1994 INTERNATIONAL SYSTEM DYNAMICS CONFERENCE
A Framework for Problem Solving Using Systems Dynamics Modelling
with a J Mi it Teck
Tom Ryan
School of Engineering Management
University of Cape Town
South Africa
Abstract
This paper reflects on a successful problem solving exercise that was carried out in a textile factory
using a systems approach together with a Japanese hniq The systems ap
used was Systems Dynamics Modelling and the Japanese management technique was CEDAC (see
dix). The paper h that problem solving can be viewed as a group learning
process and that this approach offers a useful ‘k for initiating and per
improvement. The degree of success is largely determined by the degree of learning that takes place.
Problem - Solving Methodologies, page 193
1994 INTERNATIONAL SYSTEM DYNAMICS CONFERENCE
A Framework for Problem Solving Using Systems Dynamics Modelling with
ad M. t Techni
P ) 4
1.0 Introduction
There seems to be a growing resistance to the fads that have dominated the popular management press
over the past decade. An i ition amongst iti is beginning to concede that
effective organization and management needs to go far beyond the ‘quick fix’. A few themes have
proved their value by surviving the management ‘fashion show’. These themes tend to acknowledge the
lexity of ization and rather than oversimplifying it. This is done by
highlighting the value of systems thinking (such as System Dynamics Modelling) and the importance of
managerial and organizational learning in the process. Research into the economic success of Japanese
organizations suggest similar ch istics in Japanese (Nonaka, 1988).
A large part of the management is concerned with problem solving; with changing a less desirable
behavioral pattern to one that is more desirable. Einstein said that it is not possible to solve a problem
in the same consciousness that created it (Wheatley, 1993). This implies that complex managerial
problem solving requires a change of i A change of. i is a change in the way
a problem situation is viewed. Such a change involves a learning process. The development of a
cybernetic view of an organization is a useful way of achieving this.
2.0 A Cybernetic View of Organizational Complexity
As Systems Dynamics Modelling (SDM) has its roots in the science of cybernetics it is useful to
develop a cyt ic view of izational lexity and then show how SDM can be used to deal
with the complexity. Norbert Weiner, the father of cybernetics sees it as the ‘science of information and
control in animals and machines’ (Weiner, 1948). W Ross Ashby, another pioneer in the field sees of
cybernetics as functional and dealing with all forms of behavior that are regular, determinate, or
reproducible (Ashby, 1964). Stafford Beer developed a powerful model for designing and building
viable i based on cyt ics, the Viable Systems Model (Beer, 1979). He defines
cybernetics as the science of effective organization. This paper sees cybernetics as a useful way raising
organisational r and thus i ing the i of the problem solving process.
The cybernetic approach offers a number of concepts useful in
of which the following are important when using SDM in a problem solving process: feedback,
emergence, self organization and variety (Clemson, 1984).
24 l a eeiianal
An organization is seen as a system of interacting parts. The complexity of a system is determined by
the interaction of the following factors (Schoderbek et al 1980):
1 the number of parts that make up the system
2 the number of interactions between these parts
3 the number of attributes of these parts
4 the degree of organization within the system
Problem - Solving Methodologies, page 194
1994 INTERNATIONAL SYSTEM DYNAMICS CONFERENCE
These can be seen as either static or dynamic. It is the dynamic factors, the interactions between the
parts, that dominate the behavior of organizations and are the concern of SDM. The most important
interactions affecting behavior have a feedback structure which gives rise to two important
characteristics of complex systems: and self ization. Cyt ics also introd the
concept of variety to explain and measure lexity and to offer an +h for dealing with it.
2.2 Feedback
Traditional science has thought of causality in simple linear terms, ie causes and effects were thought of
as X causes Y and Y causes Z. This could continue as far as you might want to trace it. The start of the
systems movement in the 1940's began to see causality as circular i.e. X influences Y and Y influences
Z and Z in tum influences X (Clemson, 1984). This circular causality is better known as feedback and
is the foundation of control theory. There are two types of feedback: negative feedback and positive
feedback. Negative feedback is goal seeking and works to negate disturbances experienced by the
system. Positive feedback is growth orientated and works to amplify disturbances. Feedback structures,
made up of positive and negative feedback loops, are useful for explai the ig
of complex system behavior:
at Many of the feedback loops that influence organizational behavior are not obvious. Some are
dormant under certain conditions and only become active when these conditions change. Any
managerial action that does not appreciate this has counter intuitive results.
2 For a range of inputs, ouput is largely determined by feedback rather than the input. Under
these conditions increased management intervention has little or no effect on organizational
behavior.
3 With insight into the feedback loop that infl izational behavior
management is in a far better position to take meaningful action. Lack of this insight explains
many of the di P din iP!
Different parts, attributes and interactions can result in many different behavioral patterns. The possible
number of patterns increases exponentially as the number of parts increases.
23 Emergence
An important concept developed from the systems movement is that of emergence which states that the
system as a whole exhibits properties that are not evident in its parts. These emergent properties’ of the
system are due to the interaction among the parts. These properties often produce unintend:
which can either promote or counter the purpose of the system. In a sense the interactions are co-
producers of the emergent properties. The presence or absence of an interaction influences the
emergent property. Culture and norms are examples of emergent properties of social systems.
2.4 Self Organization
All complex systems are capable of organizing themselves to some degree through the interaction of its
parts. The structural and behavioral patterns evident in complex systems are the result of such
organization. This organization happens throught the mutual adjustment of the parts of the system in a
process of inual i This self izing process is purposeful and is not ily
aligned with the overall purpose of the system. The 80/20 phenomenon can be seen as a demonstration
of the self organizing process, ie. 20% of the system produces 80% of its output.
Problem - Solving Methodologies, page 195
1994 INTERNATIONAL SYSTEM DYNAMICS CONFERENCE
Some of these resulting patterns are stable and others are not. If a system moves from a stable pattern to
one that is unstable, it will tend to gravitate back or move to another stable pattern. Any management *
action that attempts to move an organization into an unstable pattern will be resisted. A system can only
be held in an unstable pattern through high levels of resource utilization. Often, this includes coercion.
Alternatively, insightful intervention can change an unstable pattern into a stable one.
2.5 Variety
Variety is a measure of the complexity of a system. It reflects the number of distinguishable patterns a
system can assume. As the number increases, the variety of the system increases.
is largely d with ie maintaining those patterns that keep the variables
that reflect organizational performance within desired limits. To do this, management needs to be able
to cope with the variety of the organization, ie it needs to be able to respond appropriately to the
different patterns the organization can assume. The variety of management is measured in terms of the
number of ways it is able to respond to the variety of the organization. Therefore effective management
requires a measure of variety that is close to or equal to the variety of the system it has to manage. In
cybernetic terms management needs to achieve requisite variety. For any system to operate, no matter
how well or how poorly, requisite variety has to be achieved either by design of by default. An
important determinant of managerial effectiveness is the degree to which management can influence
how and where in the organization requisite variety is achieved. The implications of this concept for
management are enormous.
Organizations need to be designed and built to achieve requisite variety where it is most beneficial.
Organizations operate in high variety environments and are managed by low variety people. The
variety of management is largely limited by human information processing capability (Miller, 1963;
Simon , 1967). Many of the probl plaguing ization and can be explained in
terms of requisite variety. Stafford Beer's Viable Systems Model is an approach to achieving requisite
variety through planning and design, SDM can be seen as an approach to achieving requisite variety
through learning.
3.0 An Application of Systems Dy ics Modelling
We have found that a crucial starting point when using SDM in a complex problem solving process
such as those associated with socio-technical systems is to recognize:
1 that it is an approach to learning about, gaining insight into, and organizing your thinking about
the behavior of complex systems, and
2 that it is not an attempt to truct the total objective reality ding the problem.
It is primarily concerned about identifying plausib! ies for i desirable behavior. The
assumption is that this behavior is caused by the underlying structure of the situation. The term
structure is used to describe organizational structure, processes, norms, values, culture, and the general
pattern of resource depl within the izati Insights into structure make it possible to
identify plausible interventions that would change undesirable behavior.
Since its development in the early sixty's SDM has gone through a number of developments and has
been defined in numerous ways. A more recent definition is (Wolstenholme, 1990 p3):
Problem - Solving Methodologies. page 196
1994 INTERNATIONAL SYSTEM DYNAMICS CONFERENCE
‘A rigorous method of qualitative description, exploration and analysis of complex systems in terms of
their process, and ies; which facilitates quantitative
simulation modelling and analysis for the design of system structure and control.’
This definition is useful when dealing with socio-technical systems as long as you recognize that you
are not trying to describe, model or simulate a total objective reality. It is also important to note that
both the qualitative and itative aspects are emph
3.1 Background to the Case
The case concerns a medium sized textile manufacturer in Cape Town. The textile and clothing
industry as a whole in South Africa has been under severe economic pressures over the past five years.
The petiti of this particul was tt d as a result of excessively long
production lead times and late deliveries. We were approached by the manufacturer to assist them in
dealing with this problem. After some analysis this problem was set as a research topic for a masters
student (Van Veen, 1993). This paper reflects on the process of enquiry that followed from a learning
perspective.
3.2 The Problem Solving Process
The general approach to problem solving that we have developed involves two distinct phases. The first
is an immersion phase which involves a ‘soaking in' period during which the enquirer attempts to get an
in depth insight into the different perspectives held by the stakeholders involved in the problem
situation. The success of the problem solving process depends toa large degree on how well this phase
is done. In the second phase plausible imp} are loped based on the learning
that emerged from the first phase. It is often not necessary to explicitly select the most appropriate
strategy as the participants in the process tend to gravitate towards a shared understanding of the most
appropriate way forward.
The process started off with a q analysis of the per p facing late
deliveries and long lead times. This required an immersion into the actual situation through some form
of i ion with the : production staff and Our experience is that
SDM is not very useful in the early part of this phase, particularly when the stakeholders have no
knowledge of the process. We have found Soft Systems Methodology (Checkland, 1981) far more
useful during this immersion phase. In this case we were dealing with ‘practical’ managers at the ‘coal
face’ and therefore decided to use a Japanese management technique presented in a language they
understood. The technique used 1s known as Cause-and-Effect Diagram with the Addition of Cards
(CEDAC). The technique is described in detail in Appendix A. This is an enhancement of the classical
cause and effect diagram that has proved so useful in Quality Circles. CEDAC proved to be a powerful
group process for identifying. interpreting and communicating the issues contributing to the problem.
Once the final CEDAC representations (i.c. the immersion phase) of the problems was complete it was
possible to create a graphical SDM of the situation. This was seen as a plausible structure that produced
the undesirable behavior of the organization. It consists of feedback loop structures involving resource
flows and information flows. It provides a quali of the i ions between the parts
of the organization that give rise to the problem. An example of one of the feedback loop diagrams
created in this case study is illustrated in Fig. 1. The development of this diagram involved the group in
arich learning process and it provided a useful starting point for speculating about the determinants of
the system behavior and possible strategies for improvements. ihe process of developing the SDM
significantly enhanced the group's itivity of the and how it would
influence the success of any intervention. They were able to hypothesis more ‘intelligently’ as to which
strategies offered the most promise
Problem - Solving Methodologies, page 197
1994 INTERNATIONAL SYSTEM DYNAMICS CONFERENCE
aleve.
|
ice c) ay
Garen Sites l
eenscwarr = Quay Surecces?
Fig 1 Production feed backloop diagram
The next stage was to examine the quantitative behavior of the key systems variables over time and to
examine the validity and sensitivity of the model. This required a quantitative simulation of the SDM.
This was done using the Dynamo computer software. One of a number of strategies investigated was
the effect of changing the cycle time of various processes on the behavior of the system. The graph in
Fig 2 illustrates the effect of changing the cycle time of a particular process, on the key performance
variables.
Managerial learning is iderably ent d when the de of the particip in the process
increases. Management confi tie in SDM is significantly increased when the key variables reflect
historic behavior, both qualitatively and roughly quantitatively, when simulated over the same period.
Fig 3 compares the actual production output and the simulated output over the same period.
Considering that the group was not attempting to represent reality but to increase their understanding,
these results are close enough to make them feel comfortable with the process.
4.0 Results
Once the group had completed both the CEDAC and the SDM exercise it decided rather than implement
major changes to the system, it would focus on and ing relevant i As
an initial step it ped a series of per which reflected actual performance. These
consisted of two groups: L orientated and i A
CEDAC communications office was also established which would be accessible to all departments. Its
main purpose would be to gather and per ‘ion for decision making. The
result was a pret ing system which gave management a clear picture of
the current status of the key per » areas in the ion. This enabled them to significantly
improve their decision making. Furthermore it offered an opp ity for all stakeholder to
in driving the improvement process.
Problem - Solving Methodologies, page 198
1994 INTERNATIONAL SYSTEM DYNAMICS CONFERENCE
NUMBER OF ORDERS PRODUCED
9.2—ade-t ALEAD(S. ,9.2) 1: sentomsiontinn ORBIN(ZA. -2B.)
a 8 -B—B—8— LOELC1.,1.8)° 5 ~—e—pme=GLEAN(G.7,7.1)
7.4
8.37, WEEE
1.6 < Lm —— >
in
2. Quotep feno tine
6.9 ae
6.8
Vv NZ na (wee
6.7 8. 13. 2B. 38. a. ‘SB.
TIME
ead
Fi
10 2
TIME PERIOD
[= DYNAMO —&=— ACTUAL
Fig 3 A Comparison of Actual and Simulated Production
Problem - Solving Methodologies, page 199
1994 INTERNATIONAL SYSTEM DYNAMICS CONFERENCE
Over a and i
improvements were reported:
period of seven months the following performance
1 Average production lead time improved by 30%
2 Average ‘on time’ delivery improved by 55%
3 Average time for late delivery improved by 50%
5.0 Reflection
This improvement was gained with relatively few formal structural changes to the system. This begs
the question "what change then, caused these improvements?" One obvious change was the
in and it ion flows. What started out as a problem solving activity
tumed out to be a rich learning exercise for the participants and the organization. The answer to this
question seems to rest in understanding the managerial learning process.
The idea of a learning organization has been a recurring theme in the management literature of the past
five years. The idea can be traced back a lot further than that, to a little known paper "The Enterprise as
a Learning System’ by Reg Revans (Revans, 1982). In this paper he articulates the concept of a learning
organization and the symbiosis between work and learning. You, as manager can not create
meaningful organizational change unless you change as well, ie unless you go through a learning
process. He contends that effective managerial learning takes place in the context of work and through
interaction with fellow managers. The process involves groups of managers, who are facing Teal
problems, getting together and asking new and p ions about their probl
emerge from the process and the managers come away more competent in dealing with such problems.
Argyris has further loped the idea of izational learning with their concepts of single and
double loop learning (Argyris, 1982). The initial objective of this exercise case was single loop
learning, ie error detection and correction which is the domain of traditional problem solving. What
actually happened was double loop learning ie learning that changes current mental models, norms and
values. This involves deeper inquiry and asking new and This i
management's ability to deal with complexity.
In this exercise the CEDAC and the SDM exercises complemented each other in taking the group to a
higher level of consciousness through double loop learning. This enabled them to see and deal with
probl that were ad ly affecting 1 perf which they were not previously
aware of. In cybernetic terms the interactions that took place in the problem solving process resulted in
the emergence of double loop learning. Requisite variety was achieved at a more beneficial level within
the organization in the following way:
1 by moving toa higher level of i had d its own variety, and
2 by introdi the perfc and the CEDAC communications room the level of
self organization was increased in a positive way, resulting in an attenuation in the variety flow
in the organization as a whole.
This exercise has d d the value of problem solving as a group learning process and
managing it as such. The value can be further enhanced by adopting a systems approach such as
cybernetics and SDM in particular. When dealing with practical managers the process can be facilitated
by introducing proven Japanese management techniques.
Problem - Solving Methodologies, page 200
1994 INTERNATIONAL SYSTEM DYNAMICS CONFERENCE
References
Argyris, Chris. (1982). Reasoning, Learning and Action. Jossey Bass, San Francisco
Ashby, W Ross. (1964). An Introduction to Cybernetics. Methuen, London.
Beer, Stafford. (1979). The Heart of the Enterprise. Wiley, Chichester.
Checkland, P. (1981). Systems Thinking, Systems Practice. Wiley, New York.
Clemson, B. (1984). Cybernetics: A new Management Tool. Abacus, Turnbridge Wells.
Miller, G. (1967). The magical number, seven plus mins two. In Miller G., The Psychology of
Communication: Seven Essays. Basic Books, New York.
Fucuda, R. (1988) ial Engi ing. Productivity Press
Nonaka, I. (1988). Creating Organizational Order Out of Chaos: Self Renewal in Japanese Firms.
California Management Review, Spring, 1988.
Schoderbek, M C et al (1980). Ce I i i Business
Publications, Dallas.
Simon, H A. (1974). How Big is a Chunk? Science 183, Feb 8, 1974, p482-488.
Van Veer, H. (1993). The Need for S ic Compatibility in Approach
Unpublished Masters Thesis, University of Cape Town.
Wheatley, M J. (1992). Leadership and the New Science. Barrett-Koehler Publishers, San Francisco.
Weiner, N. (1948). Cybernetics. Wiley, New York.
Wolstenholme, E F (1990). Systems Enquiry: A Systems Dynamics Approach. Wiley,
Chichester.
Appendix
CEDAC is an acronym for Cause and Effect Diagram with the Addition of Cards. It is an enhancement
of the classical cause and effect diagram (also known as the fishbone or Ishikawa diagram). It is a
‘whole system’ rather than simply a diagram and works in the following way:
1 A problem area is identified.
2 The variable representing the behavior of the problem is plotted on a control chart together with
a target value which is set as an objective.
3 Causes of the problem are identified by operators, supervisors and management.
4 Each cause is written on an individual card.
Problem - Solving Methodologies, page 201
1994 INTERNATIONAL SYSTEM DYNAMICS CONFERENCE
5 Causes are grouped into similar categories as in a cause and effect diagram and attached to the
CEDAC diagram.
6 Possible solutions to the problems, which may be suggested by any of the actors, are written on
different coloured card and attached to the relevant cause.
An example of A CEDAC diagram in illustrated in Fig. Al
queity identified at rght. (Ona speniic quaiily of Be penduct,
(Coraei chen tor a specitic quality.
ie Ape
ig AN cot 9 a
CAUSES EFFECT
Fig Al An Example of a CEDAC Diagram
CEDAC is a ication tool for and the shopfloor for icating objectives and
ways of achieving them. It also visible monitors the progress towards these goals. It facilitates policy
deploy and p It between imp ‘ies and business objectives.
Problem - Solving Methodologies, page 202
