ADVANCED

MANAGEMENT
SCIENCE

Quantitative Approaches to
Decision Making
Course Content
• Introduction
• An Introduction to Linear
Programming
• Linear Programming: Sensitivity
Analysis and Interpretation of Solution
• Linear Programming Applications in
Marketing, Finance, and Operations
Management
• Advanced Linear Programming
Applications
Course Content
• Distribution and Network Models
• Integer Linear Programming
• Nonlinear Optimization Models
• Project Scheduling: PERT/CPM
• Inventory Models
• Waiting Line Models
• Simulation
• Decision Analysis
• ree
• Dynamic Programming
Course Content
• Multi-criteria Decisions
• Time Series Analysis and Forecasting
• Markov Processes
• Linear Programming: Simplex Method
• Simplex-Based Sensitivity Analysis and
Duality
• Solution Procedures for Transportation and
Assignment Problems
• Minimal Spanning Tree
• Dynamic Programming
INTRODUCTION
Introduction
• Management science, an approach to
decision making is based on the scientific
method and makes extensive use of
quantitative analysis.
• A variety of names exists for the body of
knowledge involving quantitative approaches
to decision making; in addition to
management science, two other widely
known and accepted names are operations
research and decision science.
• Today, many use the terms management
science, operations research, and decision
science interchangeably
 MANAGEMENT
SCIENCE IN
ACTION

Revenue Management at American Airlines*

*Based on Peter Horner, “The Sabre Story,” OR/MS
Today (June 2000).
 Revenue Management at American Airlines
• One of the great success stories in
management science involves the work done
by the operations research (OR) group at
American Airlines. In 1982, Thomas M. Cook
joined a group of 12 operations research
analysts at American Airlines. Under Cook’s
guidance, the OR group quickly grew to a
staff of 75 professionals who developed
models and conducted studies to support
senior management decision making. Today
the OR group is called Sabre and employs
10,000 professionals worldwide.
 Revenue Management at American Airlines
• One of the most significant applications
developed by the OR group came about
because of the deregulation of the airline
industry in the late 1970s. As a result of
deregulation, a number of low-cost airlines
were able to move into the market by selling
seats at a fraction of the price charged by
established carriers such as American
Airlines. Facing the question of how to
compete, the OR group suggested offering
different fare classes (discount and full fare)
and in the process created a new area of
management science referred to as yield or
revenue management.
 Revenue Management at American Airlines
• The OR group used forecasting and
optimization techniques to determine
how many seats to sell at a discount
and how many seats to hold for full
fare. Although the initial
implementation was relatively crude,
the group continued to improve the
forecasting and optimization models
that drive the system and to obtain
better data.
 Revenue Management at American Airlines
• Tom Cook counts at least four basic
generations of revenue management
during his tenure. Each produced in
excess of $100 million in incremental
profitability over its predecessor. This
revenue management system at
American Airlines generates nearly $1
billion annually in incremental
revenue.
PROBLEM SOLVING AND DECISION MAKING
PROBLEM SOLVING AND
DECISION MAKING
• Problem solving can be defined as the
process of identifying a difference
between the actual and the desired
state of affairs and then taking action to
resolve the difference. For problems
important enough to justify the time and
effort of careful analysis, the problem
solving process involves the following
seven steps:
Problem Solving Processes
1. Identify and define the problem.
2. Determine the set of alternative solutions.
3. Determine the criterion or criteria that will be
used to evaluate the alternatives.
4. Evaluate the alternatives.
5. Choose an alternative.
6. Implement the selected alternative.
7. Evaluate the results to determine whether a
satisfactory solution has been obtained.
Decision Making

• Decision making is the term generally

associated with the first five steps of the
problem solving process. Thus, the first
step of decision making is to identify and
define the problem. Decision making
ends with the choosing of an alternative,
which is the act of making the decision.
EXAMPLE
• Let us consider the following example of
the decision-making process. For the
moment assume that you are currently
unemployed and that you would like a
position that will lead to a satisfying
career. Suppose that your job search has
resulted in offers from companies in
Rochester, New York; Dallas, Texas;
Greensboro, North Carolina; and
Pittsburgh, Pennsylvania.
EXAMPLE
• Thus, the alternatives for your decision
problem can be stated as follows:
• 1. Accept the position in Rochester.
• 2. Accept the position in Dallas.
• 3. Accept the position in Greensboro.
• 4. Accept the position in Pittsburgh.
EXAMPLE
• The next step of the problem-solving process
involves determining the criteria that will be
used to evaluate the four alternatives.
Obviously, the starting salary is a factor of some
importance. If salary were the only criterion of
importance to you, the alternative selected as
“best” would be the one with the highest starting
salary. Problems in which the objective is to find
the best solution with respect to one criterion are
referred to as single-criterion decision
problems.
EXAMPLE
• Suppose that you also conclude that the
potential for advancement and the
location of the job are two other criteria of
major importance. Thus, the three criteria
in your decision problem are starting
salary, potential for advancement, and
location. Problems that involve more than
one criterion are referred to as multi-
criteria decision problems.
EXAMPLE
• The next step of the decision-
making process is to evaluate each
of the alternatives with respect to
each criterion. For example,
evaluating each alternative relative
to the starting salary criterion is
done simply by recording the
starting salary for each job
alternative.
EXAMPLE
• Evaluating each alternative with respect
to the potential for advancement and the
location of the job is more difficult to do,
however, because these evaluations are
based primarily on subjective factors that
are often difficult to quantify. Suppose for
now that you decide to measure potential
for advancement and job location by
rating each of these criteria as poor, fair,
average, good, or excellent. The data
that you compile are shown in Table 1
 EXAMPLE

You are now ready to make a choice from the available

alternatives. What makes this choice phase so difficult
is that the criteria are probably not all equally important,
and no one alternative is “best” with regard to all
criteria. Suppose that after a careful evaluation of the
data in Table 1, you decided to select alternative 3;
alternative 3 is thus referred to as the decision.
 Qualitative and Quantitative
Approaches
• The skills in the qualitative
approach are inherent in the
manager and usually increase
with experience.
• The skills of the quantitative
approach can be learned only
by studying the assumptions
and methods of management
science.
 Qualitative and Quantitative
Approaches
• A manager can increase
decision-making effectiveness
by learning more about
quantitative methodology and
by better understanding its
contribution to the decision-
making process.
 Qualitative and Quantitative
Approaches
• A manager who is knowledgeable in
quantitative decision-making
procedures is in a much better
position to compare and evaluate
the qualitative and quantitative
sources of recommendations and
ultimately to combine the two
sources in order to make the best
possible decision.
Reasons why a quantitative approach is
used in the decision-making process:
1. The problem is complex, and the manager
cannot develop a good solution without the aid
of quantitative analysis.
2. The problem is especially important (e.g., a
great deal of money is involved), and the
manager desires a thorough analysis before
attempting to make a decision.
3. The problem is new, and the manager has no
previous experience from which to draw.
4. The problem is repetitive, and the manager
saves time and effort by relying on quantitative
procedures to make routine decision
recommendations.
• Quantitative analysis begins once the
problem has been structured. It usually
takes imagination, teamwork, and
considerable effort to transform a rather
general problem description into a well-
defined problem that can be approached
via quantitative analysis.
• To successfully apply quantitative analysis to
decision making, the management scientist
must work closely with the manager or user of
the results. When both the management
scientist and the manager agree that the
problem has been adequately structured, work
can begin on developing a model to represent
the problem mathematically.
• Solution procedures can then be employed to
find the best solution for the model. This best
solution for the model then becomes a
recommendation to the decision maker. The
process of developing and solving models is the
essence of the quantitative analysis process.
Model Development
• Models are representations of real
objects or situations and can be
presented in various forms. For example,
a scale model of an airplane is a
representation of a real airplane.
Similarly, a child’s toy truck is a model of
a real truck. The model airplane and toy
truck are examples of models that are
physical replicas of real objects. In
modeling terminology, physical replicas
are referred to as iconic models.
• A second classification includes models
that are physical in form but do not have
the same physical appearance as the
object being modeled. Such models are
referred to as analog models. The
speedometer of an automobile is an
analog model; the position of the needle
on the dial represents the speed of the
automobile. A thermometer is another
analog model representing temperature.
• A third classification of models ncludes
representations of a problem by a system
of symbols and mathematical
relationships or expressions. Such
models are referred to as mathematical
models and are a critical part of any
quantitative approach to decision making.
• For example, the total profit from the sale
of a product can be determined by
multiplying the profit per unit by the
quantity sold. If we let x represent the
number of units sold and P the total
profit, then, with a profit of $10 per unit,
the following mathematical model defines
the total profit earned by selling x units:
P = 10x
• The purpose, or value, of any model is that it
enables one to make inferences about the real
situation by studying and analyzing the model.
• For example, an airplane designer might test an
iconic model of a new airplane in a wind tunnel
to learn about the potential flying characteristics
of the full-size airplane. Similarly, a
mathematical model may be used to make
inferences about how much profit will be earned
if a specified quantity of a particular product is
sold. According to the mathematical model of
equation P = 10x, we would expect selling three
units of the product (x = 3) would provide a profit
of P = 10(3) = $30.
• In general, experimenting with models requires
less time and is less expensive than
experimenting with the real object or situation.
• A model airplane is certainly quicker and less
expensive to build and study than the full-size
airplane. Similarly, the mathematical model
allows a quick identification of profit
expectations without actually requiring the
manager to produce and sell x units.
• Models also have the advantage of reducing the
risk associated with experimenting with the real
situation. In particular, bad designs or bad
decisions that cause the model airplane to crash
or a mathematical model to project a $10,000
loss can be avoided in the real situation.
• The value of model-based conclusions
and decisions is dependent on how well
the model represents the real situation.
The more closely the model airplane
represents the real airplane, the more
accurate the conclusions and predictions
will be. Similarly, the more closely the
mathematical model represents the
company’s true profit-volume
relationship, the more accurate the profit
projections will be.
• When initially considering a managerial
problem, we usually find that the problem
definition phase leads to a specific objective,
such as maximization of profit or minimization of
cost, and possibly a set of restrictions or
constraints, such as production capacities.
• The success of the mathematical model and
quantitative approach will depend heavily on
how accurately the objective and constraints
can be expressed in terms of mathematical
equations or relationships.
• A mathematical expression that describes the
problem’s objective is referred to as the
objective function.
• For example, the profit equation P = 10x would
be an objective function for a firm attempting to
maximize profit. A production capacity constraint
would be necessary if, for instance, 5 hours are
required to produce each unit and only 40 hours
of production time are available per week.
• Let x indicate the number of units
produced each week. The
production time constraint is given
by 5x ≤ 40 The value of 5x is the
total time required to produce x
units; the symbol ≤ indicates that
the production time required must
be less than or equal to the 40
hours available.
• The decision problem or question is: How
many units of the product should be
scheduled each week to maximize profit?
A complete mathematical model for this
simple production problem is:
• The x ≥ 0 constraint requires the
production quantity x to be greater than
or equal to zero, which simply recognizes
the fact that it is not possible to
manufacture a negative number of units.
• The optimal solution to this model can be
easily calculated and is given by x = 8,
with an associated profit of $80. This
model is an example of a linear
programming model.
• In the given mathematical model, the profit per unit
($10), the production time per unit (5 hours), and
the production capacity (40 hours) are
environmental factors that are not under the control
of the manager or decision maker. Such
environmental factors, which can affect both the
objective function and the constraints, are referred
to as uncontrollable inputs to the model.
• Inputs that are controlled or determined by the
decision maker are referred to as controllable
inputs to the model. In the example given, the
production quantity x is the controllable input to the
model. Controllable inputs are the decision
alternatives specified by the manager and thus are
also referred to as the decision variables of the
model.
• Once all controllable and uncontrollable inputs
are specified, the objective function and
constraints can be evaluated and the output of
the model determined.
• A flowchart of how controllable and
uncontrollable inputs are transformed by the
mathematical model into output is shown in
Figure 4 and Figure 5.
• Uncontrollable inputs can either be
known exactly or be uncertain and
subject to variation.
• If all uncontrollable inputs to a model are
known and cannot vary, the model is
referred to as a deterministic model.
• Corporate income tax rates are not under
the influence of the manager and thus
constitute an uncontrollable input in many
decision models. Because these rates
are known and fixed (at least in the short
run), a mathematical model with
corporate income tax rates as the only
uncontrollable input would be a
deterministic model.
• The distinguishing feature of a
deterministic model is that the
uncontrollable input values are known in
advance.
• If any of the uncontrollable inputs are uncertain
and subject to variation, the model is referred to
as a stochastic or probabilistic model.
• An uncontrollable input to many production
planning models is demand for the product. A
mathematical model that treats future demand—
which may be any of a range of values—with
uncertainty would be called a stochastic model.
• In the production model, the number of hours of
production time required per unit, the total hours
available, and the unit profit were all
uncontrollable inputs.
• The distinguishing feature of a stochastic
model is that the value of the output
cannot be determined even if the value of
the controllable input is known because
the specific values of the uncontrollable
inputs are unknown. In this respect,
stochastic models are often more difficult
to analyze.
DATA PREPARATION
Data Preparation

• Another step in the quantitative analysis

of a problem is the preparation of the
data required by the model.
• Data refer to the values of the
uncontrollable inputs to the model. All
uncontrollable inputs or data must be
specified before one can analyze the
model and recommend a decision or
solution for the problem.
• In the production model, the values of the
uncontrollable inputs or data were $10 per unit
for profit, 5 hours per unit for production time,
and 40 hours for production capacity.
• In the development of the model, these data
values were known and incorporated into the
model as it was being developed. If the model is
relatively small and the uncontrollable input
values or data required are few, the quantitative
analyst will probably combine model
development and data preparation into one
step. In these situations the data values are
inserted as the equations of the mathematical
model are developed.
MODEL SOLUTION
Model Solution

• Once the model development and data

preparation steps are completed, one
can proceed to the model solution step.
• In this step, the analyst will attempt to
identify the values of the decision
variables that provide the “best” output
for the model. The specific decision-
variable value or values providing the
“best” output will be referred to as the
optimal solution for the model.
Model Solution

• One procedure that might be used in the

model solution step involves a trial-and-
error approach in which the model is
used to test and evaluate various
decision alternatives. In the production
model, this procedure would mean
testing and evaluating the model under
various production quantities or values of
x.
• If a particular decision alternative does
not satisfy one or more of the model
constraints, the decision alternative is
rejected as being infeasible, regardless
of the objective function value. If all
constraints are satisfied, the decision
alternative is feasible and a candidate
for the “best” solution or recommended
decision.
• Through this trial-and-error process of
evaluating selected decision alternatives,
a decision maker can identify a good and
possibly the best feasible solution to the
problem. This solution would then be the
recommended decision for the problem.
• Although the trial-and-error solution
process is often acceptable and can
provide valuable information for the
manager, it has the drawbacks of not
necessarily providing the best solution
and of being inefficient in terms of
requiring numerous calculations if many
decision alternatives are tried.
• Model development and model solution steps
are not completely separable
• A simpler and perhaps more easily understood
model with a readily available solution
procedure is preferred even if the recommended
solution is only a rough approximation of the
best decision.
• Model testing and validation are
frequently conducted with relatively small
“test” problems that have known or at
least expected solutions.
• If the model generates the expected
solutions, and if other output information
appears correct, the go-ahead may be
given to use the model on the full-scale
problem.
• However, if the model test and validation
identify potential problems or
inaccuracies inherent in the model,
corrective action, such as model
modification and/or collection of more
accurate input data, may be taken.
Whatever the corrective action, the
model solution will not be used in
practice until the model has satisfactorily
passed testing and validation.
REPORT GENERATION
Report Generation
• An important part of the quantitative
analysis process is the preparation of
managerial reports based on the model’s
solution.
• The solution based on the quantitative
analysis of a problem is one of the inputs
the manager considers before making a
final decision.
Report Generation
• Thus, the results of the model must
appear in a managerial report that can be
easily understood by the decision maker.
The report includes the recommended
decision and other pertinent information
about the results that may be helpful to
the decision maker.
 End of
Session 1

Do you have any questions?