1981 SYSTEM DYNAMICS CONFERENCE
Discussion Paper for the Plenary Session’ on
Industrial Applications
R.G.Coyle
Management Centre
University of Bradford
July 1981
176
Introduction
This paper assesses briefly what seems to be the state of
affairs with respect to practical industrial usage of System Dynanics.
This leads to a discussion of topics which, it is felt, should be
examined as objectively as possible in the hope that steps can be
agreed which vill lead to greater practical application,
kground
The original literature on System Dynamics used the nane
‘Industrial Dynamics"; the first published applications were to
industrial problems, and the subject vas developed in a major business
school. SD seems to have great potential in management analysis as
it seems; at any rate to someone who has used both SD and the other
techniques of management science, to have the following attributes:
1. It is couparatively easy to learn.
2. Tt requires no particular mathematical skill and can, therefore,
be used by people who understand the problems,
3. It is very flexible in use, so that the problem does not have
to be forced into a given mould.
4, Tt addresses very well the question of action to be taken as
time passes and recognises that decisions made now may merely
store up trouble for the future unless these decisions are
harwonised to the structure of the system and the exogenous
shocks it encounters.
Given those attributes, and the fact that practically 25 years
have elapsed since the first publications of SD, which is about the
same as the Lifetime of Linear Programming, one vould expect to
see SD widely used. The reality seems to be that
LW
5.
‘SD is taught in perhaps 5% of Business Schools (significantly,
Warvard seems to ignore it, for all practical purposes).
There is only one small consulting firm regularly using SD.
It is almost unknown for SD papers to appear in the leading
wanagement science journals.
The software for SD is not widespread.
Where industrial applications of SD do take place that seems
to be often on a one-off basis, and rarely le
Is to continuing
activity. -
For all of these statements, exactly the opposite would be true
for, say, 1jnear programing.
Reasons?
The reasons for this a1
many and various, but they may include
the folloving.
LL
3.
Perhaps SD is simply not as good as we think it is, and there are
better and easier ways of achieving the same result.
Because SD is not taught, it is not used, and therefore it is
not taught.
Many people who have never used SD, and some who have, think that
it bas been wildly over-sold.
To people who bave not used it, the DYNAMO syntax is seen as
clumsy and restrictive, and it is certainly true that DYNAMO
itself lacks many facilities vhich would be useful.
the
Managers are not interested in/longer-term solutions which are
produced by SD and merely want to know what to do next. Perhaps
176-a
they are right in that view, and feel that the world changes
so rapidly that longer-term solutions are invalid.
6. It is certainly the view of thie author that practitioners of
SD have done themselves a grave disservice by arguing that SD
is a profes
ion in its owm right which is somehow separate from,
and superior to, the rest of management science,
eful
ns
It is, perhaps, no accident that the years from, say, 1953 to
1973 were a period of comparative stability in managerial affairs and
that the main developments in applied management science vere in LP
(for optimising essentially stable systems), and in econometrics (for
predicting essentially stationary stochastic processes), The increasing
economic and social turbulence since the mid 1970's may increase
managerial awareness of the need for analyses of controllability (though
see item 5 in the list of reasons), On the other hand, there is
increasing emphasis in the management science/econometrice literature,
on papers which basically apply the mathematics of control theory,
ignoring the work already done in SD,
The Papers in the Plenary Session
The papers have been selected to demonstrate developuents in
industrial application, which appear to be particularly important.
1. Rigorous descriptions of systems as they actually are rather
than vague ‘conceptualisation', whatever that means, of @
system capable of producing a reference mode of behaviour,
2. Considerable advances in equation-writing practice to enable
us to represent real systems as they are.
ney
176-b
7..Par more co-operation between different SD groups, and the
3. Demonstration that the validation process for SD mode
eradication of the Not Invented Here disease, which seems
done to a higher standard than is commonly believed.
to be epidemic among us.
4. Close attention to implementation, and the use of the model
to test managers’, rather than analysts’, theories about
the system.
5. Very beavy emphasis on the driven response of the system,
i.e. the firm in relation to the rest of the world, rather
than the firm seen as a closed process,
Development for the Future
It seems to the Chair that vhat is required for the future
is some or all of the following:
1. A far more modest approach to the presentation of the results
of SD analyses.
- Fever authoritative, or authoritarian, pronouncements on what
SD is, and how it should be done.
. The publication of more papers in the cainstream management
science literature, so that SD can come to be regarded merely
as part of the intellectual furniture.
Much wore care in the testing and debugging of models, and
wore effort put in to modelling the real processes of systems,
and far less into very broadly aggregated models.
5. Considerable development in the use ‘of appropriate control
theory algorithms, and their incorporation into the software.
Suitable use of optimisation facilities.
6. The development of a unified, accessible, cheap, software
package, rather than the present situation of two principal,
and several subsidiary, packages, at prices ranging from $150
to $10,000.
CC BY-NC-SA 4.0