries may be inflex-
ible. Customers are interested in convenient, fast delivery. If the products that are
delivered are perishable or fragile goods, the quality of the products delivered is
also important to the customer.
Deliveries begin with the preparation of the product and loading of the product
on the delivery resources. Determination of the best routing decisions for drivers
may depend on the number or proximity of customers waiting for the product.

14.6.7 Transportation Service

Transportation services involve the movement of people from one place to an-
other. A fundamental difference between transportation and delivery systems is
that people are being transported rather than goods. Another important difference
is that the routes in transportation services tend to be fixed, whereas the routes in
delivery services are somewhat flexible. Customers are interested in convenient,
fast transportation. Cost of transportation plays a significant role in the selection
of the service. Because set schedules and routes are used in transportation, cus-
tomers expect reliable service.
Two types of systems are used in transportation: (1) multiple pickup and
drop-off points and (2) single pickup and drop-off points. In multiple pickup and
drop-off point systems, customers enter and leave the transportation vehicle inde-
pendently. In single pickup and drop-off transportation, customers all enter at one
place and are dropped off at the same destination.
Examples include airlines, railroads, cruise lines, mass transit systems, and
limousine services.

14.7 Simulation Example: A Help Desk Operation

This section, based on an article by Singer and Gasparatos (1994), presents an ex-
ample of a simulation project performed to improve a help desk operation. This
project illustrates the flexibility of processing and how variable some of the data
can be. Even for existing service systems, radical changes can be made with little
disruption that can dramatically improve performance.
Harrell−Ghosh−Bowden: I. Study Chapters 14. Modeling Service © The McGraw−Hill
Simulation Using Systems Companies, 2004
ProModel, Second Edition

368 Part I Study Chapters

14.7.1 Background
Society Bank’s Information Technology and Operations (ITO) group offers a “help
desk” service to customers of the ITO function. This service is offered to both in-
ternal and external customers, handling over 12,000 calls per month. The client
services help desk provides technical support and information on a variety of tech-
nical topics including resetting passwords, ordering PC equipment, requesting
phone installation, ordering extra copies of internal reports, and reporting main-
frame and network problems. The help desk acts as the primary source of commu-
nication between ITO and its customers. It interacts with authority groups within
ITO by providing work and support when requested by a customer.
The old client services help desk process consisted of (1) a mainframe help
desk, (2) a phone/local area network help desk, and (3) a PC help desk. Each of
the three operated separately with separate phone numbers, operators, and facili-
ties. All calls were received by their respective help desk operators, who manually
logged all information about the problem and the customer, and then proceeded to
pass the problem on to an authority group or expert for resolution.
Because of acquisitions, the increased use of information technologies, and the
passing of time, Society’s help desk process had become fragmented and layered
with bureaucracy. This made the help desk a good candidate for a process redesign.
It was determined that the current operation did not have a set of clearly defined
goals, other than to provide a help desk service. The organizational boundaries of the
current process were often obscured by the fact that much of the different help desks’
work overlapped and was consistently being handed off. There were no process per-
formance measures in the old process, only measures of call volume. A proposal was
made to consolidate the help desk functions. The proposal also called for the intro-
duction of automation to enhance the speed and accuracy of the services.

14.7.2 Model Description

The situation suggested a process redesign to be supported by simulation of the
business processes to select and validate different operational alternatives.
Detailed historical information was gathered on call frequencies, call arrival
patterns, and length of calls to the help desk. This information was obtained from
the help desk’s database, ASIM. Table 14.1 summarizes the call breakdown by

TABLE 14.1 Types of Calls for Help Desk

Percent Percent Percent

Time Period Password Reset Device Reset Inquiries Level 1 Level 1A Level 2

7 A.M.–11 A.M. 11.7% 25.7% 8.2% 45.6% 4.6% 47.3%

11 A.M.–2 P.M. 8.8 29.0 10.9 48.7 3.6 44.3
2 P.M.–5 P.M. 7.7 27.8 11.1 46.6 4.4 45.8
5 P.M.–8 P.M. 8.6 36.5 17.8 62.9 3.7 32.2
Average 9.9% 27.5% 9.9% 47.3% 4.3% 48.4%
Harrell−Ghosh−Bowden: I. Study Chapters 14. Modeling Service © The McGraw−Hill
Simulation Using Systems Companies, 2004
ProModel, Second Edition

Chapter 14 Modeling Service Systems 369

call level. Level 1 calls are resolved immediately by the help desk, Level 1A calls
are resolved later by the help desk, and Level 2 calls are handed off to an author-
ity group for resolution.
Historically, calls averaged 2.5 minutes, lasting anywhere from 30 seconds
to 25 minutes. Periodically, follow-up work is required after calls that ranges from
1 to 10 minutes. Overall, the help desk service abandonment rate was 4 to 12 per-
cent (as measured by the percentage of calls abandoned), depending on staffing
levels.
The help desk process was broken down into its individual work steps and
owners of each work step were identified. Then a flowchart that described the
process was developed (Figure 14.1). From the flowchart, a computer simulation
model was developed of the old operation, which was validated by comparing
actual performance of the help desk with that of the simulation’s output. During
the 10-day test period, the simulation model produced results consistent with
those of the actual performance. The user of the model was able to define such
model parameters as daily call volume and staffing levels through the use of the
model’s Interact Box, which provided sensitivity analysis.
Joint requirements planning (JRP) sessions allowed the project team to col-
lect information about likes, dislikes, needs, and improvement suggestions from
users, customers, and executives. This information clarified the target goals of the
process along with its operational scope. Suggestions were collected and priori-
tized from the JRP sessions for improving the help desk process. Internal bench-
marking was also performed using Society’s customer service help desk as a ref-
erence for performance and operational ideas.

FIGURE 14.1
Customer problem, Client notification
Flow diagram of
query, or change and escalation
client services.

Automated problem Help desk Automated warning

recognition recognition

Customer self-service

Automated resolution

Qualified support assistance

Expert support assistance

Harrell−Ghosh−Bowden: I. Study Chapters 14. Modeling Service © The McGraw−Hill
Simulation Using Systems Companies, 2004
ProModel, Second Edition

370 Part I Study Chapters

A target process was defined as providing a single-source help desk (“one-

stop shopping” approach) for ITO customers with performance targets of
• 90 percent of calls to be answered by five rings.
• Less than 2 percent abandonment rate.
Other goals of the target process were to enhance the user’s perception of the help
desk and to significantly reduce the time required to resolve a customer’s request.
A combination of radical redesign ideas (reengineering) and incremental change
ideas (TQM) formed the nucleus of a target help desk process. The redesigned
process implemented the following changes:
• Consolidate three help desks into one central help desk.
• Create a consistent means of problem/request logging and resolution.
• Introduce automation for receiving, queuing, and categorizing calls for
resetting terminals.
• Capture information pertaining to the call once at the source, and, if the
call is handed off, have the information passed along also.
• Use existing technologies to create a hierarchy of problem resolution
where approximately 60 percent of problems can be resolved immediately
without using the operators and approximately 15 percent of the calls can
be resolved immediately by the operators.
• Create an automated warning and problem recognition system that detects
and corrects mainframe problems before they occur.
The original simulation model was revisited to better understand the current
customer service level and what potential impact software changes, automation,
and consolidation would have on the staffing and equipment needs and opera-
tional capacity. Simulation results could also be used to manage the expectations
for potential outcomes of the target process implementation.
Immediate benefit was gained from the use of this application of simulation to
better understand the old operational interrelationships between staffing, call vol-
ume, and customer service. Figure 14.2 shows how much the abandonment rate

FIGURE 14.2 0.20

700 calls per day
Abandonment rates for 0.18
650 calls per day
0.16
Calls abandoned

alternative staffing
levels. 0.14 600 calls per day
0.12 550 calls per day
0.10 500 calls per day
0.08 450 calls per day
0.06 400 calls per day
0.04
0.02
0
5 6 7 8 9 10
Operators staffed
Harrell−Ghosh−Bowden: I. Study Chapters 14. Modeling Service © The McGraw−Hill
Simulation Using Systems Companies, 2004
ProModel, Second Edition

Chapter 14 Modeling Service Systems 371

will change when the average daily call volume or the number of operators varies.
The importance of this graph is realized when one notices that it becomes increas-
ingly harder to lower the abandonment rate once the number of operators increases
above seven. Above this point, the help desk can easily handle substantial
increases in average daily call volume while maintaining approximately the same
abandonment rate.
After modeling and analyzing the current process, the project team evaluated
the following operational alternatives using the simulation model:
• The option to select from a number of different shift schedules so that
staffing can easily be varied from current levels.
• The introduction of the automated voice response unit and its ability to
both receive and place calls automatically.
• The ability of the automated voice response unit to handle device resets,
password resets, and system inquiries.
• The incorporation of the PC and LAN help desks so that clients with
PC-related problems can have their calls routed directly to an available
expert via the automated voice response unit.
• The ability to change the response time of the ASIM problem-logging
system.
Additionally, two alternative staffing schedules were proposed. The alterna-
tive schedules attempt to better match the time at which operators are available for
answering calls to the time the calls are arriving. The two alternative schedules
reduce effort hours by up to 8 percent while maintaining current service levels.
Additional results related to the Alternative Operations simulation model were
• The automated voice response unit will permit approximately 75 percent
of PC-related calls to be answered immediately by a PC expert directly.
• Using Figure 14.2, the automated voice response unit’s ability to aid in
reducing the abandonment rate can be ascertained simply by estimating
the reduction in the number of calls routed to help desk operators and
finding the appropriate point on the chart for a given number of operators.
• Improving the response time of ASIM will noticeably affect the operation
when staffing levels are low and call volumes are high. For example, with
five operators on staff and average call volume of 650 calls per day, a
25 percent improvement in the response time of ASIM resulted in a
reduction in the abandonment rate of approximately 2 percent.

14.7.3 Results
The nonlinear relationship between the abandonment rate and the number of op-
erators on duty (see Figure 14.2) indicates the difficulty in greatly improving per-
formance once the abandonment rate drops below 5 percent. Results generated
from the validated simulation model compare the impact of the proposed staffing
changes with that of the current staffing levels. In addition, the analysis of the
Harrell−Ghosh−Bowden: I. Study Chapters 14. Modeling Service © The McGraw−Hill
Simulation Using Systems Companies, 2004
ProModel, Second Edition

372 Part I Study Chapters

effect of the automated voice response unit can be predicted before implementa-
tion so that the best alternative can be identified.
The introduction of simulation to help desk operations has shown that it can
be a powerful and effective management tool that should be utilized to better
achieve operational goals and to understand the impact of changes. As the au-
tomation project continues to be implemented, the simulation model can greatly
aid management and the project team members by allowing them to intelligently
predict how each new phase will affect the help desk.

14.8 Summary
Service systems provide a unique challenge in simulation modeling, largely due
to the human element involved. Service systems have a high human content in the
process. The customer is often involved in the process and, in many cases, is the
actual entity being processed. In this chapter we discussed the aspects that should
be considered when modeling service systems and suggested ways in which
different situations might be modeled. We also discussed the different types of
service systems and addressed the modeling issues associated with each. The ex-
ample case study showed how fluid service systems can be.

14.9 Review Questions

1. What characteristics of service systems make them different from
manufacturing systems?
2. What internal performance measures are linked to customer satisfaction?
3. Why are service systems more reconfigurable than manufacturing
systems?
4. Identify a service system where the customer is not part of the process.
5. What is abandonment or reneging in simulation terminology?
6. Explain how simulation was used to create the abandonment rate chart
shown in Figure 14.2.
7. What are three questions that simulation might help answer in designing
bus routes for a city bus line?
8. Is a transportation system entity-driven or resource-driven? Explain.
9. Describe conceptually how you would model jockeying in a simulation
model.

References
Aran, M. M., and K. Kang. “Design of a Fast Food Restaurant Simulation Model.” Simu-
lation. Norcross, GA: Industrial Engineering and Management Press, 1987.
Collier, D. A. The Service/Quality Solution. Milwaukee: ASQC Quality Press, 1994.
Harrell−Ghosh−Bowden: I. Study Chapters 14. Modeling Service © The McGraw−Hill
Simulation Using Systems Companies, 2004
ProModel, Second Edition

Chapter 14 Modeling Service Systems 373

Hancock, W.; D. C. Magerlein; R. H. Stores; and J. B. Martin. “Parameters Affecting

Hospital Occupancy and Implications for Facility Sizing.” Health Services Research
13, 1978: pp. 276–89.
Harrington, J. H. Business Process Improvement: The Breakthrough Strategy for Total
Quality, Productivity, and Competitiveness. New York: McGraw-Hill, 1991.
Harrington, J. H. Business Process Improvement. New York: McGraw-Hill, 1994.
Kharwat, A. K. “Computer Simulation: An Important Tool in the Fast Food Industry.”
Proceedings of the 1991 Winter Simulation Conference, eds. B. Nelson, W. D. Kelton,
and G. M. Clark. Piscataway, NJ: Institute of Electrical and Electronics Engineers,
1991.
Sasser, W. E.; R. P. Olsen; and D. D. Wyckoff. Management of Service Operations. Boston:
Allyn & Bacon, 1978, pp. 8–21.
Schmenner, Roger W. Plant and Service Tours in Operations Management. 4th ed. New
York: MacMillan, 1994.
Singer, R., and A. Gasparatos. “Help Desks Hear Voice.” Software Magazine, February
1994, pp. 24–26.
Zilm, F., and R. Hollis. “An Application of Simulation Modeling to Surgical Intensive Care
Bed Needs Analysis.” Hospital and Health Services Administration, Sept/Oct. 1983,
pp. 82–101.