Models and Modeling - Powell and Baker
Models and Modeling - Powell and Baker
CHAPTER 1
INTRODUCTION
1
c01.qxd 6/30/03 5:06 PM Page 2
2 CHAPTER 1 INTRODUCTION
Mathematical models take many forms and are used throughout science, engineer-
ing, and public policy. For instance, a groundwater model helps determine where flood-
ing is most likely to occur, population models predict the spread of infectious disease,
and exposure-assessment models translate the impact of toxic spills. In other settings,
traffic-flow models predict the buildup of highway congestion, fault-tree models help
reveal the causes of an accident, and reliability models suggest when equipment may
need replacement. Mathematical models can be extremely powerful, especially when
they give clear insights into the driving forces behind a particular result.
Many of the models business analysts create are used in one-time decision problems.
A corporate valuation model, for example, might be used during merger negotiations to
make a go/no-go decision, but the model may never be used again. In other situations, a
one-time model might be created to evaluate the profit impact of a promotion campaign,
or to help select a health insurance provider, or to structure the terms of a supply contract.
One-time models are usually built by decision makers themselves, often under time pres-
sure. Managerial judgment is often used as a substitute for empirical data in such models,
due to time constraints and data limitations. Most importantly, this type of model involves
the user intensively, because the model is usually tailored to a particular decision-making
need. One major benefit of studying modeling is to make the use of one-time models much
more effective.
Decision-support systems are computer systems that tie together models, data, analy-
sis tools, and presentation tools into a single integrated package. These systems are
intended for repeated use, either by executives themselves or by their analytic staff.
Decision-support systems are used in research and development planning at pharmaceuti-
cal firms, pricing decisions at oil companies, and product-line profitability analysis at
manufacturing firms, to cite just a few examples. Decision-support systems are usually the
province of computer personnel, but they represent the routine use of what were once one-
time decision models. After a one-time model is established, it can be adapted for broader
and more frequent use in the organization. Thus, the models within decision-support sys-
tems may initially be developed by managers and business analysts, but subsequently
streamlined by information systems staff for a less intensive level of human interaction.
An additional benefit of studying modeling is to recognize possible improvements in the
design and operation of decision-support systems.
Embedded models are models that are contained within computer systems that per-
form routine, repeated tasks with little or no human involvement. Many inventory replen-
ishment decisions are made by automated computer systems. Loan payments on auto
leases or prices for stock options are also determined by automated systems. Routine real
estate appraisals may also be largely automated. In these cases, the models themselves are
somewhat hidden behind the software. Many users of embedded models are not aware of
the underlying models, but simply assume that somehow the “system” knows how to make
the right calculations. One ancillary benefit of studying modeling is to become more
aware, and perhaps more questioning, of these embedded models.
Models are useful not only in the business world, but also in the academic world in
which business analysts are educated. The modern business curriculum is heavily depend-
ent on models for delivering basic concepts as well as for providing numerical results. An
introductory course in Finance might include an option-pricing model, a cash-management
model, and the classical portfolio model. A basic Marketing course might include demand
curves for pricing analysis, a diffusion model for new-product penetration, and clustering
models for market segmentation. In Operations Management, we might encounter inven-
tory models for stock control, allocation models for scheduling production, and the
newsvendor model for trading off shortage and surplus outcomes. Both micro- and macro-
economics are taught almost exclusively through models. Aggregate supply-and-demand
curves are models, as are production functions. Most of the models used in education are
highly simplified, or stylized, in order to preserve clarity. Stylized models are frequently
used to provide insight into qualitative phenomena, not necessarily to calculate precise
numerical results. In this book, we frequently use models from business education as exam-
ples, so that we can combine learning about business with learning about models.
Modeling can benefit business decision making in a variety of ways. For one thing, it
can provide more timely information than alternative approaches. For example, while
extensive surveys could be used to determine the potential demand for a product, effective
c01.qxd 6/30/03 5:06 PM Page 4
4 CHAPTER 1 INTRODUCTION
modeling can often give useful bounds on the likely range of demand in far less time.
Modeling can also save costs: surveys are expensive to create, distribute, and process. An
effective modeler may be able to provide the same level of information at much lower
cost. Another way in which modeling reduces cost is that it allows us to make inexpensive
errors. Wind-tunnel tests are used in airplane design partly because if every potential wing
design had to be built into a full-scale aircraft and flown by a pilot, we would lose far too
many pilots. Similarly, using a model, we can propose ideas and test them in the model,
without having to suffer the consequences of bad ideas in the real world. Yet another ben-
efit of modeling is that it allows us to explore the impossible. Many companies have poli-
cies, procedures, or habits that prevent them from making certain choices. Sometimes,
these habits prevent them from discovering better ways of doing business. Modeling can
be used to explore these “impossible” alternatives, and if some of them prove attractive,
modeling can be used to convince the skeptics to try them out.
A final benefit from studying modeling is that it can improve business intuition. As
we have said, a model is a laboratory in which we perform experiments to learn. We can
usually learn faster from laboratory experiments than from experience in the real world.
With a model, we can try thousands of combinations that would take many years to test in
the real world. We can also try extreme ideas that would be too risky to test in the real
world. And we can learn about how the world works by simulating a hundred years of
experience, all in an afternoon.
Even among those who do not build models, skill in working with models is very
important. Most students eventually find themselves on a team charged with recommend-
ing a course of action. If these teams do not build models themselves, they often work with
internal or external consultants who do. Experience in building and analyzing models is,
in our minds, the best training for working effectively on problem-solving teams. People
who have not actually built a few models themselves often accept model results blindly or
become intimidated by the modeling process. A well-trained analyst not only appreciates
the power of modeling but also remains skeptical of models as panaceas.
We believe that modeling skills are useful to a very broad range of businesspeople,
from junior analysts without a business degree working in consulting to senior vice pres-
idents of start-ups who do their own analysis. Many recent MBAs have only a superficial
knowledge of these tools, because the typical MBA program still emphasizes passive
consumption of other people’s models rather than active model building. Thus, there is
considerable potential even among those who hold MBAs to improve their modeling
skills so they can become more capable of carrying out independent analyses of impor-
tant decisions. The only absolute prerequisite for using this book and enhancing this skill
is a desire to use logical, analytic methods to reach a better level of understanding in the
decision-making world.
This book is organized around the three skills we believe business analysts most need in
their modeling work:
■ spreadsheet engineering
■ modeling craft
■ management science and statistics
Spreadsheet engineering deals with how to design, build, test, and perform analysis
with a spreadsheet model. Modeling craft refers to the nontechnical but critical skills that
an expert modeler employs, such as abstracting the essential features of a situation in a
model, debugging a model effectively, and translating model results into managerial