Harrell−Ghosh−Bowden: I. Study Chapters 5.

Getting Started © The McGraw−Hill

Simulation Using Companies, 2004
ProModel, Second Edition

C H A P T E R

5 GETTING STARTED

“For which of you, intending to build a tower, sitteth not down first, and counteth
the cost, whether he have sufficient to finish it? Lest haply, after he hath laid the
foundation, and is not able to finish it, all that behold it begin to mock him,
Saying, This man began to build, and was not able to finish.”
—Luke 14:28–30

5.1 Introduction
In this chapter we look at how to begin a simulation project. Specifically, we
discuss how to select a project and set up a plan for successfully completing it.
Simulation is not something you do simply because you have a tool and a process
to which it can be applied. Nor should you begin a simulation without forethought
and preparation. A simulation project should be carefully planned following basic
project management principles and practices. Questions to be answered in this
chapter are
• How do you prepare to do a simulation study?
• What are the steps for doing a simulation study?
• What are typical objectives for a simulation study?
• What is required to successfully complete a simulation project?
• What are some pitfalls to avoid when doing simulation?
While specific tasks may vary from project to project, the basic procedure for
doing simulation is essentially the same. Much as in building a house, you are
better off following a time-proven methodology than approaching it haphazardly.
In this chapter, we present the preliminary activities for preparing to conduct a
simulation study. We then cover the steps for successfully completing a simulation
project. Subsequent chapters elaborate on these steps. Here we focus primarily

103
Harrell−Ghosh−Bowden: I. Study Chapters 5. Getting Started © The McGraw−Hill
Simulation Using Companies, 2004
ProModel, Second Edition

104 Part I Study Chapters

on the first step: defining the objective, scope, and requirements of the study. Poor
planning, ill-defined objectives, unrealistic expectations, and unanticipated costs
can turn a simulation project sour. For a simulation project to succeed, the objec-
tives and scope should be clearly defined and requirements identified and quantified
for conducting the project.

5.2 Preliminary Activities

Simulation is not a tool to be applied indiscriminately with little or no fore-
thought. The decision to use simulation itself requires some consideration. Is the
application appropriate for simulation? Are other approaches equally as effective
yet less expensive? These and other questions should be raised when exploring
the potential use of simulation. Once it has been determined that simulation is the
right approach, other preparations should be made to ensure that the necessary
personnel and resources are in place to conduct the study. Personnel must be iden-
tified and trained. The right software should also be carefully selected. The nature
of the study and the policies of the organization will largely dictate how to prepare
for a simulation project.

5.2.1 Selecting an Application

The decision to use simulation usually begins with a systemic problem for which
a solution is being sought. The problem might be as simple as getting better
utilization from a key resource or as complex as how to increase throughput while
at the same time reducing cycle time in a large factory. Such problems become
opportunities for system improvement through the use of simulation.
While many opportunities may exist for using simulation, beginners in simu-
lation should select an application that is not too complicated yet can have an im-
pact on an organization’s bottom line. On the one hand, you want to focus on
problems of significance, but you don’t want to get in over your head. The first
simulation project should be one that is well defined and can be completed within
a few weeks. This pilot project might consist of a small process such as a manu-
facturing cell or a service operation with only a few process steps. The goal at this
point should be focused more on gaining experience in simulation than making
dramatic, enterprisewide improvements. Conducting a pilot project is much like a
warm-up exercise. It should be treated as a learning experience as well as a confi-
dence and momentum builder.
As one gains experience in the use of simulation, more challenging projects
can be undertaken. As each project is considered, it should be evaluated first for
its suitability for simulation. The following questions can help you determine
whether a process is a good candidate for simulation:
• Is the process well defined?
• Is process information readily available?
Harrell−Ghosh−Bowden: I. Study Chapters 5. Getting Started © The McGraw−Hill
Simulation Using Companies, 2004
ProModel, Second Edition

Chapter 5 Getting Started 105

• Does the process have interdependencies?

• Does the process exhibit variability?
• Are the potential cost savings greater than the cost of doing the study?
• If it is a new process, is there time to perform a simulation analysis?
• If it is an existing process, would it be less costly to experiment on the
actual system?
• Is management willing to support the project?
The wisdom of asking these questions should be clear. You don’t want to
waste time on efforts or problems that are too unstructured or for which simulation
is not well suited. The last question, which relates to management support, is a
reminder that no simulation project should be undertaken if management is un-
willing to support the study or take the recommendations of the study seriously.
Obviously, multiple projects may fit the criteria just listed, in which case the list of
candidates for simulation should be prioritized using a cost–benefit analysis
(a.k.a. biggest-bang-for-the-buck analysis). Management preferences and timing
may also come into play in prioritization.

5.2.2 Personnel Identification

One of the first preparations for a simulation project is to have management assign
a project leader who works alone or with a project team. When assessing personnel
requirements for the simulation, consideration should be given to factors such as
level of expertise, length of the assignment, and organizational structure. Large
companies, for example, often have a central simulation group that provides sim-
ulation services to entities within the organization. In some organizations, simulation
is used no more than once or twice a year, such as when a major change in opera-
tion is being considered. In such situations, it may be more cost-effective to hire a
simulation consultant than to do the simulation in-house. While the decision to
use a consultant is usually based on the level and availability of in-house simulation
expertise as well as the difficulty of the project, the opportunity to train additional
people or acquire increased simulation expertise to apply to future projects should
not be overlooked.
Assuming the simulation will be done internally, either a simulation special-
ist should be assigned to lead the project or an engineer or manager who works
closely with the specialist should be assigned. The person or group doing the
modeling should be well trained and proficient in the use of one or more simula-
tion packages. If no such experienced individual exists, management must be
willing to invest in employee training in simulation.
The process owner should take an active role in the project and, if possible,
even lead the effort. If the process owner refuses or doesn’t have time to get
involved in the simulation project, the effort will be largely wasted. It is impera-
tive to have someone involved who not only understands the process and what
possible design options are feasible, but who also has responsibility for the
system’s performance.
Harrell−Ghosh−Bowden: I. Study Chapters 5. Getting Started © The McGraw−Hill
Simulation Using Companies, 2004
ProModel, Second Edition

106 Part I Study Chapters

5.2.3 Software Selection

Once the responsibility for doing simulation has been assigned, software tools, if
not already in possession, need to be acquired to develop the simulation model.
Having the right tool for the job can mean the difference between success and fail-
ure in a simulation project. Modeling efforts sometimes go on for months without
much progress because of limited capabilities in the software. Simulation soft-
ware can’t make a simulation succeed, but it can make it fail.
Selecting the right simulation software requires that an assessment first be
made of the simulation requirements. Then alternative products should be evalu-
ated against those requirements. If simulation is to be used in multiple applica-
tions that are quite diverse, it may be desirable to have more than one simulation
product so that the right tool is used for the right job.
In selecting simulation software, attention should be given to the following
criteria:
• Quality—Reliable algorithms, absence of bugs, and graceful handling of
user input errors.
• Features and capability—Ability to model systems of any size and
complexity and meet simulation needs in terms of external interfacing,
animation, reporting, and so forth.
• Ease of use—Intuitive editing and built-in constructs that minimize the
time and effort to learn and use the product.
• Services—Quality and level of training and technical support, including
the long-term sustainability of this service.
• Cost—The total cost of ownership (TCO), which includes cost of
procurement, training, support, and labor to use the software.
Although fewer than a dozen major products are on the market, there are over
a hundred products from which to choose. It is best to conduct an initial screening
to develop a short list. Then a more in-depth evaluation can be conducted on the
three or four finalists. In conducting the initial screening, different products
should be screened in terms of their ability to meet minimum requirements in the
areas just identified. Often vendor literature or buying guides are helpful at this
stage of the evaluation. Magazines such as IIE Solutions and OR/MS Today run
regular reviews of simulation products.
In the final evaluation phase, the products should be examined with greater
scrutiny in each of the areas just listed. Since all major software products have
bugs, it is best to look for a product that is mature and has a large user base. This
will help ensure that even rare bugs have been identified and fixed. Simulation
features are pretty easy to assess because most vendors have feature lists avail-
able. One of the most difficult criteria to assess in a simulation product is ease of
use. Most vendors will likely claim that their product is the easiest to use, so it is
always a good idea to have the vendor being considered build a simple model in
your presence that is representative of the target application. This is the best way
to test the suitability and ease of use of the software. At a minimum, it is good to
experiment building simple models with the software to see how user friendly it is.
Harrell−Ghosh−Bowden: I. Study Chapters 5. Getting Started © The McGraw−Hill
Simulation Using Companies, 2004
ProModel, Second Edition

Chapter 5 Getting Started 107

Many vendors offer guarantees on their products so they may be returned after
some trial period. This allows you to try out the software to see how well it fits
your needs.
The services provided by the software provider can be a lifesaver. If working
late on a project, it may be urgent to get immediate help with a modeling or soft-
ware problem. Basic and advanced training classes, good documentation, and lots
of example models can provide invaluable resources for becoming proficient in
the use of the software.
When selecting simulation software, it is important to assess the total cost of
ownership. There often tends to be an overemphasis on the purchase price of the
software with little regard for the cost associated with learning and using the
software. It has been recommended that simulation software be purchased on
the basis of productivity rather than price (Banks and Gibson 1997). The purchase
price of the software can sometimes be only a small fraction of the cost in time
and labor that results from having a tool that is difficult to use or inadequate for
the application.
Other considerations that may come into play when selecting a product in-
clude quality of the documentation, hardware requirements (for example, is a
graphics accelerator card required?), and available consulting services.

5.3 Simulation Procedure

Once a suitable application has been selected and appropriate tools and personnel
are in place, the simulation study can begin. Simulation is much more than build-
ing and running a model of the process. Successful simulation projects are well
planned and coordinated. While there are no strict rules on how to conduct a sim-
ulation project, the following steps are generally recommended:
Step 1: Define objective, scope, and requirements. Define the purpose
of the simulation project and what the scope of the project will be.
Requirements need to be defined in terms of resources, time, and budget
for carrying out the project.
Step 2: Collect and analyze system data. Identify, gather, and analyze the
data defining the system to be modeled. This step results in a conceptual
model and a data document on which all can agree.
Step 3: Build the model. Develop a simulation model of the system.
Step 4: Validate the model. Debug the model and make sure it is a
credible representation of the real system.
Step 5: Conduct experiments. Run the simulation for each of the
scenarios to be evaluated and analyze the results.
Step 6: Present the results. Present the findings and make
recommendations so that an informed decision can be made.
Each step need not be completed in its entirety before moving to the next
step. The procedure for doing a simulation is an iterative one in which activities
Harrell−Ghosh−Bowden: I. Study Chapters 5. Getting Started © The McGraw−Hill
Simulation Using Companies, 2004
ProModel, Second Edition

108 Part I Study Chapters

are refined and sometimes redefined with each iteration. The decision to push
toward further refinement should be dictated by the objectives and constraints of
the study as well as by sensitivity analysis, which determines whether additional
refinement will yield meaningful results. Even after the results are presented,
there are often requests to conduct additional experiments. Describing this itera-
tive process, Pritsker and Pegden (1979) observe,
The stages of simulation are rarely performed in a structured sequence beginning with
problem definition and ending with documentation. A simulation project may involve
false starts, erroneous assumptions which must later be abandoned, reformulation of
the problem objectives, and repeated evaluation and redesign of the model. If properly
done, however, this iterative process should result in a simulation model which prop-
erly assesses alternatives and enhances the decision making process.

Figure 5.1 illustrates this iterative process.

FIGURE 5.1
Define objective,
Iterative nature of scope, and
simulation. requirements

Collect and
analyze system
data

Build model

Validate model

Conduct
experiments

Present
results
Harrell−Ghosh−Bowden: I. Study Chapters 5. Getting Started © The McGraw−Hill
Simulation Using Companies, 2004
ProModel, Second Edition

Chapter 5 Getting Started 109

The remainder of this chapter focuses on the first step of defining the objec-
tive, scope, and requirements of the study. The remaining steps will be discussed
in later chapters.

5.4 Defining the Objective

The objective of a simulation defines the purpose or reason for conducting the
simulation study. It should be realistic and achievable, given the time and resource
constraints of the study. Simulation objectives can be grouped into the following
general categories:
• Performance analysis—What is the all-around performance of the system
in terms of resource utilization, flow time, output rate, and so on?
• Capacity/constraint analysis—When pushed to the maximum, what is
the processing or production capacity of the system and where are the
bottlenecks?
• Configuration comparison—How well does one system configuration
meet performance objectives compared to another?
• Optimization—What settings for particular decision variables best
achieve desired performance goals?
• Sensitivity analysis—Which decision variables are the most influential on
performance measures, and how influential are they?
• Visualization—How can system dynamics be most effectively visualized?
Naturally, a simulation project may have multiple objectives and therefore
fall under two or more of the categories listed here. For example, a sensitivity
analysis may be conducted as part of an optimization study to determine the
highest-leveraging variables on which to optimize. After the model is optimized,
it may be necessary to focus attention on enhancing the visualization of the model
to more effectively sell the solution to management.
Sometimes the objectives change or expand as the project progresses. The
insights provided by simulation give rise to previously unforeseen opportunities
for improvement. Furthermore, the very process of doing a simulation helps refine
and clarify objectives. Staying as close as possible to the initial scope of the pro-
ject is important, and changes to the scope should be documented and assessed in
terms of their impact on budget and schedule. Otherwise, “scope creep” sets in
and the project never ends.
Defining objectives for a simulation study should take into account the ulti-
mate intended use of the model. Some models are built as “throwaway” models
that are used only once and then discarded. Other models will be reused for
ongoing what-if analyses. Some simulations are run to obtain a single summary
statistical measure. Others require realistic animation to convince a skeptical cus-
tomer. Some models are only for the use of the analyst and need not have a user-
friendly interface. Other models are intended for use by managers with little
Harrell−Ghosh−Bowden: I. Study Chapters 5. Getting Started © The McGraw−Hill
Simulation Using Companies, 2004
ProModel, Second Edition

110 Part I Study Chapters

simulation background and therefore require a customized, easy-to-use interface.

Some models are used to make decisions of minor consequence. Other models
serve as the basis for major financial decisions.
Following is a list of sample design and operational questions that simulation
can help answer. They are intended as examples of specific objectives that might
be defined for a simulation study.

Design Decisions
1. What division and sequence of processing activities provide the best
flow?
2. What is the best layout of offices, machines, equipment, and other work
areas for minimizing travel time and congestion?
3. How many operating personnel are needed to meet required production
or service levels?
4. What level of automation is the most cost-effective?
5. How many machines, tools, fixtures, or containers are needed to meet
throughput requirements?
6. What is the least-cost method of material handling or transportation that
meets processing requirements?
7. What are the appropriate number and location of pickup and drop-off
points in a material handling or transportation system that minimizes
waiting times?
8. What are the optimum number and size of waiting areas, storage areas,
queues, and buffers?
9. What is the effect of localizing rather than centralizing material storage,
resource pools, and so forth?
10. What automation control logic provides the best utilization of resources?
11. What is the optimum unit load or batch size for processing?
12. Where are the bottlenecks in the system, and how can they be
eliminated?
13. How many shifts are needed to meet a specific production or service
level?
14. What is the best speed to operate conveyors and other handling or
transportation equipment to meet move demands?

Operational Decisions
1. What is the best way to route material, customers, or calls through the
system?
2. What is the best way to allocate personnel for a particular set of tasks?
3. What is the best schedule for preventive maintenance?
4. How much preventive maintenance should be performed?
5. What is the best priority rule for selecting jobs and tasks?
Harrell−Ghosh−Bowden: I. Study Chapters 5. Getting Started © The McGraw−Hill
Simulation Using Companies, 2004
ProModel, Second Edition

Chapter 5 Getting Started 111

6. What is the best schedule for producing a mix of products or

customers?
7. What is the best inventory replenishment policy?
8. How much raw material and work-in-process inventory should be
maintained?
9. What is the best production control method (kanban, for example)?
10. What is the optimum sequence for producing a particular set of jobs or
processing a given set of customers?
11. How many personnel should be scheduled for a particular shift?

When the goal is to analyze some aspect of system performance, the tendency
is to think in terms of the mean or expected value of the performance metric. For
example, we are frequently interested in the average contents of a queue or the
average utilization of a resource. There are other metrics that may have equal or
even greater meaning that can be obtained from a simulation study. For example,
we might be interested in variation as a metric, such as the standard deviation in
waiting times. Extreme values can also be informative, such as the minimum and
maximum number of contents in a storage area. We might also be interested in a
percentile such as the percentage of time that the utilization of a machine is less
than a particular value, say, 80 percent. While frequently we speak of designing
systems to be able to handle peak periods, it often makes more sense to design for
a value above which values only occur less than 5 or 10 percent of the time. It is
more economical, for example, to design a staging area on a shipping dock based
on 90 percent of peak time usage rather than based on the highest usage during
peak time. Sometimes a single measure is not as descriptive as a trend or pattern
of performance. Perhaps a measure has increasing and decreasing periods such as
the activity in a restaurant. In these situations, a detailed time series report would
be the most meaningful.
It is always best if objectives can be clearly, completely, and precisely stated.
You need to be able to set your sight clearly on what you are trying to accomplish
in the simulation study. You also need to be able to know whether you hit your tar-
get objective. To be effective, an objective should be one that

• Has high potential impact (reduce throughput rate costs, not reduce
backtraveling).
• Is achievable (20 percent inventory reduction, not zero inventory).
• Is specific (reduce waiting time in a particular queue, not eliminate
waste).
• Is quantifiable (reduce flow time by 40 percent, not reduce flow time).
• Is measurable (increase yield by 10 percent, not improve morale by
10 percent).
• Identifies any relevant constraints (reduce turnaround time by
20 percent without adding resources, not just reduce turnaround time by
20 percent).
Harrell−Ghosh−Bowden: I. Study Chapters 5. Getting Started © The McGraw−Hill
Simulation Using Companies, 2004
ProModel, Second Edition

112 Part I Study Chapters

Examples of well-stated objectives might be the following:

• Find the lowest-cost solution to reduce patient waiting time in an emergency
room so that no more than 10 percent of patients wait longer than 15 minutes.
• Increase the throughput rate of an assembly line by an average of
20 percent without additional capital expenditure.
While well-defined and clearly stated objectives are important to guide the
simulation effort, they should not restrict the simulation or inhibit creativity.
Michael Schrage (1999) observes that “the real value of a model or simulation may
stem less from its ability to test a hypothesis than from its power to generate use-
ful surprise. Louis Pasteur once remarked that ‘chance favors the prepared mind.’
It holds equally true that chance favors the prepared prototype: models and simu-
lations can and should be media to create and capture surprise and serendipity. . . .
The challenge is to devise transparent models that also make people shake their
heads and say ‘Wow!’” The right “experts” can be “hypervulnerable to surprise
but well situated to turn surprise to their advantage. That’s why Alexander Flem-
ing recognized the importance of a mold on an agar plate and discovered peni-
cillin.” Finally, he says, “A prototype should be an invitation to play. You know
you have a successful prototype when people who see it make useful suggestions
about how it can be improved.”

5.5 Defining the Scope of Work

With a realistic, meaningful, and well-defined objective established, a scope of
work can be defined for achieving the stated objective. The scope of work is im-
portant for guiding the study as well as providing a specification of the work to be
done upon which all can agree. The scope is essentially a project specification that
helps set expectations by clarifying to others exactly what the simulation will in-
clude and exclude. Such a specification is especially important if an outside con-
sultant is performing the simulation so that there is mutual understanding of the
deliverables required.
An important part of the scope is a specification of the models that will be
built. When evaluating improvements to an existing system, it is often desirable to
model the current system first. This is called an “as-is” model. Results from the
as-is model are statistically compared with outputs of the real-world system to
validate the simulation model. This as-is model can then be used as a benchmark
or baseline to compare the results of “to-be” models. For reengineering or process
improvement studies, this two-phase modeling approach is recommended. For
entirely new facilities or processes, there will be no as-is model. There may, how-
ever, be several to-be models to compare.
To ensure that the budget and schedule are realistic, a detailed specification
should be drafted to include
• Model scope.
• Level of detail.
Harrell−Ghosh−Bowden: I. Study Chapters 5. Getting Started © The McGraw−Hill
Simulation Using Companies, 2004
ProModel, Second Edition

Chapter 5 Getting Started 113

• Data-gathering responsibilities.
• Experimentation.
• Form of results.

5.5.1 Determining Model Scope

Model scope refers to what system elements should be represented in the model.
Determining the scope of a model should be based on how much impact a partic-
ular activity has on achieving the objectives of the simulation.
One decision regarding model scope relates to the model boundaries. A com-
mon tendency for beginners is to model the entire system, even when the problem
area and all relevant variables are actually isolated within a smaller subsystem. If,
for example, the objective is to find the number of operators needed to meet a
required production level for a machining cell, it probably isn’t necessary to
model what happens to parts after leaving the cell. Figure 5.2 illustrates how the
extent of the model should be confined to only those activities whose interactions
have a direct bearing on the problem being studied. If the objective of the simula-
tion is to find how many operators are needed to perform activity C, upstream and
downstream activities that do not impact this decision should be omitted from the
model. In Figure 5.2, since the input rate to activity B is reasonably predictable,
based on knowledge of how activity A behaves, activity A is excluded from the
model and only modeled as an arrival rate to location 2. Since activity E never
causes entities to back up at D, activity E can also be ignored.
In addition to limiting the model boundaries, other system elements may be
excluded from a model due to their irrelevance. Factors such as off-shift mainte-
nance activities or machine setups that are brief and occur only infrequently can
usually be safely ignored. The determination of what elements can be eliminated
should be based on their relevance to the objectives of the study.

5.5.2 Deciding on Level of Detail

Where model scope is concerned more with model breadth, the level of detail de-
termines the depth of a model. More specifically, it defines the granularity or res-
olution of the model. At one extreme, an entire factory can be modeled as a single
“black box” operation with a random activity time. At the other extreme is a

FIGURE 5.2
Confining the model to
impacting activities.
Activity Activity Activity Activity Activity
A B C D E

Scope of model
Harrell−Ghosh−Bowden: I. Study Chapters 5. Getting Started © The McGraw−Hill
Simulation Using Companies, 2004
ProModel, Second Edition

114 Part I Study Chapters

FIGURE 5.3
Effect of level of detail on model development time.

One-to-one
correspondence

Minimum
required

Level of
detail

Model development time

“white box” model that is very detailed and produces a one-to-one correspon-
dence between the model and the system.
Determining the appropriate level of detail is an important decision. Too
much detail makes it difficult and time-consuming to develop and debug the
model. Too little detail may make the model unrealistic by oversimplifying the
process. Figure 5.3 illustrates how the time to develop a model is affected by
the level of detail. It also highlights the importance of including only enough de-
tail to meet the objectives of the study.
The level of detail is determined largely by the degree of accuracy required in
the results. If only a rough estimate is being sought, it may be sufficient to model just
the flow sequence and processing times. If, on the other hand, a close answer is
needed, all of the elements that drive system behavior should be precisely modeled.

5.5.3 Assigning Data-Gathering Responsibilities

In any simulation project, data gathering is almost always the most difficult and
time-consuming task. Identifying data requirements and who will be responsible
for gathering the data is essential if this activity is to succeed. If the modeler is re-
sponsible for gathering data, it is important to have the full cooperation of those
who possess the data, including equipment vendors, engineers, technicians, direct
labor personnel, and supervisors. It is also a good idea to get those individuals
involved who have a stake in the results, such as managers and possibly even cus-
tomers. The more fingerprints you can get on the knife, the stronger the commit-
ment you’ll have to the success of the project.
Harrell−Ghosh−Bowden: I. Study Chapters 5. Getting Started © The McGraw−Hill
Simulation Using Companies, 2004
ProModel, Second Edition

Chapter 5 Getting Started 115

Dates should be set for completing the data-gathering phase because, left un-
controlled, it could go on indefinitely. One lesson you learn quickly is that good data
are elusive and you can always spend more time trying to refine the data. Nearly all
models are based partially on assumptions, simply because complete and accurate
information is usually lacking. The project team, in cooperation with stakeholders
in the system, will need to agree on the assumptions to be made in the model.

5.5.4 Planning the Experimentation

The number and nature of the scenarios or alternative configurations to be evalu-
ated should be planned from the outset to ensure that adequate time is allotted.
Often this decision itself is influenced by the time constraints of the study. Where
only slightly different scenarios are being evaluated, a single model can be devel-
oped and modified appropriately to run each scenario. If alternative configura-
tions are significantly different, separate models may need to be built. It may
require nearly as much effort modeling each configuration as it does developing
the initial model.
For studies considering improvements to an existing system, it is often help-
ful and effective to model the current system as well as the proposed system. The
basic premise is that you are not ready to improve a system until you understand
how the current system operates. Information on the current system operation is
easier to obtain than information on the proposed change of operation. Once a
model of the current system is built, it is easier to visualize what changes need to
be made for the modified system. Both systems may even be modeled together in
the same simulation and made to run side by side. During the final presentation of
the results, being able to show both the as-is and to-be versions of the system
effectively demonstrates the differences in performance.

5.5.5 Determining the Form of Results

The form in which the results are to be presented can significantly affect the time
and effort involved in the simulation study. If a detailed animation or an extensive
written report is expected, the project can easily stretch on for several weeks after
the experimentation phase has been completed. Many times the only result re-
quired is a simple verification of whether a system is capable of meeting a produc-
tion requirement or service level. In such cases, a single number may suffice. In
other instances, a complete document may be required that details all of the goals,
data sources, assumptions, modeling procedures, experiments, analysis of results,
and recommendations. The key to determining the kind and quantity of informa-
tion to present to the decision maker or stakeholder is to ask what decision is being
made with the simulation and what the background is of the person(s) who will be
making the decision. Attention should then be focused on providing adequate in-
formation and effective visualization to enable the right decision to be made.
Sometimes the simulation animation can effectively show bottleneck areas or
compare alternative scenarios. When showing animation, use graphics and displays
Harrell−Ghosh−Bowden: I. Study Chapters 5. Getting Started © The McGraw−Hill
Simulation Using Companies, 2004
ProModel, Second Edition

116 Part I Study Chapters

that focus attention on the area of interest. If lots of color and detail are added to the
animation, it may detract from the key issues. Usually the best approach is to keep
stationary or background graphics simple, perhaps displaying only a schematic of
the layout using neutral colors. Entities or other dynamic graphics can then be
displayed more colorfully to make them stand out. Sometimes the most effective
presentation is a realistic 3-D animation. Other times the flow of entities along a
flowchart consisting of simple boxes and arrows may be more effective.
Another effective use of animation in presenting the results is to run two or
more scenarios side by side, displaying a scoreboard that shows how they com-
pare on one or two key performance measures. The scoreboard may even include
a bar graph or other chart that is dynamically updated to compare results.
Most decision makers such as managers need to have only a few key items of
information for making the decision. It should be remembered that people, not the
model, make the final decision. With this in mind, every effort should be made to
help the decision maker clearly understand the options and their associated
consequences. The use of charts helps managers visualize and focus their atten-
tion on key decision factors. Charts are attention grabbers and are much more ef-
fective in making a point than plowing through a written document or sheets of
computer printout.

5.6 Defining Project Requirements

With the scope of work defined, resource, budget, and time requirements can be
determined for the project. Most of the resources (personnel, hardware, and so on)
should have already been put in place when the decision was made to do a simu-
lation. The primary task at this point is to develop a budget and schedule for the
project. A budget or schedule overrun can sour an otherwise great modeling effort.
It does little good if simulation solves a problem but the time to do the simulation
extends beyond the deadline for applying the solution, or the cost to find the solu-
tion exceeds the benefit that is derived. Realistic budgeting and scheduling are
based on a complete knowledge of the requirements of the study. It is not uncom-
mon to begin a simulation project with aspirations of developing an impressively
detailed model or of creating a stunningly realistic animation only to scramble at
the last minute to throw together a crude model that barely meets the deadline.
Obviously, the time to perform a simulation project will vary depending on
the size and difficulty of the project. If input data cannot be readily obtained, it
may be necessary to add several additional weeks to the project. Assuming input
data are readily available, a small project typically takes two to four weeks, while
large projects can take two to four months. A simulation schedule should be based
on realistic projections of the time requirements, keeping in mind that
• Data gathering and analysis can take over 50 percent of the overall project
time.
• Model building usually takes the least amount of time (10 to 20 percent).
Harrell−Ghosh−Bowden: I. Study Chapters 5. Getting Started © The McGraw−Hill
Simulation Using Companies, 2004
ProModel, Second Edition

Chapter 5 Getting Started 117

• Once a base model is built, several additional weeks may be needed to

conduct all of the desired experiments, especially if the purpose of the
project is to make system comparisons.
• At least several days must be allowed to clean up the models and develop
the final presentation.
• Simulation projects follow the 90–10 rule, where the first 90 percent takes
10 percent of the time, and the last 10 percent takes the other 90 percent.

5.7 Reasons Why Simulation Projects Fail

No matter how good the intentions, simulation projects can fail if not carefully
planned and executed. Many specific reasons contribute to simulation failure.
Some of the most common causes of unsuccessful projects include
• Unclear objectives.
• Unskilled modelers.
• Unavailable data.
• Unmanaged expectations.
• Unsupportive management.
• Underestimated requirements.
• Uninvolved process owner(s).
If the right procedure is followed, and the necessary time and resources are com-
mitted to the project, simulation is always going to provide some benefit to the
decision-making process. The best way to ensure success is to make sure that
everyone involved is educated in the process and understands the benefits and
limitations of simulation.

5.8 Summary
Simulation projects are almost certain to fail if there is little or no planning. Doing
simulation requires some preliminary work so that the appropriate resources and
personnel are in place. Beginning a simulation project requires selecting the right
application, defining objectives, acquiring the necessary tools and resources, and
planning the work to be performed. Applications should be selected that hold the
greatest promise for achieving company goals.
Simulation is most effective when it follows a logical procedure. Objectives
should be clearly stated and a plan developed for completing the project. Data
gathering should focus on defining the system and formulating a conceptual
model. A simulation should then be built that accurately yet minimally captures
the system definition. The model should be verified and validated to ensure that
the results can be relied upon. Experiments should be run that are oriented toward
meeting the original objectives. Finally, the results should be presented in a way
Harrell−Ghosh−Bowden: I. Study Chapters 5. Getting Started © The McGraw−Hill
Simulation Using Companies, 2004
ProModel, Second Edition

118 Part I Study Chapters

that clearly represents the findings of the study. Simulation is an iterative process
that requires redefinition and fine-tuning at all stages in the process.
Objectives should be clearly defined and agreed upon to avoid wasted efforts.
They should be documented and followed to avoid “scope creep.”Aconcisely stated
objective and a written scope of work can help keep a simulation study on track.
Specific items to address when planning a study include defining the model
scope, describing the level of detail required, assigning data-gathering responsi-
bilities, specifying the types of experiments, and deciding on the form of the re-
sults. Simulation objectives together with time, resources, and budget constraints
drive the rest of the decisions that are made in completing a simulation project.
Finally, it is people, not the model, who ultimately make the decision.
The importance of involvement of process owners and stakeholders through-
out the project cannot be overemphasized. Management support throughout the
project is vital. Ultimately, it is only when expectations are met or exceeded that
a simulation can be deemed a success.

5.9 Review Questions

1. Why are good project management skills important to the success of a
simulation project?
2. If you are new to simulation and have the option of simulating either a
small system with a small payback or a large system with a large
payback, which project would you choose and why?
3. How can simulation software make or break a simulation study?
4. When evaluating ease of use in simulation software, how can you make
an informed decision if every vendor advertises ease of use?
5. Why is it important to have clearly defined objectives for a simulation
that everyone understands and can agree on?
6. For each of the following systems, define one possible simulation
objective:
a. A manufacturing cell.
b. A fast-food restaurant.
c. An emergency room of a hospital.
d. A tire distribution center.
e. A public bus transportation system.
f. A post office.
g. A bank of elevators in a high-rise office building.
h. A time-shared, multitasking computer system consisting of a server,
many slave terminals, a high-speed disk, and a processor.
i. A rental car agency at a major airport.
7. How can you effectively manage expectations during a simulation
project?
8. Suppose you are determining the size of a waiting area for a service
facility (such as a clinic, a car repair shop, or a copy center) where the
Harrell−Ghosh−Bowden: I. Study Chapters 5. Getting Started © The McGraw−Hill
Simulation Using Companies, 2004
ProModel, Second Edition

Chapter 5 Getting Started 119

arrival rate fluctuates throughout the day. On which of the following

metrics would you base the waiting area capacity and why?
a. The average number of customers waiting.
b. The peak number of customers waiting.
c. The 95th percentile of customers waiting.
9. What step in doing a simulation study generally takes the longest?
10. What stage in model building usually gets cut short?
11. What factors should determine the scope of a model?
12. What factors should determine the level of detail to include in a model?
13. What is the advantage of building as-is and to-be models?
14. What should determine the form of results of a simulation study?

5.10 Case Studies

CASE STUDY A
AST COMPUTES BIG BENEFITS USING SIMULATION
John Perry
Manufacturing Engineer

Source: AST Research Inc. Used with permission by ProModel, 1999.

In the world of computer manufacturing and assembly, success is measured in terms of

seconds. The PC business poses some unique challenges: it is intensely competitive with
thin profit margins and is characterized by constantly changing technologies and products.
The race is to excel in rapid delivery to the marketplace through high-velocity manufac-
turing. A manufacturer that can minimize the time needed to assemble a computer, move
nimbly to respond to change, and get maximum productivity out of its capital investment
has an edge. At AST Research, we use computer simulation to realize these goals and, so
far, have achieved significant benefits. The application area, which is product assembly, is
an area where almost anyone can benefit from the use of simulation.
Harrell−Ghosh−Bowden: I. Study Chapters 5. Getting Started © The McGraw−Hill
Simulation Using Companies, 2004
ProModel, Second Edition

120 Part I Study Chapters

The Problem
AST Research Inc., founded in 1980, has become a multibillion-dollar PC manufacturer.
We assemble personal computers and servers in Fort Worth, Texas, and offshore. For a long
time, we “optimized” (planned) our assembly procedures using traditional methods—
gathering time-and-motion data via stopwatches and videotape, performing simple
arithmetic calculations to obtain information about the operation and performance of the
assembly line, and using seat-of-the-pants guesstimates to “optimize” assembly line output
and labor utilization.

The Model
In December 1994 a new vice president joined AST. Management had long been commit-
ted to increasing the plant’s efficiency and output, and our new vice president had experi-
ence in using simulation to improve production. We began using ProModel simulation as a
tool for optimizing our assembly lines, and to improve our confidence that changes pro-
posed in the assembly process would work out in practice. The results have been signifi-
cant. They include
• Reduced labor input to each unit.
• Shortened production cycle times.
• Reduced reject rate.
• Increased ability to change assembly instructions quickly when changes become
necessary.
Now when we implement changes, we are confident that those changes in fact will improve
things.
The first thing we did was learn how to use the simulation software. Then we attempted
to construct a model of our current operations. This would serve two important functions: it
would tell us whether we really understood how to use the software, and it would validate
our understanding of our own assembly lines. If we could not construct a model of our
assembly line that agreed with the real one, that would mean that there was a major flaw
either in our ability to model the system or in our understanding of our own operations.
Building a model of our own operations sounded simple enough. After all, we had an
exhaustive catalog of all the steps needed to assemble a computer, and (from previous data
collection efforts) information on how long each operation took. But building a model that
agreed with our real-world situation turned out to be a challenging yet tremendously edu-
cational activity.
For example, one of our early models showed that we were producing several thousand
units in a few hours. Since we were not quite that good—we were off by at least a factor of
10—we concluded that we had a major problem in the model we had built, so we went back
to study things in more detail. In every case, our early models failed because we had over-
looked or misunderstood how things actually worked on our assembly lines.
Eventually, we built a model that worked and agreed reasonably well with our real-
world system out in the factory. To make use of the model, we generated ideas that we
thought would cut down our assembly time and then simulated them in our model.
We examined a number of changes proposed by our engineers and others, and then sim-
ulated the ones that looked most promising. Some proposed changes were counterpro-
ductive according to our simulation results. We also did a detailed investigation of our test-
ing stations to determine whether it was more efficient to move computers to be tested into
Harrell−Ghosh−Bowden: I. Study Chapters 5. Getting Started © The McGraw−Hill
Simulation Using Companies, 2004
ProModel, Second Edition

Chapter 5 Getting Started 121

the testing station via FIFO (first-in, first-out) or LIFO (last-in, first-out). Modeling showed
us that FIFO was more efficient. When we implemented that change, we realized the gain
we had predicted.
Simulation helped us avoid buying more expensive equipment. Some of our material-
handling specialists predicted, based on their experience, that if we increased throughput
by 30 percent, we would have to add some additional, special equipment to the assembly
floor or risk some serious blockages. Simulation showed us that was not true and in prac-
tice the simulation turned out to be correct.
We determined that we could move material faster if we gave material movers a specific
pattern to follow instead of just doing things sequentially. For example, in moving certain
items from our testing area, we determined that the most time-efficient way would be to
move shelf 1 first, followed by shelf 4, then shelf 3, and so on.
After our first round of making “serious” changes to our operation and simulating them,
our actual production was within a few percentage points of our predicted production.
Also, by combining some tasks, we were able to reduce our head count on each assembly
line significantly.
We have completed several rounds of changes, and today, encouraged by the experience
of our new investor, Samsung, we have made a significant advance that we call Vision 5.
The idea of Vision 5 is to have only five people in each cell assembling computers.
Although there was initially some skepticism about whether this concept would work, our
simulations showed that it would, so today we have converted one of our “focused
factories” to this concept and have experienced additional benefits. Seeing the benefits
from that effort has caused our management to increase its commitment to simulation.

The Results
Simulation has proven its effectiveness at AST Research. We have achieved a number of
useful, measurable goals. For competitive reasons, specific numbers cannot be provided;
however, in order of importance, the benefits we have achieved are
• Reduced the reject rate.
• Reduced blockage by 25 percent.
• Increased operator efficiency by 20 percent.
• Increased overall output by 15 percent.
• Reduced the labor cost of each computer.
Other benefits included increased ability to explain and justify proposed changes to
management through the use of the graphic animation. Simulation helped us make fewer
missteps in terms of implementing changes that could have impaired our output. We were
able to try multiple scenarios in our efforts to improve productivity and efficiency at com-
paratively low cost and risk. We also learned that the best simulation efforts invite partici-
pation by more disciplines in the factory, which helps in terms of team-building. All of
these benefits were accomplished at minimal cost. These gains have also caused a cultural
shift at AST, and because we have a tool that facilitates production changes, the company
is now committed to continuous improvement of our assembly practices.
Our use of simulation has convinced us that it produces real, measurable results—and
equally important, it has helped us avoid making changes that we thought made common
sense, but when simulated turned out to be ineffective. Because of that demonstrable
payoff, simulation has become a key element of our toolkit in optimizing production.
Harrell−Ghosh−Bowden: I. Study Chapters 5. Getting Started © The McGraw−Hill
Simulation Using Companies, 2004
ProModel, Second Edition

122 Part I Study Chapters

Questions
1. What were the objectives for using simulation at AST?
2. Why was simulation better than the traditional methods they were using to achieve
these objectives?
3. What common-sense solution was disproved by using simulation?
4. What were some of the unexpected side benefits from using simulation?
5. What insights on the use of simulation did you gain from this case study?

CASE STUDY B
DURHAM REGIONAL HOSPITAL SAVES $150,000
ANNUALLY USING SIMULATION TOOLS
Bonnie Lowder
Management Engineer, Premier

Used with permission by ProModel, 1999.

Durham Regional Hospital, a 450-bed facility located in Durham, North Carolina, has been
serving Durham County for 25 years. This public, full-service, acute-care facility is facing
the same competition that is now a part of the entire health care industry. With that in mind,
Durham Regional Hospital is making a conscious effort to provide the highest quality of
care while also controlling costs.
To assist with cost control efforts, Durham Regional Hospital uses Premier’s Customer-
Based Management Engineering program. Premier’s management engineers are very
Harrell−Ghosh−Bowden: I. Study Chapters 5. Getting Started © The McGraw−Hill
Simulation Using Companies, 2004
ProModel, Second Edition

Chapter 5 Getting Started 123

involved with the hospital’s reengineering and work redesign projects. Simulation is one of
the tools the management engineers use to assist in the redesign of hospital services and
processes. Since the hospital was preparing to add an outpatient services area that was to
open in May 1997, a MedModel simulation project was requested by Durham Regional
Hospital to see how this Express Services area would impact their other outpatient areas.
This project involved the addition of an outpatient Express Services addition. The
Express Services area is made up of two radiology rooms, four phlebotomy lab stations, a
patient interview room, and an EKG room. The model was set up to examine which kind
of patients would best be serviced in that area, what hours the clinic would operate, and
what staffing levels would be necessary to provide optimum care.

The Model
Data were collected from each department with potential Express Services patients. The
new Express Services area would eliminate the current reception desk in the main radiol-
ogy department; all radiology outpatients would have their order entry at Express Services.
In fiscal year 1996, the radiology department registered 21,159 outpatients. Of those, one-
third could have had their procedure performed in Express Services. An average of 18 out-
patient surgery patients are seen each week for their preadmission testing. All these patients
could have their preadmission tests performed in the Express Services area. The laboratory
sees approximately 14 walk-in phlebotomy patients per week. Of those, 10 patients are
simple collections and 4 are considered complex. The simple collections can be performed
by anyone trained in phlebotomy. The complex collections should be done by skilled lab
personnel. The collections for all of these patients could be performed in Express Services.
Based on the data, 25 patients a day from the Convenient Care Clinic will need simple
X rays and could also use the Express Services area. Procedure times for each patient were
determined from previous data collection and observation.
The model was built in two months. Durham Regional Hospital had used simulation in
the past for both Emergency Department and Ambulatory Care Unit redesign projects and
thus the management team was convinced of its efficacy. After the model was completed, it
was presented to department managers from all affected areas. The model was presented to
the assembled group in order to validate the data and assumptions. To test for validity, the
model was run for 30 replications, with each replication lasting a period of two weeks. The
results were measured against known values.

The Results
The model showed that routing all Convenient Care Clinic patients through Express
Services would create a bottleneck in the imaging rooms. This would create unacceptable
wait times for the radiology patients and the Convenient Care patients. Creating a model
scenario where Convenient Care patients were accepted only after 5:00 P.M. showed that
the anticipated problem could be eliminated. The model also showed that the weekend vol-
ume would be very low. Even at minimum staffing levels, the radiology technicians and
clerks would be underutilized. The recommendation was made to close the Express Ser-
vices area on the weekends. Finally, the model showed that the staffing levels could be
lower than had been planned. For example, the workload for the outpatient lab tech drops
off after 6:00 P.M. The recommendation was to eliminate outpatient lab techs after 6:00 P.M.
Further savings could be achieved by cross-training the radiology technicians and possibly
the clerks to perform simple phlebotomy. This would also provide for backup during busy
Harrell−Ghosh−Bowden: I. Study Chapters 5. Getting Started © The McGraw−Hill
Simulation Using Companies, 2004
ProModel, Second Edition

124 Part I Study Chapters

times. The savings for the simulation efforts were projected to be $148,762 annually. These
savings were identified from the difference in staffing levels initially requested for Express
Services and the levels that were validated after the simulation model results were ana-
lyzed, as well as the closing of the clinic on weekends. This model would also be used in
the future to test possible changes to Express Services. Durham Regional Hospital would
be able to make minor adjustments to the area and visualize the outcome before imple-
mentation. Since this was a new area, they would also be able to test minor changes before
the area was opened.
The results of the model allowed the hospital to avoid potential bottlenecks in the radi-
ology department, reduce the proposed staffing levels in the Express Services area, and val-
idate that the clinic should be closed on weekends. As stated by Dottie Hughes, director of
Radiology Services: “The simulation model allowed us to see what changes needed to be
made before the area is opened. By making these changes now, we anticipate a shorter wait
time for our patients than if the simulation had not been used.” The simulation results were
able to show that an annual savings of $148,762 could be expected by altering some of the
preconceived Express Services plan.

Future Applications
Larry Suitt, senior vice president, explains, “Simulation has proved to be a valuable tool for
our hospital. It has allowed us to evaluate changes in processes before money is spent on
construction. It has also helped us to redesign existing services to better meet the needs of
our patients.” Durham Regional Hospital will continue to use simulation in new projects to
improve its health care processes. The hospital is responsible for the ambulance service for
the entire county. After the 911 call is received, the hospital’s ambulance service picks up
the patient and takes him or her to the nearest hospital. Durham Regional Hospital is plan-
ning to use simulation to evaluate how relocating some of the ambulances to other stations
will affect the response time to the 911 calls.

Questions
1. Why was this a good application for simulation?
2. What key elements of the study made the project successful?
3. What specific decisions were made as a result of the simulation study?
4. What economic benefit was able to be shown from the project?
5. What insights did you gain from this case study about the way simulation is used?

References
Banks, Jerry, and Randall R. Gibson. “Selecting Simulation Software.” IIE Solutions, May
1997, pp. 29–32.
Kelton, W. D. “Statistical Issues in Simulation.” In Proceedings of the 1996 Winter
Simulation Conference, ed. J. Charnes, D. Morrice, D. Brunner, and J. Swain, 1996,
pp. 47–54.
Pritsker, Alan B., and Claude Dennis Pegden. Introduction to Simulation and SLAM.
New York: John Wiley & Sons, 1979.
Schrage, Michael. Serious Play: How the World’s Best Companies Simulate to Innovate.
Cambridge, MA: Harvard Business School Press, 1999.