System. Dynamics as a Tool for Corporate Planning
Erich Zahn
Corporate Modelling and Planning Research Group
Betriebswirtschaftliches Institut
Universitat Stuttgart
1, Introduction
Corporate modelling in general and System Dynamics modelling in
particular have now a history of more than two decades. Despite
this fact impacts on the corporate planning process have not
been very satisfactory. The reason is that in many cases system
dynamics models (as well as other tynes of corporate models) had
not been constructed, validated and implemented adequately for
managerial use. They did not provide the. information support
which is needed in order to make the necessary decisions in
the various phases of a complex planning process that has a
lot to do with major changes in markets, products, production
processes, technologies, governmental regulations etc. Here,
formal decision rules as used in operational planning are
impractical in most cases.
In corporate modelling we have observed almost the same pitfalls
as we have experienced with management information systems.
Failures in potn areas ao indicate that wherever non-routine
decisions have to be made generalized solutions will be very
limited. Instead ad hoc usable decision support devices seem
to be a much more effective approach. With the help of such
a decision support system the appropriate information can be
generated and fed directly into the planning process. The kind
of demand for information that does exist in most planning
situations requires models that are problem-oriented,
S28)
-2-
sufficiently detailed, and easy to develope and to adjust.
Their structure should be understood by the manager and reflect
his own way of thinking. Our research indicates that this type
of model has to be developed within an intensive dialogue
between model builder and model user. If this modelling
process is orgdnized and carried out well the result might
be an experience-generalized model which is. really needed in
strategic planning and which therefore is much more likely
to be accepted by managers, because it seems to enhance managers
‘thinking processes'.
In this paper we will discuss how the potential of system
dynamics ‘models in supporting corporate ‘planning could be
realized more effectively. Our remarks on model conceptualization,
vaiiaation ana impiementation wili pe pased on two modei-projects.
Both of these are carried out with a system dynamics model that
in each case does represent a real world problem in a concrete
firm.
2. A System Dynamics Model for Operational Planning
a) The problem to be studied with the model
The first model deals with an operational planning- problem; the
simultaneous planning of marketing logistics at a German shoe-
producer. The model is used in order to gain. better insights
into tne complex “saies-inventory-manufacturing".
In reality this complex is characterized by a rather strong
seasonal development of order entries and deliveries on the
one side (as illustrated in figures la+b) and the need for
continuous production in order to realize high capacity
utilization rates on the other side.
=3-
-3-
a
(pa|
eal]
me OU
(CT tale Ta Taga tr [erected] Tamar [rire
Figure la: Observed seasonal dynamics of order flow
Ul
ss
rey
va Se ]
(aarp TT ior] rt Test Serene ottober[aovemter[Oexenber] Jamar [Febru]
Figure 1b: Observed seasonal dynamics of delivery flow
Orders are differentiated in those for the account and in
those that have to be delivered within a short time. Seasonal
dynamics of orders of the first category - their entry, backlog
and delivery-are represented in Figure lc.
-4-
(Cit Tipe TT ors [iret Titi [tober [vane [zener [Taresr [rebar]
Figure Ic: Observed seasonal dynamics of orders for
the account
The stock of finished goods in its yearly avarage lasts for
50 work days (see figure 1d). Resulting inventory costs are
about 1.5 Mill. DM per year.
Cua apt TT ioe Tots Tigo st [iriober [Resto [tecente [Tancar_[fetrat
Figure 1d: Observed dynamics of inventory (finished goods}>
“5s
Sales effectiveness during season significantly depends on
low delivery delays. Operational goals of the firm are a
stock of finished goods that last for 20 days (seasonal compo-
nant excluded),a delivery delay of 24 hours for goods that
are ordéred for current sales, production in large lot sizes,
. and a continuous production output.
b) Objectives of the model
The particular objectives of the model are:
- to identify areas within the firm and its markets that seem to be
relevant in order to understand the investigated problem;
- to analyse changes in systems behaviour caused by distur-
bances in the market, and
- to formulate policies for improving the real world systems
behaviour in terms of performances.
c) The structure of the model
The model structure includes the production process which is
devided into three phases, capacity, material and man power
control,as well as all relations of the firm to its labour,
machine-toole,and material-markets, that are relevant for
operational planning. The basic flow diagram is illustrated
in figure 2.
d) The behaviour of the model
After validating the model with real world data and experiences
from middie level and upper level managers the model was used
for several tests, including a constant demand, a sudden jump
-6-
investment.
capacity
utiiizatio
divestment
man powe
capacity capeeity
Ss
Bndproduction Fe-Productis
inventory
stage 1
inventory’
—l goods
orders
wennaee” marketing
saSDsTsen="baliches
ae. 1 |
backlog decision S°/very
ayailability Liman MM
inventory
stage 2
serveral
information
flows
produc> short term
‘success
Figure 2: Basic Causal diagram of the model.
sds
Fa:
in demand, a large order, increasing and decreasing demand,
a flop in product innovation, a too optimistic sales forecast,
seasonal and shorter fluctuations in the flow of orders.
In figures 3a-f a selected number of test results are
illustrated,
lo 8 8
a tr oor
vt
Tog ind te
Lo}
seo ap
a ohooh eres
(iar peta Tor Tt Taogust “Senter [ortober [Ttvenber]oezenber[ ianuar [retrcar_]
x
Figure 3b: Response of production rates according to an
unexpected demand jump of 10%.
Figure 3a: Dynamics of important level variables with
stability conditions.
Ta OT [toranter[tesvebe | Tancar [Fetroat J
Figure 3c: Response of inventories according to an un-
expected demand jump of 10%.
a Ce
Figure 3d: Changes in manpower capacity according to an
unexpected demand jump of 10%.
(eae Tat Tea eT itt Tigut [septa [riober[evenber rvater] [rtrwr")
Figure 3e: Changes in additional work hours according
to an unexpected demand jump of 10%,
-10-
Figure 3¥: Development of short-term goals according to
an unexpected demand jump of 10%.
As a result of an unexpected demand jump of 10% total costs
will increase by 33.900,-- DM, where 15.600,-- DM are due
to more overtime working hours and 18.300,-- DM are due to
an increase of current assets.
Summarizing these test results two recommendations could be
given to the firm for improving the performances of its sub-
system "sales-inventory-manufacturing”:
First, simulated market conditions show that reserves in all
three inventory stages.seem high enough in order to allow
an average decrease of stocks by 15 to 20%. The situation
of reserves looks more comfortable at inventories for raw
materials and partly finished goods than for finished goods.
-it-
-ll-
Second, a shortening of the reaction time from now 20 to
15 days would lead to a significant improvement of the time-
variant behaviour. For operational planning such. an action
would mean that observed gaps in inventories have to be
eliminated faster. Similarly a speeding up of the planning
rhythm or a faster implementation of plans would result in
a better dynamic behaviour too. All policies for applying
the recommendations have been discussed with the management
in detail; it turned out that they are feasible.
-12-
3. A System Dynamics Model of Coal Dynamics
a) The problem and objective of the study
The second model deals with a strategic planning problem; the
decision to mine or to import coal, and to allocate coal to
different sectors of consumption:. steal~production, conversion
to electric power, and heating. The model=project which is a
consulting work-for a large West German coal company is still
ina pilot phase. It is intentea to use the model as an early-
warning instrument. In todays energy business conditions do
change fast and often very sudden. Therefore management needs
something that can help to forewarn of changing conditions
such as jumps in prices for crude ojiland gas, supply cuts,
import stops, modifications in EEC-contracts, problems con-
cerning the acceptance of nuclear energy, technological
substitutions etc. Informations about such kind of environ-
mental changes are crucial for formulating effective business
strategies: To give management decision support in this respect
is the-main objective of the model-project.
b) The structure of the model..and preliminary results
According to a broadly stated request.by our. customer the
model tries to represent all aspects of todays and tomorrows
coal business which seem to be relevant for strategic planning.
The basic structure of the model, illustrated in figure 4,
shows the main elements: a mining sector, a buying sector,
three sectors of coal consumption, connected by an imaginary
variable coal inventory and an allocation mechanism. The
Tatter is determined by demand structures in the various
consumption sectors on the one side and by long-term contracts
as well as. short-term reactions on the other side.
1
seenarioe scentzice scenario. |
|
T T T
1 \ 1 !
tooL- bi ii 4. 4d
Figure 4: Basic structure of the coal model.
-14-
Consumption sectors are modelled seperately as modules for
isolated operations. Besides transformation processes, demand
structures, and marketing relations they also-do represent
special governmental regulations and £EC-contracts. Inter-
faces are constructed in a way that allows easily a connection
of the different modules.
The model-project is at the end of a pilot phase, now. It's
result is a relative simple model which gives the following
performances:
- it captures the essential structure of the real world system
as seen by the management,
- it generates time-variant trajectories which fit roughly
to historical time series, and
- it enables preliminary analysis of policies.
A second phase of the model-project is planned. Here a system
of models with much more detail will be developed, as outlined
in figure 5.
4. Some thoughts about applying system dynamics models
In the two model-projects mentioned above we made experiences
similar to those mentioned inthe technical paper of James
Lyneis and in earlier publications by Edward Roberts (1973)
and Henry Weil. These experiences could be summarized as
follows: Firet, start with a relatively simple model which
captures the basic structure of the system studied!
Such a model helps modelling experts to learn about the systems,
and decision makers to become familiar with system dynamics
thinking and modelling. In both projects we developed what
Burpuezsuapun wajgoud o1uaxshs jo uoLqoauip
Ress es
RAG-
integration model
Submodels for
- allocation
= re-allocation
- financial planning
Figure 5: Planned model system.
MoH Mouy WaLqoud pa[teyep Jo UOLZD0ALp
“16
we call a demonstration model. It helped us to better
communicate and structure ideas about the problem. Especially
in the second project where we had to cooperate with a project
team of staff people and line managers from different
divisions and various backgrounds, this approach was very
fruitful. It helped us to interpret the various, often
fuzzy opinions about the system or its parts and to integrate
them into a holistic picture, all participants could agree
upon. Here it turned out to be very important that each
member of the project team finds his view of the problem
reflected sufficiently well. Otherwise he would probably
reject the study and participate not any longer. Form a more
technical point of view this procedure looks too time
consuming and not successful at all. It is sometimes
frustrating, indeed. For example, our first coal model
didn't look very much like a system dynamics model. But step
by step we could form it into one, and at the end of the pilot
phase the project members were ready to agree upon the structure,
the generated behaviour, and further steps for developing a
detailed model system.
Second, try to develope a model that corresponds in cetail
to the problems of the major parts of the organization
(divisions, strategic units, etc.)!
This seems to be a neccessary condition for a model that
the
really may support managers' minds and thus helps to cc the
strategic planning work.
To follow this demand a modular approach of modelling should
be used. It has several advantages. First of all it may help
to make model building more effectively and efficiently.
e7e
The reason is that modelling experts and specific decision
makers could work together more intensively. A lot of use~
less discussion normaly observed in larger project teams
might be avoided. A modular approach delivers managers
responsible for a division or a strategic unit a submodel
which could be used seperately or in connection with other
modules. Sometimes competition between division managers
is hard and therefore they will not feel very comfortable
to make their problems transparent for opponents. If this
is the case good data may only be collected if the division
manager could be convinced that he can run the model
exclusively with his own data.
Third, according to the last thesis it will be neccassary
for model acceptance that inputs and outputs of the mode
do correspond to available company data.
Last not least success of model applications will significantly
depend on appropriate management involvement in all phases
of the model building process from model conceptualization
to model implementation. Managers participating in model-
projects should come from those hierarchical levels and
functional areas where they are responsible for making and
executing decisions that are supposed to be supported by the
model.
