PROBLEM SOLVING

Problem solving is difficult to teach; it is more art than science. To learn problem solving, you must
do it. As in riding bicycle, explanations are helpful, but eventually you must do it yourself. You may fall,
that is usually necessary in learning to ride a bike.

A problem exists when what is happening differs from what should happen. What is happening is the
current state, and what should happen is the goal state. We know the current and goal states in well-
structured problems. “Finding the minimum of a quadratic function is an example. Other problems
are ill-structured problems that may not have a clear goal state, nor even a well-defined current state.
“Improving the quality of life is an ill-structured problem.”

Problems may be one-time problems or re-curing problems. “Deciding how many products from each
type to produce each month” is a re-curing problem, but determining “why a machine failed is one-
time problem”. Re-curing problems require continuing data collection, reports and other
infrastructure.

Every problem has a life span. Some problems must be solved quickly, but others are not so urgent.

Problems also have different impact. Problems solved should be worthy of the resources required to
solve them. If it costs more to do the study than it will save, do not do it. The problem impact should
determine the amount of effort we put into solving the problem.

Economic studies can justify solving some problems. Others are hard to justify solely on cost and profit.
Identifying all impacts, including those difficult to quantify, is critical to knowing which problems
deserve effort. Problems typically do not go away unless we do something to resolve them. This
intervention is called problem solving. We look for easy and quick solutions, but complex problems
often require complex solutions. Solving most well-structured problems is easy. “To find the minimum
of a quadratic function, we use first and second derivatives”. Solving ill-structured problems is not
obvious. A large part of problem solving is transforming ill-structured problems into well-structured
problems.

Ackoff discusses four approaches to a problem: Absolution, Resolution, Solution, and Dissolution.

Absolution ignores the problem and hopes it goes away, which is seldom a good approach.

Resolution finds an acceptable solution to the problem using common sense; resolution is usually
better than absolution, but the answer may not be a good one sometimes.

Solution uses quantitative and experimental methods to get the “best” answer under the current
conditions.
Dissolution redesigns the system to eliminate the cause of the problem. This approach if possible and
not too costly is the preferred one.

Van Gundy’s five conditions required to solve a problem:

1. A gap between the current state and the goal state,

2. An awareness of the gap,
3. Motivation to close the gap,
“The problem is important to someone and has impact and resources will be committed to
solve it.”
4. The ability to measure the size of the gap,
5. The ability and resources to close the gap.

If one or more of these conditions is missing, the successful solution of the problem is unlikely. The
person who has the problem or someone paid to solve problems could be the problem solver. In the
production area, the problem solver may be a manager, analyst, or industrial engineer. Often there
are several problem solvers working together. Because problem solvers are people, they are not
infallible. Personal values, biases, and judgement all affect the problem solving process. Whether a
problem exists or no is affected by a person’s point of view, but recognizing that bias exists should
minimize its impact.
The knowledge and experience of a problem solver also comes into play. If a person has more tools,
that person has a wider range of options for solving problems. Experience teaches which tools to use
in certain situations and even helps in inventing new ones and adopting old tools to new situations. If
a particular tool is not in your tool kit, you will not use it.
The general problem solving approach includes;
• problem identification,
o analyze the system,
o identify the problem,
o understand the problem,
o develop a model,
o verify the model,
• Generation of solutions,
o solve the model,
o choose a solution,
o verify and validate the solution,
o interpret the model solution to problem solution,
 • Implement the selected solution,
• Monitor the results of implementation for problem solving,

The first step in problem solving is trying to identify the problem by analyzing the system, which
includes identifying the problem owners, and together determining the problem mission. Assumptions
are made at this step for the uncontrolled environmental issues. After successfully identifying the
problem, the problem solver and owners must understand the problem which includes examining the
system within which the problem occurs, specifying problem characteristics including goals and
possibly made more assumptions. Validation ensures that the correct problem is to be solved at later
steps. If this step is unsuccessful, a return to problem identification will be necessary.

Once the problem solver and owners agree on the problem, a formal model is developed. Modelling
concepts and data availability determine a representation for the model. Then, boundaries,
constraints, objectives, and relationships are used to produce a formal model. Validate the model to
make sure it does what the problem solver intended to do. Failure at this step may require a return to
either problem identification or problem understanding.

After constructing the formal model and collecting data, an appropriate algorithm is used to solve the
model. Again, unsatisfactory results force a return to a previous step. Once, we have the model
solution, we interpret it considering the actual problem. Robustness, solution, validation, and
judgement lead to a problem solution. If necessary return to a previous step to resolve differences.

Finally, implement the solution. Implementation begins by proposing a solution. Once the solution is
accepted, resources are committed to solve the problem. Appropriate people are trained and the ne
solution is implemented in parallel with the old procedure. Be sure to monitor progress or continued
success.

PROBLEM IDENTIFICATION

Problem identification converts a “mess” into a simple problem statement. A problem never
recognized will never be solved. Ackoff believes that “it is better to get the wrong solution to the
right problem than to get the right solution to the wrong problem”. Note that, solutions to the wrong
problem are ignored, but wrong answers to the right problem create interest and are corrected soon
and used.
Two sources of problems are need and opportunity. The most important phase of problem
identification is to determine the problem mission. The mission is the overall purpose – what we want
to accomplish. The mission will later be translated into goals, and then objectives. Different missions
result in different solutions.

Defining the correct mission will prevent confusing symptoms with problems. problem identification
has much in common with a doctor making a diagnosis. The problem solver looks at symptoms and
determine the correct problem statement. To differentiate between symptoms and causes, Ohno
recommends asking “why?” several times. Once the right mission has been chosen, identifying the
problem is easier. Do not try to incorporate solutions into problem identification. Brain storming with
a group of knowledgeable people can help clarifying the problem.

Problem owners are people who must live with the solution. Continually involve the owners in the
solution process. Make the problem a common enemy so that he problem owners work with you
(problem solvers) rather than against you.

Often the owners of the problem only recognizes symptoms and do not recognize the problem. A good
diagnosis is necessary and requires a continuous dialogue with the problem owners. If you work on
well-structured problems, assumptions may not be necessary. However, relationships between various
parts of most problems are uncertain, which requires the problem solver to make assumptions about
them. It is very important to state assumptions explicitly.

Once you identify a problem, write down a “formal” problem statement. Include a one-or-two-
sentence description of the mission and a brief description of the current state and the goal state. Do
not include restrictions, list all assumptions.

UNDERSTAND THE PROBLEM

We must describe the system within which it occurs. Once the boundaries of the problem are
determined, the problem solver and owners identify problem characteristics. They also make the
problem mission more specific by identifying solution goals.

A system is a collection of interacting components; its function cannot be done by any single
component. Problems occur in the ay parts of the system interact with each other. So, we must
understand problems within the systems framework (perspective).
Analysis is one way to study a system: The system is taken apart and each component studied
separately to see how it works. Then, the knowledge of the components is combined to gain
knowledge about the system, which can usually tells us how a system works.

Synthesis is another way to look at a system. Synthesis vies the system as a component of a larger
system, and we try to explain the behavior of the larger system. Then, we can explain the original
system as it functions within the larger system. Rather than tell us how the system operates, synthesis
tells us why it operates as it does. When examining a problem within the systems perspective,
important questions to ask are who, what, why, when, where, and how.

The mission is our overall purpose, but the goal should be one or more accomplishments that will
lead to fulfilling the mission. The time frame in which the problem exists is important. If it is a one-
time problem, the solution must be given before the problem disappears or changes. Thus, the time
available to solve a problem determines the approach we take to solve it. Recurring problems often
require more resources. Reports, data collection, and other infrastructure needed for recurring
problems are more elaborate then for one-time problems. The characteristics and impact of the
problem determine how much time and effort can be spent for solving the problem.

Important problems with a long time frame are worthy of time and effort. However, if the potential
payoff is small, we should resort to “quick and dirty” solutions or not to solve the problem at all. Quick
and dirty methods may also be right for important problems with a short time horizon.

Our understanding of the problem is an abstraction of the real problem. If we understand the actual
problem, our abstraction should capture its important features. Validating understanding is not easy.
Try describing the problem to the problem owners. Critically question every part of the problem
description. It helps to have someone else critically question your understanding of the problem. Make
the problem statement more formal by developing a formal model.

A model is a representation of something. The problem solver uses available data, modelling
concepts, and assumptions to choose a type of model. Then, specific data and the problem
boundaries help generating an objective and constraints applicable to the model. A model is proposed
and structurally validated, that ensures the model work as it should.

Models can be formal or informal, qualitative or quantitative. They are used to test an alternative,
predict the behavior of a system, determine the best of many alternatives, or explore what-if-
questions. We can usually test more alternatives with a model than we could by direct
experimentation.
Iconic models are physical representations that usually have different scales than what they represent.
These models are easy to explain, because they look like the real thing.

Analog models behave the same way the system does but do not look like the system.

Symbolic models are an important class of models. As the name implies, the system is represented by
symbols. This class includes graphical, tabular, and mathematical models. Systems too complicated for
tables and graphs use more complicated models.

Simulation models that use relationships between components, are symbolic models.

Mathematical models capture relationships between components using mathematical equations.

Linear programming formulations are typical mathematical models.

Data are used to validate assumptions, estimate model parameters, and validate models and are
usually collected at every step of the problem solving process. Availability of data may dictate the
type of model used. Data are not information; information describes or explains data. Data are used
to get information. Do not collect data for the sake of collecting data.

Company records are probably the best source of data. People are another source of data. People,
problem owners are source of data, as are customers, vendors, and even competitors.

Interviews and questionnaires are two ways to elicit information from people, but the questions asked
should be thought out so that the data will provide useful information. Governments and trade
associations often provide data. Research reports and regulatory publications are also sources of data.

The cost to get data will vary greatly depending on the source. Data must reflect the physical
phenomenon, which we call data integrity. Data must be accurate, and the accuracy required should
match the problem.

No matter where data come from, do not trust the data without verifying it yourself. If not verified
by observation or another source, consider the data questionable. Plan to use sensitivity analysis on
the values of the data when you generate solutions.

Modelling is an art, there is no recipe to tell how to do it.