D-4079-1 1

Market Growth as Influenced by

Capital Investment∗
Jay W. Forrester
Massachusetts Institute of Technology

The manager, by evolving the policies that guide decision making, designs a corporate
system. Such systems have a feedback loop structure that determines the growth and
stability of the enterprise. The complexity of these systems usually precludes intuitive
determination of how a policy change will affect the total system. A simulation model of
the feedback structure and policies allows one to try policy changes to see how the
system reacts. But to build such a model, one should be guided by principles of System
structure that help one to interrelate the diverse observations that arise from any real
system.

In this paper, structure is organized in a hierarchy. A closed system should be defined,

within which the difficulties are generated and solutions are to be sought. Within the
boundary, feedback loops form the system substructure. Feedback loops contain only
two types of variables - levels and rates - into which all aspects of the real system can be
fitted. The rates are the system policies and in turn contain component concepts - goal,
observed conditions, discrepancy, and action.

Growth and stagnation of a new product is taken as an example to show how the ideas
of feedback structure can explain a common occurrence in market behavior. Very often
early sales grow rapidly, only to level off even while the market demand continues to
rise, as shown, by competitors capturing an increasing share of the market. One cause is
found in the capital investment policy of the firm, as shown in a model involving
salesmen, market reaction to delivery delay, and capital equipment expansion.

Introduction to Systems To speak of systems implies a structure of interacting

functions. Both the separate functions and the interrelationships as defined by the
structure contribute to the system behavior. To describe a system, one must describe not
only the separate functions but their method of interconnection. To identify the structure
of a specific system, one should understand the fundamental nature of the structure
common to all dynamic systems.

A dynamic system is one which changes with the progress of time. The parts interact to
create a progression of system conditions. There is a basic structure common to all such
systems, whether they be the systems encountered in engineering, in management, in
economics, in nature, in psychology, or in any purposeful relationship of components.

• This paper was presented at the Ninth Annual Paul 0. Converse Awards Symposium, University of Illinois, April
13, 1967. It will appear in [4]. The computer runs in this paper were obtained using the facilities of Project
MAC. Work reported herein was supported in part by Project MAC, an M.I.T. research project sponsored by
the Advanced Research Projects Agency, Department of Defense, under Office of Naval Research Contract
No. Nonr— 4102 (01). Reproduction of this report in whole or in part is permitted for any purpose of the U.S.
Government.
• Originally published in the Industrial Management Review, MIT Sloan School, Vol. 9, No. 2, 1968. Reprinted
in: Forrester, Jay W., 1975. Collected Papers of Jay W. Forrester. Waltham, MA: Pegasus Communications.
284 pp.
D-4079-1 2

The theory of system structure will here be described and exemplified in terms of four
steps in a hierarchy:

A Closed boundary
1 Feedback loops
a Levels
b Rates
(1) Goal
(2) Observed condition
(3) Discrepancy
(4) Desired action

If one has a theory of structure and confidence that the structure is universal in its
applicability, such a theory greatly expedites the process of identifying and classifying
the available information about an actual system. It is here asserted that such a theory of
structure does exist, that it can be precisely stated, that it can be rigorously applied, and
that it is of major practical value in organizing knowledge. The four hierarchies in the
general theory of structure will now be discussed briefly and then illustrated by an
example.1

Closed Boundary In defining a system, we start at the broadest perspective with the
concept of the closed boundary. The boundary encloses the system of interest. It states
that the modes of behavior under study are created by the interaction of the system
components within the boundary. The boundary implies that no influences from outside
of the boundary are necessary for generating the particular behavior being investigated.
So saying, it follows that the behavior of interest must be identified before the boundary
can be determined. From this it follows that one starts not with the construction of a
model of a system but rather one starts by identifying a problem, a set of symptoms, and
a behavior mode which is the subject of study. Without a purpose, there can be no
answer to the question of what system components are important. Without a purpose, it
'is impossible to define the system boundary.

But given a purpose, one should then define the boundary which encloses the smallest
permissible number of components. One asks not if a component is merely present in
the system. Instead, one asks if the behavior of interest will disappear or be improperly
represented if the component is omitted. If the component can be omitted without
defeating the purpose of the system study, the component should be excluded and the
boundary thereby made smaller. An essential basis for identifying and organizing a
system structure is to have a sharply and properly defined purpose.

Feedback Loops as Building Blocks Inside the closed boundary one finds a structure
of interacting feedback loops. The feedback loop is the structural setting within which all
decisions are made. The feedback loop is a closed path. A decision is based on the
observed state of the system. The decision produces action which alters the state of the
system and the new state gives rise to new information as the input to further decisions.
The feedback loop implies the circularity of cause and effect, where the system produces
the decision which produces the action which produces change in the system. One has
not properly identified the structure surrounding a decision point until the loops are
closed between the consequences of the decision and the influence of those
consequences on future decisions.

1
For a more complete discussion of structure see [1]
D-4079-1 3

Level and Rate Variables Within the feedback loop we find the next lower hierarchy of
structure. To represent the activity within a feedback loop requires two and only two
distinctly different kinds of variables -the levels and the rates. The levels represent the
system condition at any point in time. In engineering, the level variables are often
referred to as the system state variables. In economics, the system levels are often
spoken of as stocks. The levels are the accumulations within the system. Mathematically
they are integrations.

The rate variables represent the system activity. The rate equations are the policy
statements in, the system which define how the existing conditions of the system
produce a decision stream controlling action.

The clear separation of system concepts into the two classes of variables - levels and
rates - has interesting and useful consequences. The level variables are the integrations
of those rates of flow which cause the particular level to change. It follows that a level
variable depends only on the associated rates and never depends on any other level
variable. Furthermore, in any system, be it mechanical, physical, or social, rates of flow
are not instantaneously observable. No rate of flow can depend on the simultaneous
value of any other rate. Rates depend only on the values of the level variables. If levels
depend only on rates and rates depend only on levels, it follows that any path through
the structure of a system will encounter alternating level and rate variables.

Policy Structure An important substructure exists within the equation that defines a
rate variable, A rate equation defining a rate variable is a statement of system policy.2
Such a policy statement describes how and why decisions are made. A policy statement
incorporates four components - the goal of the decision point, the observed conditions as
a basis for decision, the discrepancy between goal and observed conditions, and the
desired action based on the discrepancy.

A decision is made for a purpose. The purpose implies a goal that the decision process
is trying to achieve. The policy statement that determines a rate variable does so in an
attempt to bring the system toward the goal. The goal is sometimes adequately
represented as a constant objective; more often the goal is itself a result of the past
history of the system that has established traditions to guide present action. Whether or
not the goal is actually achieved depends on how the system as a whole responds to the
particular decision point. Usually, the competition for resource allocation results in the
system falling short of most of the goals. The goal at the particular decision point is
compared with the observed system condition as a guide to action.

One must distinguish observed conditions from the actual conditions of a system. A
system model must incorporate both actual and apparent system levels (the levels
describe the condition or state of the system). Where an important difference can exist
between what the system is and what it is thought to be (and these differences are
especially prevalent in the marketing sector of a company), one represents both, and
explicitly shows how the apparent states arise out of the true states. A decision can be
based only on the observed conditions, that is, the available information. Very often,
substantial deviations exist between the true conditions of a system and the observed
conditions. The discrepancy can arise from delay in recognizing changes in the system,
random error, bias in not wanting to believe what is visible, distortion, insensitivity, and
misinterpretation of meaning.

2
For a more complete description of a policy statement see [2], Chapter 10.
D-4079-1 4

The policy statement makes a comparison of the goal and apparent condition to detect a
discrepancy. The discrepancy may be in the form of a difference, a ratio, or some other
indicator of lack of agreement.

On the basis of the discrepancy, the policy describes the action to be taken.

A System Example in Marketing The preceding concepts of system structure will now
be illustrated in terms of a set of relationships often encountered in the growth of a new
product.

Marketing couples the resources of a company to the desires of the customers. As such
it represents the interface across which flow goods, services, money, and information.
But these flows across the boundary are a consequence of interactions within the
company, within the market, and between the two. Market dynamics can be understood
only in the context created by other company functions because these other functions
produce the variables with which marketing must deal.

As stressed above, one can identify a system only in terms of an objective. Here the
objective is to identify and to explain one of the systems which can cause stagnation of
sales growth even in the presence of an unlimited market. In particular, we deal here
with that system which causes sales stagnation, or even sales decline, to arise out of an
overly cautious capital investment policy. In this system inadequate capacity limits the
growth in product sales.

Figure 1 illustrates the scope of the system being considered. The closed boundary
surrounds the relationships shown. No other influences from the outside are necessary
for creating the sales growth and stagnation patterns which will presently be developed.

Within the closed boundary the system consists of interacting feedback loops as
illustrated in Figure 1. Three major loops are shown. Loop 1 is a positive feedback loop
involving the marketing effort here described in terms of hiring of salesmen. It provides
the driving power for sales growth. Only positive feedback loops can produce sustained
growth. A positive loop is one in which activity changes the condition of the system in
such a direction as to produce still greater activity. Assuming a favorable set of
conditions around the loop, here is a situation in which salesmen book orders followed by
product delivery which generates revenue which produces the sales budget which
permits hiring still more salesmen. In short, salesmen produce revenue to pay for the
further expansion of the sales effort.

However, Loop 2, on the upper right, involves delivery delay and sales effectiveness and
can make the product sufficiently unattractive that the sales loop is no longer able to
generate revenue greater than its current expenditures. The delivery delay in Loop 2 can
convert the salesmen-hiring in Loop 1 from positive-feedback growth behavior to
negative-feedback goal-seeking behavior. Negative loops are goal seeking and adjust
activity toward some target value. Here Loop 2 is a negative feedback loop and tends to
D-4079-1 5

1 Loop Structure for Sales Growth, Delivery Delay, and Capacity Expansion.

adjust the incoming order rate to equal the production capacity. It is common to think of
the order rate as determining the production capacity, but under many circumstances
production capacity is instead determining the order rate. This phenomenon takes place
within Loop 2. Orders booked increase the order backlog which increases the delivery
delay which makes the product less attractive and reduces the order rate. Were the
order rate to be sustained above the production capacity, the backlog and the delivery
delay would continue to increase until the product could no longer be sold.

Production capacity is determined in Loop 3. Here a very simplified capital investment

policy will be represented to keep the example within permissible size. The ordering of
new production capacity is a function of delivery delay only. Rising order backlog, as
indicated by delivery delay, is taken as an indication of inadequate capacity, and orders
for more capacity are placed. These orders, after an acquisition delay, add to the
production capacity. Loop 3 is a negative feedback loop which is attempting to change
D-4079-1 6

2 Salesmen-Hiring Loop with Sales Generating Revenue to Support Selling Effort.

D-4079-1 7

production capacity to adjust the order backlog to a value determined by a management

goal for proper delivery delay. As the delivery delay rises, production capacity is raised
to bring down the delivery delay.

These loops will be examined in turn to show their detailed structure and their behavior.
The flow diagrams and system equations define a complete simulation model of the
simplified company-market system so that the time sequences implied by the system
description can be computed and plotted.3

Salesmen-Hiring Loop The detailed structure of the positive feedback loop governing
the hiring and level of salesmen is shown in Figure 2.

In the flow diagrams, the level equations are shown as rectangles, as for salesmen in
this figure. The rate variables are shown by the valve symbol, as for salesmen hired.
The circles are "auxiliary" variables which are algebraically substitutable into the
following rate equations and are structurally part of the rate equation.

Considering that the auxiliary variables are part of the associated rate variables, we see
in Figure 2 the alternating rate and level substructure within a feedback loop. The
salesmen-hiring rate feeds the salesmen level. The salesmen level controls the orders
booked rate. Orders booked as a rate flows into the backlog level. The backlog level is
depleted by the delivery rate. The delivery rate is an input to the delivery rate average,
which is a level. (All averages are generated by an accumulation process and by both
mathematical form and structural location are necessarily system levels.) The delivery
rate average, being a level, feeds into the salesmen-hired rate.

In Figure 2 the "Salesmen Switch" SSW at the top of the figure has been put in to agree
with the specific equations in the Appendix. The switch allows activation or deactivation
of the loop in simulation model runs. For the purpose of this discussion it should be
considered in Position 1.

The positive feedback character of Loop 1 shown in Figure 2 gives this market system
its growth tendencies. With a sufficiently attractive product and a sufficiently high
fraction of revenue devoted to the sales budget, conditions are such that salesmen
produce orders booked which increase backlog which increases delivery rate which
increases the budget which increases the indicated salesmen that can be supported
which causes a salesmen-hiring rate which increases the number of salesmen.

Such a positive feedback loop has an exponential growth character as shown in Figure
3. Here, Loop 1 alone causes ever increasing growth without limit in the loop variables.
The growth rate depends on the delays around the loop and on the conversion
coefficients that determine loop amplification. Delays around the loop occur in the order
backlog, in the delivery rate averaging, which here represents the billing and collection
delay, and in the salesmen adjustment time SAT, which here represents the delays in
budgeting and the delay in finding and training salesmen. The value of 20 months for
SAT is probably shorter than correct for most systems. If any of these delays are
increased, the growth rate will be slower.

3
The detailed equations and their descriptions are given in the Appendix. Equation structure is described in [2].
Equation details conform to the DYNAMO compiler as described in [3].
D-4079-1 8

3 Unlimited Exponential Growth in Loop 1.

4 Initial Growth in Loop 1 Followed by Decline when Sales Effectiveness is Reduced.

The effect of changing the gain around the sales-hiring loop in Figure 2 can be shown by
making a large change in the sales effectiveness. If sales effectiveness is reduced, it
means that a given number of salesmen will book fewer orders and produce less
revenue and thereby support a smaller sales budget. It the sales effectiveness were
made small enough, a given number of salesmen would produce revenues too small to
support themselves. Under these circumstances the indicated salesmen would be less
than the existing number of salesmen and salesmen-hiring would become salesmen
reduction. Under such circumstances, the positive feedback loop would have been
converted to a negative feedback loop tending toward zero salesmen and zero activity.
This change in sales effectiveness is shown in Figure 4. Here conditions are as in Figure
3 until week 36, At week 36 the sales effectiveness has been reduced from 400
units/man-month (400 units per month sold by each salesman) to 100 units/man-month.
In other words, the imaginary condition has been created where the product
D-4079-1 9

5 Market Loop with Delivery Delay Determining Product Attractiveness.

D-4079-1 10

is four times harder to sell. Orders booked drop immediately but rise again by a small
amount because the number of salesmen is still increasing. After the time for the lower
order rate to propagate through the order backlog and the delay in delivery rate
averaging, the number of salesmen starts to decline, and, along with the declining
salesmen there is a corresponding decline in orders booked and in backlog. Figure 4 is
included to give a feeling for the behavior of the loop in Figure 2. The sudden change in
product attractiveness and the four-fold decrease in sales effectiveness would of course
not be expected in an actual system.

Market Loop In Figure 5, the major loop, Loop 2 from Figure 1, connects delivery delay
of the market, generates sales effectiveness, and influences the rate of orders booked. A
minor loop relates order backlog and production capacity to generate the delivery rate.

The delivery delay of a product is given approximately by the ratio of backlog to delivery
rate. In other words, the time to fill an order is indicated by how long the present delivery
rate will require to work its way through the present order backlog. The delivery delay
indicated, IDDI, is the ratio of the present backlog to the present short-term average of
the delivery rate, DRA. But this present condition of delivery delay, as implied by present
backlog and present delivery rate, ordinarily does not immediately reach the attention of
decision makers within the system. The delivery delay recognized by the company,
DDRC, is a delayed version of delivery delay indicated, DDI. The delivery delay
recognized by the company is an input to the production capacity ordering decision and
also forms the basis of delivery quotations to the market. The market takes time to
respond to changing delivery delay quotations and so a further delay intervenes before
the delivery delay is recognized by the market, at IDDRIVI. On the basis of DIDRIVI the
attractiveness of the product to the customers is determined. Figure 6 shows the general
kind of relationship which must necessarily exist between delivery delay and sales
effectiveness. The figure shows the sales effectiveness from delay as a multiplier
SEDIVI, which is a fraction given in terms of its maximum value. The maximum value of
unity occurs at zero delivery delay. For very small increases in delivery delay the sales
are unaffected. As delivery delay becomes long enough to be of concern to the
customer, sales effectiveness drops rapidly and then levels out as the remaining
customers are those who particularly want this specific product and are unwilling to
change to competitive suppliers unless delivery delay becomes too long.

The switch at SEDS in Figure 5 permits opening of the market loop in the simulation
runs. The sales effectiveness, SE, is given by multiplying the sales effectiveness from
delay multiplier, SEDM, by the value of sales effectiveness maximum, SEM. SEM
represents the sales effectiveness when delivery delay is zero, assuming some particular
and constant set of conditions with respect to price, quality, competence of salesmen
and other influences on the selling process. In Figures 5 and 6, one can see that, as the
order backlog increases (assuming some constant production capability and therefore a
limited delivery rate), the delivery delay indicated will increase. After a delay, the
delivery delay recognized by the company, IDDRC, increases and after a further delay
the delivery delay recognized by the market, DDRIVI, increases. This causes the sales
effectiveness multiplier, SEDIVI, to decrease as shown in Figure 6, which causes sales
effectiveness, SE, to decline and thereby reduces orders booked, OB, until the order
backlog, BL, no longer rises.
D-4079-1 11

6 Table for Sales Effectiveness from Delay Multiplier as It Depends on DDRM.

7 Table for Production Capacity Fraction as It Depends on Delivery Delay Minimum.

In Figure 5 the small lower negative loop determines delivery rates in terms of backlog
and production capacity. These relationships have two objectives. First, for a given
production capacity they should properly relate delivery delay to backlog. When the
backlog is low, the delivery rate should be such that the delivery delay is the minimum
order filling and manufacturing time, taken here as two months. As the backlog rises, the
delivery rate increases but gradually levels off as it approaches the production capacity.
As the second objective, the relationships should permit changing the level of production
capacity while retaining a proper relationship between backlog and delivery rate. This is
done by first generating the concept of delivery delay minimum, DDM, which is the ratio
of backlog to the maximum production capacity. DDM then enters a table as given in
Figure 7, which yields the fraction of the production capacity actually utilized. The rising
slope of the curve in Figure 7 determines the minimum delivery delay caused by order
handling and minimum manufacturing time. The curve is for a delivery delay of
D-4079-1 12

two months when capacity is lightly loaded. To see the effect, assume that backlog
divided by maximum production capacity would yield an implied one-month delivery for
DDM. The value of one month enters on the horizontal axis of Figure 7 and yields a
production capacity fraction of 0.5. In other words, only half of the production capacity
will be utilized. A delivery rate which is half the production capacity will work its way
through the backlog in two months. Likewise, smaller backlogs yielding smaller values of
delivery delay minimum will produce still smaller values of the production capacity
fraction. Going in the other direction, as the backlog increases, the fraction of capacity
utilization increases but not as rapidly as the increase in backlog. As shown in Figure 7
the delivery delay minimum must increase to five months before the theoretical
maximum capacity is achieved. This extra time permits scheduling, ordering of
materials, rearrangement of work load, and backlogging items in front of each machine
so that it can be fully utilized.

A negative feedback loop, as in Loop 2 of Figure 5, is goal seeking. Here the loop tends
to adjust the rate of order booking to equal the delivery rate. If order booking is too high,
the backlog rises and decreases the rate of order booking and vice versa. But such
adjustment does not necessarily progress smoothly to the equilibrium conditions.
Because of the three delays around the loop— in the order backlog, in the company
recognition of delivery delay, and in the market recognition of delivery delay— the
adjustments may occur too late and cause a fluctuating condition in the system. Such is
shown in Figure 8, which illustrates the behavior of Loop 2 when the number of
salesmen is constant and the production capacity is constant (although delivery rate still
depends on order backlog in the region below maximum production capacity). Figure 8
starts with a backlog and a delivery rate below their equilibrium values for the number of
salesmen and the production capacity which have been used. The rate of order booking
is initially too high because of the low backlog and the low delivery delay. But the order
rate in excess of delivery rate causes backlog to rise and causes the delivery delay
recognized by the market to rise. Sales effectiveness and orders booked fall. The rate of
order booking declines below the delivery rate, thereby causing a decline in the order
backlog. Fluctuations of decreasing amplitude continue over the period of 100 months
shown in the figure.

This fluctuating condition is often found in actual market situations. Many factors enter to
create the behavior in addition to those identified in Figure 5. When delivery delays
become long, not only are the customers unwilling to order, but, previously placed orders
are likely to be delivered later than had been promised and salesmen spend time
explaining late orders rather than obtaining new orders. Salesmen become demoralized
and feel there is no point in trying to sell a product which is not available. These factors
reinforce one another to cause a downturn in orders. But orders must fall below the
delivery capability before there is a reduction in the order backlog. One then finds a
reverse situation. Backlog is falling, inventories are rising, deliveries are improving,
pressure is put on salesmen to sell the product, the product becomes more attractive,
and the order rate rises.

Combined Salesmen and Market Loops If Loops 1 and 2 of Figure 1, as detailed in

Figures 2 and 5, are combined, system performance is as shown in Figure 9. Here a
constant production capacity is assumed. The system starts with an initial number of
D-4079-1 13

8 Fluctuation Caused by Delayed Responses within Loop 2.

9 Sales Growth and Stagnation Caused by Interaction of Loops 1 and 2.

salesmen but the number can change. The capacity is greater than needed to produce
what the salesmen can sell. In Figure 9 exponential growth occurs until about the 30th
month. But at that time delivery delay recognized by the market begins to increase.
Sales effectiveness begins to decrease, order rates begin to level off, and revenue no
longer supports the rapid expansion of the sales force. The right hand section of Figure 9
shows generally the kind of behavior already seen in Figure 8. Sales overshoot and then
return toward production capacity. The number of salesmen continue to increase for a
time as the sales effectiveness continues to fall toward a value low enough that the
salesmen generate only sufficient sales budget to maintain themselves.

The combined system consisting of Loops 1 and 2. as illustrated in Figure 9, can serve
to show one of the broad classes of hazards in the changing of policies. Very often the
symptoms of a difficulty superficially suggest a policy change which the system
D-4079-1 14

10 Higher Budget to Sales Effort

itself will defeat. Suppose that in the preceding system one were not aware of the
importance of delivery delay in limiting sales. Such lack of awareness is found
repeatedly on the industrial scene. As illustrated in a later figure, the lack of appreciation
of the influence of delivery delay can arise simply from becoming accustomed to a
particular level of order backlog. Or the lack of awareness can occur because the
industrial system is so complex that few people see all of its manifestations at one time
and are unaware of the implications on one section, such as the marketing effort, of
conditions that exist elsewhere, as in production. Suppose, therefore, that the leveling off
and stagnation of sales in Figure 9 were interpreted as a signal/to increase the marketing
effort. Such would be represented in this model by increasing the fraction of revenue
going to sales RS in Figure 2. The particular numbers used here are incidental but, for
the sake of the example, IRS in Figure 9 represented a $12 contribution from revenue to
sales for each unit of product sold where the unit selling price is $50. Suppose that this
revenue to sales is raised from $12 to $13.60. One then finds the behavior shown in
Figure 10. Here again the constant production capacity is 12,000 units per month.
However, production capacity is often not clear and evident when one product is
immersed in a manufacturing organization involving a multitude of products. The effect
in Figure 10 is to increase slightly the early rate of product growth. The peak in orders
booked is somewhat higher and occurs at 36 months instead of 42 months. However, in
the long run, the effect of the higher budget allocation to sales is to increase the total
number of salesmen and to decrease the sales effectiveness in exactly the right
proportions so that the average orders booked continue to equal the available production
capacity. The higher sales effort increases the selling pressure on the market, increases
the order backlog, increases the product delivery delay, decreases the sales
effectiveness, and results in the same level of sales. Profitability is, however, reduced
because of the greater expenditure to support sales and may also be reduced because
of the higher order backlog, the greater likelihood of factory confusion caused by
rearranging schedules to expedite priority orders and a less than optimum use of
production equipment.
D-4079-1 15

11. Capacity Extension Loop

D-4079-1 16

The differences between Figure 9 and Figure 10 illustrate a very common characteristic
of multiple-loop, non-linear systems. A change in policy can have quite unexpected
equilibrium effects as well as having unexpected dynamic effects. The change in policy
may be aimed at a specific point in the system, here to increase the rate of order
booking. The consequence, however, may be an indirect warping of the system to defeat
the intended result. Here the number of salesmen does indeed go up but the sales
effectiveness goes down correspondingly.

Capital Investment Clearly, the fixed production capacity of Figures 9 and 10 is not
typical. One should then explore capital investment policies. For the sake of illustration,
one of these has been incorporated into the present example to show how capital
investment policy can couple to market variables.

Figure 11 shows a basic part of the decision making process for addition of production
capacity. The addition of capacity is of course contingent on many factors such as
projection of sales trends, financial condition, and product profitability. However, all of
these tend to produce variations on a basic capacity addition rate which depends on the
adequacy of the present capacity in terms of market demand. One of the most
persuasive indicators of the adequacy of existing capacity is the size of the order
backlog and the length of time the customer must wait for delivery. As the delivery delay
rises above the company's goal, the pressure increases for expanding capacity. Figure
11 relates delivery delay to production capacity. In an earlier section of this paper it was
suggested that a policy (rate equation) contains a statement of the decision making goal
and a detection of the related system state. In Figure 11, the delivery delay operating
goal, DDOG, is generated in Equations 19, 20, and 20.1. Here the goal can be a fixed
goal determined by management and adhered to rigidly as a constant goal. Or, at the
other extreme, the goal can be the goal of matching the traditional delivery delay. Or any
intermediate weighting factor can be used to blend these two extremes of goals.
Consider first a fixed delivery delay goal given by the constant DDMG which is here
taken as two months and equals the minimum order processing and manufacturing time.
The delivery delay condition, DDC, in Equation 21 is the ratio of the delivery delay
recognized by the company, DDRC, to the delivery delay goal, DIDOG. From this ratio is
subtracted a delivery delay bias, DDB, which simply represents the competition for
resources throughout the company and represents the deviation between goal and
performance that is necessary to sustain any given level of resource allocation. A
company under great pressure on resources would fall further behind its goals than a
company not under pressure. The coefficient DDB would, in a more complete model, be
a variable generated by such things as financial pressure within the organization and by
the extent to which other operating goals were not being met. Once the ratio of
performance to goal has been established and offset by the bias created by other
pressures, the resulting delivery delay condition, DDC, is conceived to operate through a
relationship similar to that in Figure 12. Here, as DDC rises above 1.0, capacity
expansion occurs. When DDC falls, pressure on production capacity is reduced and
resources are diverted to other areas. The vertical scale is given in terms of the fraction
of existing capacity which is ordered each month. Again, the switch at CEFSW exists not
in the real system but in the model to permit de-activating the capacity expansion loop.
A minor positive feedback loop exists in the figure where capacity ordering adds to
capacity, which adds to the capacity ordering rate because the variable
D-4079-1 17

12 Table for Capacity Expansion Fraction as it Depends on Delivery Delay

Condition

13 Capital Expansion Loop 3 Responds to Delivery Delay and Restores Growth

coming in from CEF is in terms of the fraction per month of the existing capacity. The
effect is to make capacity ordering a function of the operating scale that the system has
currently achieved.

Consider first the case of a constant goal of management to hold a two-month delivery
delay. The effect of delivery delay tradition is not active. The coefficient DDW in
Equation 20.1 is zero. Figure 13 shows the consequence and should be compared with
Figure 9. Production capacity starts at 12,000 units per month which is well above the
initial rate of sales. As a consequence, production capacity is diverted to bring capacity
more in line with the rate of order booking. For this reason the initial increase in sales is
D-4079-1 18

somewhat slower because sales reach capacity and delivery delay rises sooner than in
Figure 9. The rate of order booking rises above the production capacity at about the 24th
week.

14 Delivery Delay Goal Based on Past Performance

At this time order backlog is rising rapidly, capacity is failing slowly, and the delivery
delay is climbing steeply. As delivery delay rises above the two-month goal, the capacity
expansion fraction moves from a small negative value to a large positive value. As a
result of the capacity ordering, the production capacity rises steeply after the 40th week.
It overtakes the orders booked, which have fallen because of the earlier poor deliveries;
delivery delay as recognized by the market again falls; this signals a reduction in the rate
of capacity expansion; and production capacity levels off around the 70th week. Because
delivery delay is low and sales effectiveness is high, the order rate again climbs and
exceeds production capability. Production capacity continues to climb but with a
repeating fluctuation of capacity and order rate crossing one another. One sees here a
classic form of growth instability. High delivery delay simultaneously causes the
expansion of capacity and the suppression of orders. Very sharp crossovers of capacity
and orders occur which at the 84th week are almost in a right angle relationship to one
another.

But Figure 13 is based on management consistently adhering to its delivery delay goal of
two months even under circumstances where actual delivery delays have fluctuated
between 2.5 and 4.5 months. An organization which has experienced sales growth and
periods in which capacity exceeded sales would be inclined to shift its goal structure.
The shift would be to compare present performance with historical or traditional
performance. The historical performance simply means the average perception of past
performance. In Figure 11 this average is generated in Equation 19 for the delivery delay
traditional, DDT. Suppose that this tradition is initially set at two months and the
weighting factor, DDW, is set at 1, meaning that full weight is given to tradition and none
to the delivery delay management goal, DDMG. Here a rather short averaging time of 12
months has been taken for the time to establish delivery delay tradition, TDDT. The goal
structure of the organization is now floating. It simply strives to achieve its historical
accomplishments. For the more subtle goals in an organization, this striving to equal the
past, and conversely being satisfied if one equals the past, is a strong influence.
D-4079-1 19

The result of changing the goal structure from a fixed goal to a goal set by tradition is
shown in Figure 14, which should be compared with Figure 13. The delivery delay
recognized by the market, DDRM, and the delivery delay operating goal, DDOG, of the
company are shown. After delivery delay rises, the operating goal rises after a time
delay. (The goal is responsive to DDRC, not plotted, which is six months earlier than
DDRM, which is plotted as the symbol R.) This means that delivery delay does not
produce the degree of concern that it did when the goal was fixed and low. As a
consequence, expansion does not seem so justified or so important. In Figure 14 the
goal structure continues to collapse. Delivery delay continues to rise, the traditional goal
rises after it, the discrepancy is never great enough to produce active expansion of
capacity, and there is a constant erosion of capacity. As capacity goes down, the rate of
order booking declines to correspond, because in the long run, average orders can not
exceed capacity. Sales effectiveness declines, the revenue to sales declines, and the
revenue becomes insufficient to support the existing number of salesmen. After about
the 70th week the number of salesmen begins to decrease and stagnation has turned
into decay.

Implications for Marketing Marketing is not an activity limited to one corporate

subdivision. Marketing interacts with research to guide product design. Marketing
interacts with price policy to determine profitability. Marketing interacts with production to
determine product availability and quality. Marketing interacts with personnel policy and
training to determine skill in the marketing activities. Marketing interacts with the
corporate information system in determining how market information guides resource
allocation within the company. The relationship of marketing policy to the market, to the
other corporate functions, and to competition is particularly intimate. There are many
interacting feedback loops. These loops are capable of producing product growth or
stagnation and decline. How the loops are balanced determines product profitability.
Misplaced emphasis can raise costs, as illustrated by the increased marketing cost in
Figure 10, whereas the solution to product growth lies in the capital investment policies.

Numerous feedback loops of importance exist in the market complex beyond those
illustrated here. The preceding example assumes a market of unlimited extent. Although
the market responds to delivery delay, for a fixed delay the rate of order booking is
proportional to the number of salesmen. This assumption of an unlimited market is true
for a much wider range of market situations than normally believed. But given a specific
product, there is an upper limit to demand and the demand limiting process within the
market represents additional important loops in the market system. As an example, one
of the most important market loops concerns the kind of product where the product
inventory in the hands of the customer generates customer satisfaction rather than the
rate of product consumption. Customer satisfaction is obtained from the size of the
customer inventory in capital items such as automobiles and household appliances. Here
the sale of a new product fills an inventory and sales then decline to a replacement level.
The recent steep rise and decline in the sale of pin-setting machines for bowling alleys
illustrates a situation where customer inventory can dominate the dynamic behavior.

Market interactions are so complex that they cannot be intuitively appreciated. One must
D-4079-1 20

have a concept of feedback system structure for organizing his knowledge about the
system. Such a structure has been suggested at the beginning of this paper. The
structure can be used for arranging the parts of a system so that behavior can be
simulated as shown in the preceding examples. Market research should be directed
toward identifying the structures that can cause the many dynamic modes that exist in
actual market situations. Emphasis should be reduced on numerical data gathering in
market research and should be increased on conceptualizing the structures which
produce typical classes of market behavior.

Appendix The following pages give the equations used for the system simulation in this
paper. Knowledge of the system structure described in Industrial Dynarnics4 and the
DYNAMO User's Manual5 is assumed. To understand fully the equation format and
structure the reader will need to be familiar with those books. These equations are
included primarily for those who may want to experiment further with the model herein
described.

Following is the full set of equations as they appear in the model. The basic system
structure extends through Equation 27.2. Control cards follow and then the variations of
coefficients used to produce the different computer runs in this paper for Figures 3, 4, 8,
9, 10, 13, and 14.

4
See [2]
5
See [3]
D-4079-1 21

PAGE 1 FILE 41+B42 MARKET LOOPS 01/03/68 1724.9

0.1 MARKET LOOPS

0.2 RUN STD
0.3 NOTE U. OF ILL.--CONVERSE AWARDS SYMPOSIUM, APRIL 13, 1967
0.4 NOTE
0.5 NOTE POSITIVE LOOP--SALESMEN
0.6 NOTE
1 1L S.K=S.J+(DT)(SH.JK+O)
1.1 6N S=10
2 49A SSW.K=SWITCH(SCT,S.K,SWl)
2.1 C SCT=60
2.2 C SWl=0
3 12R OB.KL=(SSW.K)(SE.K)
4 12A B.K=(DRA.K)(RS)
4.1 C RS=12
5 20A IS.K=B.K/SS
5.1 C SS=2000
6 21R SH.KL=(l/SAT)(IS.K-S.K)
6.1 C SAT=20
6.4 NOTE
6.5 NOTE NEGATIVE LOOP--MARKET
6.6 NOTE
7 IL BL.K=BL.J+(DT)(OB.JK-DR.JK)
7.1 6N BL=8000
8 20A DDM.K=BL.K/PC.K
9 58A PCF.K=TABHL(TPCF,DDM.K,0,5,.5)
9.1 C TPCF*=O/.25/.5/.67/.8/.87/.93/.95/.97/.98/1
10 12R DR.KL=(PC.K)(PCF.K)
11 3L DRA.K=DRA.J+(DT)(l/DRAT)(DR.JK-DRA.J)
11.1 6N DRA=DR
11.2 C DRAT=l
12 20A DDI.K=BL.K/DRA.K
13 3L DDRC.K=DDRC.J+(DT)(l/TDDRC)(DDI.J-DORC.J)
13.1 6N DDRC=DDI
13.2 C TDDRC=4
14 3L DDRM.K=DDRM.J+(DT)(I/TDDRM)(DDRC.J-DDRM.J)
14.1 6N DDRM=DDRC
14.2 C TDDRM=6
15 58A SEDM.K=TABHL(TSEDM,DDRM.K,0,10,1)
15.1 C TSEDM*=l/.97/.87/.73/.53/.38/.25/.15/.08/.03/.02
16 51A SEDC.K=CLIP(SEDF,SFDI,TlME.K,SEDCT)
16.1 C SEDF=l
16.2 C SEDI=l
16.3 C SEDCT=36
17 49A SEDS.K=SWITCH(SEDC.K,SEDM.K,SW2)
17.1 C SW2=0
18 12A SE.K=(SEDS.K)(SEM)
18.1 C SEM=400
18.4 NOTE
18.5 NOTE CAPITAL INVESTMENT
18.6 NOTE
19 3L DDT.K=DDT.J+(DT)(l/TDDT)(DDRC.J-DDT.J)
19.1 6N DDT=DDRC
19.2 C TDDT=12
20 15A DDOG.K=(DDT.K)(DDW)+(DDMG)(DDWC)
20.1 7N DDWC=1-DDW
20.2 C DDW=O
20.3 C DDMG=2
21 27A DDC.K=(DDRC.K/DDOG.K)-DDB
D-4079-1 22

PAGE 2 FILE 41+B42 MARKET LOOPS 01/03/68 1724.9

21.1 C DDB= .3
22 58A CEF.K=TABHL(TCEF,DDC.K,0,2.5,.5)
22.1 C TCEF*= -.07/-.02/0/.02/.07/.15
23 49A CEFS14.K=SWITCH(O,CEF.K,SW3)
23.1 C SW3=0
24 12R PCO.KL=(PC.K)(CEFSW.K)
25 39R PCR.KL=DELAY3(PCO.JK,PCRD)
25.1 C PCRD=12
26 1L PCOO.K=PCOO.J+(DT)(PCO.JK-PCR.JK)
26.1 12N PCOO=(PCO)(PCRO)
27 IL PC.K=PC.J+(DT)(PCR.JK+0)
27.1 6N PC=PCI
27.2 C PCI=12000
27.5 NOTE
27.6 NOTE CONTROL CARDS
27.7 NOTE
27.8 PLOT OB=*,PC=C(0,24000)/SE=E(0,400)/S=S(0,80)
27.9 NOTE B42, RERUNS OF B41
28 NOTE
28.1 RUN A
28.2 NOTE UNLIMITED EXPONENTIAL GROWTH
28.3 SPEC DT=.5/LENGTH=100/PRTPER=100/PLTPER=2
28.4 PRINT 1)S
29 C SWl=l
29.1 C PCI=100000
29.4 PLOT OB=*(0, 24000)/SE=E(0,400)/BL=B(0,120000)/S=S(0,80)
29.5 RUN B
29.6 NOTE GROWTH AND DECLINE
30 C SW1=l
30.1 C SEDF=.25
30.2 C PCI=100000
30.5 RUN C
30.6 NOTE NEGATIVE LOOP OSCILLATION
31 C SW2=1
31.3 PLOT 0B=*,DR=D(0,24000)/SE=E(0,400)/DDRM=R(2,6)/BL=B(0,120000)
31.4 RUN D
31.5 NOTE SALES STAGNATION
32 C SW1=1
32.1 C SW2=1
32.4 PLOT OB=*(0,24000)/SE=E(0,400)/SH=H(0,2)/DDRM=R(2,6)/S=S(0,80)
32.5 RUN E
32.6 NOTE INCREASED SALES BUDGET ALLOCATION
33 C SWI=l
33.1 C SW2=1
33.2 C RS=13.6
33.5 RUN F
33.6 NOTE CAPACITY EXPANSION
34 C SW1=1
34.1 C SW2=1
34.2 C SW3=1
34.5 PLOT
OB=*,PC=C(0,24000)/SE=E(0,400)/DDRM=R,DDOG=G(2,6)/S=S(0,30)/CEF=
34.6 X1 -.06,.18)
34.7 RUN G
34.9 NOTE GOAL=TRADITION WITH DELIVERY DELAY BIAS PRESSURE
35 C SWl=l
35.1 C SW2=1
35.2 C SW3=1
35.3 C DDW=l
D-4079-1 23

References

1 Forrester, Jay W. "Industrial Dynamics - After the First

Decade," forthcoming Management Science, March,
1968.

2 Forrester, Jay W. Industrial Dynamics. Waltham, MA:

Pegasus Communications, 1961.

3 Pugh, Alexander L. Ill. DYNAMO User's Manual,

Second Edition. Cambridge, MA: M.I.T. Press, 1963.

4 Slate, Daniel M. and Ferber, Robert (eds.). Systems

Analysis: Research and Implications for Marketing.
Urbana, IL: Bureau of Economic and Business
Research, University of Illinois, forthcoming, 1968.

References added, July 1999, while electronic

scanning of the original document:

5 Forrester, Jay W. 1975. Collected papers of Jay W.

Forrester. Waltham, MA: Pegasus Communications.
284 pp.

6 Forrester, Jay W. 1971. World Dynamics. (1973

second edition.). Waltham, MA: Pegasus
Communications. 144 pp. Second edition has an added
chapter on physical vs. social limits.

7 Forrester, Jay W. 1969. Urban Dynamics. Waltham,

MA: Pegasus Communications. 285 pp.

8 Forrester, Jay W. 1968. Principles of Systems.

Waltham, MA: Pegasus Communications. 391 pp.