SYSTEM DYNAMICS MODELING FOR
LONG RANGE STRATEGIC PLANNING
Merle McKenzie
Jet Propulsion Laboratory
4800 Oak Grove Drive
Pasadena, California
ABSTRACT
A system dynamics model of a major telecommunica-
tions network has been developed to support
managers in the function of long range strategic
planning. Application of system dynamics to the
strategic planning area was found to be, in some
respects, quite unique. The article discusses this
type of application in the areas of model require-
ments, sponsorship, scope, development, and review.
In the area of requirements, it was found that a
system dynamics model developed to support long
range strategic planning should be quite broad in
scope, must satisfy a potentially large Bommonity
of planners, yet also must pass the review 0:
tactical planners as well. A baseline-model
approach is proposed as an effective way to satisfy
these requirements. Guidelines for the modeler are
proposed for obtaining sponsorship, for avoiding
pitfalls in the model development process, and for
interacting with model users and reviewers. The
baseline-model approach, coupled with the guide-
lines, has been found to work quite effectively
within one organization to support long range
strategic planning.
SYSTEM_DXYNAMICS..MODELING_FOR
LONG_BANGE. STRATEGIC. PLANNING
A system dynamics model of a major telecommunications
network has been developed at the Jet Propulsion Laboratory.
Its purpose is to support the management function of long
range strategic planning for the roughly $100 million per year
network. The application of system dynamics modeling to the
area of strategic planning is not a subject heavily
represented in the literature, and the team developing the
above Telecommunications Network (TN) model found that such
modeling is in some respects quite unique, The following
sections describe these unique aspects as they are found in
the areas of: requirements for system dynamics models in the
long range strategic planning environment (Section I);
sponsorship for such models (Section II); model scope (Section
III); model development (Section IV); and interaction with
users and reviewers (Section v).
i Requirements
According to George A. Steiner, the strategic planning
process is the formulation and implementation of strategy.[1]
He further refers to Peter Drucker for the definition of
strategy formulation as:
254
"the task of thinking through the mission of the
business, that is, of asking the question ‘what is our
business and what should it be?! This leads to the
setting of objectives, the development of strategies and
plans, and the making of today's decisions for
tomorrow's results. This clearly can be done only by an
organ of the business that can see the entire business;
that can make decisions that affect the entire business;
that can balance objectives and the needs of today
against the needs of tomorrow; and that can allocate
resources of men and money to key results."[2]
Drucker's definition adequately captures the purview of
the long range strategic planning organization for which the
above model was developed. Implicit within this definition
are the following characteristics of strategic planning
organizations:
1. The viewpoint of the managers is quite broad,
encompassing up to the entire system to be planned.
2. The planning horizon is correspondingly long, In
the case of long range strategic planning of the
author's experience, the planning horizon is ten to
fifty years.
3. The amount of planning work involved is substantial
and, thus, a number of managers can be heavily
involved in the planning process. This was indeed
the case for the TN model where the group of
managers involved: in strategic planning included up
to fifteen individuals.
4, A strong connection. may exist between the strategic
plan and the tactical plan for detailed
implementation, The managers of both strategic and
tactical concerns will thus be quite interested in
any tool used to support long range strategic
planning.
Any successful system dynamics model developed for a
strategic planning organization must deal effectively with the
above organizational characteristics. The author's group
found that the long range strategic planning environment
and organizational characteristics uniquely affect the
modeling effort in. three ways:
1. . It is quite diggicult to limit the focus of the
model to one or two questions. As stated by Jay W.
Forrester; "The model must reflect the point of
view of the person whose questions are to be
answered by the model experimentation."[3] In the
case of long range strategic planning, the view-
point is quite broad-and the persons many. This
will be discussed further in Section III.
255
2. There may be a substantial number of potential
users of the model.
3. There may also be a good number of people who will
review the model, even though they are not users.
Thus, the requirements of the model include: that it be
useful for long range strategic planning, that it be quite
broad in scope, that it satisfy a substantial community of
strategic planning users, and yet that it also pass the review
of tactical planners as well.
II. Model_sponsorship
As stated by Steiner;
"Years ago, the managerial emphasis in the typical
corporation was on operations. A major question for
management was how to use effectively those scarce
resources at its disposal in producing goods and ser-
vices at prices consumers were willing to pay.
Today, efficient use of scarce resources is still a
commanding concern of managements of all organizations,
but today, because of a turbulent and rapidly changing
environment, the ability of an organization to adapt
properly to environment, internal and external, is
becoming more critical in survival."(4]
Thus, in the terms of A. Porter, the recent emphasis in
organizational planning has moved from "optimization" to
“adaptation,"[5] and this change is in response to an
increasingly dynamic environment.
As most proponents of system dynamics would agree, this
need for adaptive (or strategic) planning is eminently
suitable for application of system dynamics, Yet, some
managers will question the ability of any model to capture a
quickly changing environment, much less address the equally
dynamic company plans, An effective strategy for a modeler to
pursue in this case is as follows:
1, Find the strategic planner who is at least willing
to grant the benefit of the doubt to the ability of
system dynamics to cope with such a situation. Use
this person or person(s) as the foundation of
support,
2. Build additional support slowly. Once the model is
developed and approved, use it to first address
questions that are of medium, rather than extreme
concern. Once confidence in the model is
256
established with respect to these medium-concern
issues, the community of managers will be more
likely to seriously apply the model to higher-
impact planning issues.
It is of interest to note that an organization being
subjected to major upheavals is not likely to have the manage-
ment time to devote to developing a new system dynamics model,
even though such could be of extreme value, Yet, if the
organization already has such a model, a few enhancements
might very well enable the manager to model the effect of
recent perturbing changes under various policy scenarios,
This can be a major selling point to a set of prospective
sponsors,
III. Model scope
It has been stated numerous times in the system dynamics
literature that a system dynamics model must be developed to
address a single issue or question for the model to be focused
and for the effort to be productive. John D. W. Morecroft
claims: "The point of departure for any strategy support
modeling project is a clear definition of the strategic
problem at issue... ."[6] Jay W. Forrester states: "The
first guiding principle is to state a clear purpose. One
should at first identify the problem, the phenomenon, or the
question to which the model is to be addressed,"[7]
The opposite extreme of a single-issue or single-
problem model is the general purpose model, of which Edward B.
Roberts says:
". . + most [corporate] models were not built to solve a
problem; they were instead built merely to represent (or
simulate) a corporation. If you build general-purpose
corporate models, you should assume in advance that you
will achieve little or no implementation of results.
You must build models that are motivated by a real
problem that exists."[8]
Unfortunately, a single-issue model, if not impossible,
is not as productive as a more general model when applied to
the area of long range strategic planning, The reasons are as
follows:
1. There are numerous managers likely to be involved
in a model designed to support long range strategic
planning.
2. These managers can have quite a broad view of the
system to be modeled.
257
Each manager has one or more issues that need to be
addressed, and the overlap among managers is not
total. In fact, most planning organizations
delegate planning areas to different managers both
to accomplish more and to gain the advantage of
multiple viewpoints.
Two common questions asked by a strategic planner
of a system dynamics model are: How does the real
system work, and what behavior does the system
produce? Yet, these are the very questions that the
single-issue modeling approach singles out to be
avoided.
Ultimately, credibility in the model and belief in
its usefulness can not. be easily achieved with a
Single-issue model. The multiple users and
reviewers of the model will claim that such a model
is too limited, addresses issues unimportant to
them, is too expensive per result obtained, and is
too incomplete. Even if the model is built for a
single, strategic planner with a single issue in
mind, to implement the strategy effectively that
planner must have cooperation of the others in the
area. They too must be convinced of the model's
completeness and veracity.
It is not impossible to build a model broadly
satisfactory to a strategic planning organization in a single-
issue fashion ~- it is just less likely. Basically, the model
must be broadened one issue at a time, with critics at the
door each time a recommendation is made. The author's group
came to the conclusion that in such a mode, the model would
very likely be shelved before its full potential could be
demonstrated.
There is actually evidence in the literature supporting
such a conclusion, Roberts states: ". .. the problem or
opportunity you select has to be seen as important to the
‘client' of your project, whom I define as the individual or
small group whose approval is needed for change to be
implemented."(9] In the case of strategic planning, a single
issue satisfying this criterion is not common. Roberts even
notes that after proceeding with a model built in the single-
issue mode, a problem may occur in which a recommendation on
the issue has unexpected effects on another part of the
system, He states that the question must be asked: "Will
this recommendation impact other systems or other parts of the
system?"[10] This is a question quite likely to be posed by
critics of a single-issue model at the point of recommen-
dation, These critics are likely to be managers of other
parts of the system who were not overly involved in the
258
AL
development of the single-issue model. Even Jorgen Randers,
who warns of the “tendency to ramble due to lack of an
explicit goal,"{11] recommends that if the complexity of the
model must be reduced, that the modeler should "reduce the
amount of detail (depth), rather than scope (breadth).
oe M112]
An alternative to the single-issue modeling procedure is
proposed as follows:
1. Develop a more general model of the system, where
the system is defined as the entity planned by the
strategic planning organization. This more general
model, which can be referred to as the baseline
model, should incorporate the variables and
parameters of interest to the strategic planners
(thus capturing the system reference mode) and
specifically address: How does our system operate;
and, given the existing set of conditions,
policies, and plans, how does it behave? This
baseline model does not have to be huge, just
comprehensive. ‘The TN model uses roughly 600
DYNAMO instructions,
2. During the development process, maintain a list of
questions the various managers do want answered.
This list will help the modelers incorporate the
proper parameters, and when prioritized, the list
will supply a roadmap for the enhancement phase.
3. Subject the baseline model to critical review.
Testing the ability of the baseline model to
address some of the managers questions will be a
test of its completeness. Yet managers will not
expect the baseline model to answer all questions.
4, Enhance the model, once approved, to address other
questions.
The advantages of the above procedure are respectable.
The baseline model can describe the real system and system
behavior. In baseline form, it can thus be used to identify
areas in which managers of the system disagree, provoke
discussion of strategic planning issues, and stimulate
strategic thinking in new areas by virtue of its breadth. If
the modelers are even somewhat careful, the baseline model can
also address significant managerial problems even before
enhancement. When the baseline model is done, the
enhancements require less time because the system structure
already exists and is approved. In this mode, the model is
less likely to be used just once and shelved, and as a model
of the whole system, it becomes a reference point for new
strategic planning. Enhancements are built and the model
259
13
becomes a thriving, evolving representation of the dynamic
system that it models.
All, however, is not roses. The founder(s) of the model
must be willing to wait longer for the baseline model than
they would for a single-issue model. This process can only be
expedited to the degree that the many necessary people can be
interviewed, that the separate observations and perceptions
can all be combined in a single consistent model, that the
data can be obtained, and that the reviewers can all be
brought together as required. Yet, the result is worth the
effort in the case of a system dynamics model for long range
strategic planning.
Iv. Model_Development
In the area of model development, there are a few guide-
lines that, while true for any system dynamics model, are also
especially relevant to a strategic planning model. The first
is a point noted by Roberts, that "To the maximum degree
1. Consistent is used here to mean that the model presents a
single, rather than multiple, representation of the
system. The baseline model can of course include system
inconsistencies if such exist in the real world.
4
possible, the work should be executed by knowledgeable in-
house employees."[13] Following this procedure was indeed
found to be quite effective, after the initial requisite
training in the system dynamics technique. The second guide~
line, also noted by Roberts, is that “The effective model will
contain a level of detail sufficient to demonstrate the
problem system, and sufficient to persuade participating
management of the model's adequacy."[14] 4H. B. Weil agrees,
saying"... clients are more comfortable with and confident
in a model that they consider ‘realistic.' Since these
attitudes are an absolute requisite for successful implementa-
tion, we are generally very accommodating to client desires
for more detail."[15] In the case of long range strategic
planning, the “problem system" mentioned by Roberts may be the
entire system being managed, and the level of detail required
may well be the baseline model as described in Section III.
In addition to the above, there are warnings issued by
modelers in other areas that are equally applicable to
modeling for long range strategic planning. Even with the
baseline model approach in mind, as proposed in Section III,
the following are common ways in which the modeling effort can
get sidetracked into interesting, but potentially deadly
paths:
260
15
1. Building the model to answer tactical rather than
strategic questions. ven though the tactical
planner may have some review authority over the
model, it should be made clear that the model is
designed for strategic, not tactical, use, as they
are quite different in detail, boundaries, and time
horizon.
2. Overdevelopment of the model sector(s) best known
to the modelers. This is extremely difficult to
combat, as it is most certainly a part of human
nature.
3. Overdevelopment of the model sector(s) best known
by the model's strongest supporters. Use of
system-level reviews helps to minimize this
tendency.
4, Starting a new system dynamics modeling effort
before completing the current one. An excrucia-
tingly obvious trap, but nevertheless quite
Powerful when a new system appears on the horizon,
just begging for some system dynamics analysis.
The modeler can most effectively deal with these sirens
of the modeling world by being aware of their existence and
wary of their song.
16
v. Users_apd_Reviewers
As noted above, using system dynamics to support long
range strategic planning may involve the review and/or
approval of the baseline model by quite a few individuals.
These individuals can be roughly categorized as follows:
1. Those funding the effort.
2. Those whose approval is required before a recommen-
dation can be implemented,
3. Those whose review is required as an input to those
approving in 2.
4, Those who are interested for their own reasons and
whose good will is helpful, if not absolutely
necessary.
Clearly, most effort will be spent on satisfying the
first two categories of individuals; however, none should be
neglected or alienated. Even more challenging, these multiple
users and reviewers will undoubtedly have multiple and
disparate views of: the system, current plans, current
policies, strategy changes, effects of strategies, and
evaluation of effects, This is probably inherent in the
nature of strategic planning and the managing of a complex
system. Yet, it is difficult to please everyone,
261
The author's group has found a few techniques to be of
use in interacting with users and reviewers during this
Process.
1.
They are as described below.
Use the same review process and documentation as
employed for other strategic planning efforts. The
users and reviewers will be most comfortable with
this known approach, and if approval is achieved,
the model's recommendations have a greater chance
for successful implementation,
Within the above review format, be sure to include
background on those items peculiar to system
dynamics, as is well discussed in the literature.
Employ a varying review audience: use the full
user-reviewer audience when you want broad comment
and critique; and use a more limited audience for
review of specific parts of the model. Similar to
the process of gradual validation testing,
confidence in the model can be built with a series
of iterative reviews by slightly varying audiences.
Be ready to respond to the query: "But how can you
Possibly test such a model?" Keeping a list of the
18
standard system dynamics tests, and the status of
the model with respect to each, has been found to
be quite effective.2
Be ready to compare the system dynamics technique
to standard econometrics techniques, It is not
uncommon for someone familiar with modeling in
general to impugn your model for those very
difficulties that system dynamics was created to
alleviate.
Do not be discouraged in the need to repeat your-
self with respect to the model and the technique.
The user may need to repeat views to you too in the
Process of evolving a model in the environment of
multiple users and reviewers, A combination of
Working Memoranda, as suggested by R. G. Coyle[16]
viewgraph presentations, viewgraph reviews, and
individual conferences are effective in supporting
good communications,
This technique was suggested by Dr. Peter C. Gardiner,
University of Southern California.
262
Once the baseline model is reviewed and approved, the
model can be applied by either the modeler or the user to
answer policy questions. As stated earlier, some questions
will require enhancement by the modeler, Yet with a fully
discussed and approved baseline model, the policy analysis
proceeds more smoothly. The dialectic method of strategy
analysis and selection, as proposed and well supported by John
D. W. Morecroft, is quite applicable in this stage. In his
words, the model will be "a vehicle for extending argument and
debate . . ."{17] to arrive at a new or changed system policy,
VI. Summary
The application of system dynamics to the area of long
range strategic planning has some unique aspects. Such
modeling will involve a large audience, members of which will
contribute to the model, review the model, and use the model.
To adapt to this larger audience, an approach is proposed in
which a more general model of the system, called the baseline
model, is developed and approved. This baseline model is then
selectively enhanced as necessary to address more user
questions.
263
20
This approach, together with quidelines and experiences
outlined above, has been found to work quite effectively
within one organization to support long range strategic
planning.
ca
REFERENCES
George A. Steiner, Strategic_Planping. What Every
Nanager_Must Know (New York: The Free Press, 1979),
P. 6.
Peter F. Drucker, Managementi_Tasks. Responsibilities,
Practices (New York: Harper & Row, 1974), p. 611.
day W. Forrester, "Modeling the Dynamic Process of
Corporate Growth," Reprint from Proceedings_of_the_IBM
Scientific Computing Sympsium_ on Simulation Models_and
Gaming (Dec. 7-9, 1964), p. 33.
Steiner, p. 5.
A Porter, “The Process of Adaptation and Its
Implications For Management," Canadian_Aeronautics_and
Space Journal (Vol. 15, Feb. 1969), p. 43.
10.
21
REEERENCES (continued)
John D. W. Morecroft, "Strategy Support Models" (Sloan
School Working Paper WP 425-83: July, 1982), p. 11.
Jay W. Forrester, "Modeling the Dynamic Process of
Corporate Growth," p. 23.
Edward B. Roberts, "Strategies for Effective Implemen-
tation of Complex Corporate Models," Edward B, Roberts,
ed., Managerial Applications_of_Systen_Dynanics
(Cambridge, Massachusetts: The MIT Press, 1981), p. 78.
Edward B. Roberts, “Strategies for Effective
Implementation," p. 78.
Edward B. Roberts, “Strategies for Effective
Implementation," p, 82.
Jorgen Randers, “Guidelines for Model Conceptuali-
zation," Jorgen Randers, ed., EBlements_of the_System
Dynamics_Method (Cambridge, Massachusetts: The MIT
Press, 1980), p. 136.
12.
13.
14.
15.
16.
17,
22
REEERENCES (continued)
Jorgen Randers, "Guidelines for Model Conceptuali-
zation,” p, 138,
Edward B, Roberts, "Some Insights Into Implementation,"
Edward B. Roberts, ed., Managerial_Applications_of
System_Dypamics (Cambridge, Massachusetts: The MIT
Press, 1981), p. 156.
Edward B, Roberts, “Some Insights Into Implementation,"
p. 158.
H, B. Weil, “Approach for Achieving Implemented
Results," Jorgen Randers, ed., Elements_of the System
Dynamics Method, op. cit., p. 284.
R. G. Coyle, Nanagement_System_Dynapics (London: John
Wiley & Sons, 1977), p. 363.
John D. W. Morecroft, "Strategy Support Models," p. 8.
264
