D-3419-1
RATIONALITY AND STRUCTURE IN
BEHAVIORAL MODELS OF BUSINESS SYSTEMS
By
John D. W. Morecroft
System Dynamics Group
Massachusetts Institute of Technology
Cambridge, Massachusetts
To be presented at the 1983 International system
Dynamics Conference, Chestnut Hill, Massachusetts,
July 27-30, 1983
February 1983
Revised April 1983
The author received valuable help and suggestions on an earlier
draft of this paper from Jay W. Forrester, John M. Richardson,
John D, Sterman, George P. Richardson, and Robert L. Eberlein.
D-3419-1 2
ABSTRACT
Rationality is an underutilized concept for creating and
analyzing behavioral simulation models of business systems. Much
explanatory power and insight can be gained by assuming that
business decisionmaking is intendedly rational, examining the
factors that limit rational adjustment in business decisions, and
exposing in simulation experiments the rationality that underlies
even the most counterintuitive total-system behavior.
The paper begins by defining rationality and illustrating
the difference between objective rationality, which is common in
classical economic models of decisionmaking, and bounded ration-
ality, which is common in behavioral models of decisionmaking.
Two methods of analysis are then proposed for clarifying the
theory implicit in a simulation model, The first method is
premise description. In describing decision: functions and model
equations attention should be drawn to the organizational
processes of factoring, goal formation, routine and tradition
that limit the area of rational adjustment in business decision-
making. The second method is partial model testing. A sequence
of partial model tests should be designed to examine the intended
rationality of decisionmaking. ‘The intuitively clear and
sensible behavior of partial tests should be contrasted with the
more complex and often counterintuitive behavior of the whole
model.
The application of these methods is illustrated with a
simulation model of a sales organization containing linked
decision functions for sales objectives and salesman overtime,
and a behavioral function for sales force motivation.
D-3419-1 3
INTRODUCTION
Suppose we learn that a magazine publisher goes out of business
and that the circumstances leading up to the collapse are record
losses and, at the same time, record revenues and circulation--
circumstances that have occurred a number of times in the past.
What do we conclude about the rationality of the publisher's
strategy? It is clearly unreasonable to assume that the business
failure occurred because strategy was deliberately designed to
cause major losses. Rather, the separate policies comprising the
strategy were intendedly rational, but when linked in a commer—
cial setting they produced an unexpected and undesirable outcome.
The thesis of this paper is that rationality is an
underutilized concept for creating and analyzing behavioral
simulation models of such business problems.. There is a great
deal of explanatory power and insight to be gained by assuming
that business decisionmaking is intendedly rationai, examining
the factors that limit rational adjustment in business decisions,
and exposing in simulation experiments the rationality that
underlies even the most nonrational total-system behavior.
For example, suppose we have a simulation model of the
familiar industrial production and distribution system’. The
model is composed of many interrelated decision functions for
ordering, inventory control, forecasting, labor adjustment, and
so forth. It is well known that such a system produces costly
D-3419-1 4
fluctuations in production, ordérs ‘and labor force. (Mack 1967;
The behavior is at first surprising, because we assume that re-
tailers, manufacturers, and vendors are not intending to create a
costly system--in fact, quite the reverse. Simulation analysis
by itself is capable of showing that fluctuation is a possibi-
lity. But simulatiori coupled with an analysis of rationality can
reveal how such fluctuations arise from intendedly rational
decisionmaking and therefore why the behavior is likely to
Persist.
In, the paper we will first explore the concept of
rationality in decisionmaking, drawing on the ideas of the
Carnegie school. We will then describe two methods of model
analysis--premise description and partial model testing--that can
clarify and better communicate the theory-implicit in a
simulation model.
OBJECTIVE AND BOUNDED RATIONALITY
Simon (1982, p. 209-238) has characterized theories of
objectively rational behavior as
++.those that employ as their central
concepts the notions of (1) a set of
alternative courses of action presented to
the individual's choice; (2) knowledge and
information that permit the individual to
Predict the consequences of choosing any
alternative; and (3) a criterion for
deciding which set of consequences he
. prefers.
D-3419-1 5
Objectively rational behavior is possible, when the
conditions surrounding decisionmaking are very simple. Consider,
for example, an individual faced with the choice between wearing
and not wearing a raincoat in a violent rainstorm. The
consequences are to stay dry or get wet in the rainstorm, and the
presumed criterion is that the owner wishes to stay dry.
Objectively rational behavior is likely because the choices are
very limited (only two), the consequences obvious, and the
criterion straightforward to apply and unlikely to change.
By contrast, our magazine business strategy is much more
complex. It is very difficult to reliably deduce the
consequences of the interacting policies underlying strategy.
Consider pricing policy as one element of the overall business
strategy (and bear in mind that-in the magazine publishing
industry subscription prices affect circulation, which in turn
affects advertising revenue: advertisers.will pay more to
advertise in a magazine with high circulation). (Hall 1976) To
anticipate the effects of a price change, and therefore to set
prices rationally, one must fully understand customer needs, the
Structure and maturity of the market, competitor prices and price
responses, not to mention the circulation requirements of
advertisers and the general economics of the publishing business.
The consequences of alternative prices are contingent on an
enormous chain of linked events, Objectively rational magazine
pricing is much more difficult than objectively rational
weatherproofing!
D-3419-1 6
It is helpful for our later discussion to picture an
objectively rational decisionmaking process. Figure 1 shows
information (say, about weather conditions) flowing from the
environment into the decisionmaking process. The information
prompts the selection of an appropriate course of action from a
set of alternatives. ' Given-a criterion--the desire to stay
dry--and the exercise of intelligence--the ability to deduce that
a raincoat ensures dryness--a rational course of action is
selected. In general, a wide range of alternatives may be
considered. A course of action is selected that optimizes a
criterion function, shown in the figure as £(x,y,z). Adequate
intelligence, shown as a memory matrix, is available for the
storage of information and the prediction of consequences.
The figure readily allows for considerable complexity in the
decision process. It is possible that a course of action will
influence the environment (shown by a return flow of informa
tion). Objectively rational behavior will have available the
mental capacity to factor these interactions into the selection.
It is possible that the set of alternatives is very large, and
the criteria subtle and interdependent. Objectively rational
behavior handles this complexity with ease and precision.
D-3419-1
ENVIRONMENT
Complete
Information
T
1
1
‘
1
1
'
1
1
l
DECISIONMAKING
\ PROCESS
Criterion: Mox f(x,y,z)
Intelligence a
A= 1506
Figure 1. Picture of Objectively
Rational Decisionmaking
D-3419-1 8
Bounded Rationality
Problems of choice in business organizations are very
complex. To set an objectively rational production plan in a
manufacturing firm, one would want to know the personal
circumstances and state of mind of all potential customers, the
sales plans of all retailers, the exact status of retail and
distribution inventories, the status of manufacturing
inventories, the condition and availability of plant and
equipment, the willingness to work of the workforce, the
production plans of all suppliers, and the capacity, production;
and promotion plans of all competitors. In addition, one would
need a sound and detailed knowledge of the economics of the
industry to compute and compare the financial consequences of
alternative production plans.
No’ individual can possibly hope to solve problems posed in
such complex terms. Yet organizations exist, they are managed by
ordinary people, often with great success. This seeming paradox
is readily explained when we realize that organizations are
structured to "transform intractable decision problems into
tractable ones." (Simon 1979, p. 501) Individuals in
organizations exhibit only bounded rationality--they make
rational decisions under conditions of choice that have been
deliberately simplified.? -There is usually a rationale for any
business decision. Whether that rationale "makes sense" for the
D-3419-1 |
organization as a whole deperids on whether the simplified
conditions of choice lead to actions that support the goals of
other parts of the enterprise.
The challenge to the modeler and theorist is to be sensitive
to the many ways in which conditions of decisionmaking can be
simplified, and to develop the vocabulary for recognizing and
describing such situations. As Simon (1982, p. 215) has pointed
out: :
Significant models can be constructed by
singling out for attention, and for
embodiment in them, the significant
limiting conditions that serve as
boundaries to the area of rationality in
human behavior.
Factored Decisionmaking
One way to simplify a complex decisionmaking process is to
factor it into small pieces as shown in figure 2. Within each
decision function there are few alternatives, simple criteria to
be satisfied, and intelligence matched to the simplified problem.
Factored decisionmaking is an inescapable empirical feature of
all organizations. (Allison 1971; Cyert and March 1963) There
are important structural implications of such an arrangement.
Information is distributed among the various decision nodes of
the system. Each node receives only part of the available flow
of information--an amount sufficiently small to allow timely
processing and action. Organizations are clearly a long way from
monolithic thinking; they are systems of weakly coupled,
distributed thinkers. Models should properly reflect this
obvious structural feature of the information network.
D-3419-1 10
ENVIRONMENT )
\
\
/ \ Portiol
t \ \\ Information
Portiol
Information ,'
Fo es
Alternotives
Criterion: Sot
sinteltigence FF
*Few
Alternatives: to
“Criterion: Sotisfice x,
sintettigence FA]
Portiol
Information
Few C
Alternatives
“Criterion: Satisfice 2
intelligence FF]
As1807
Figure 2. Factored Decisionmaking
and Bounded Rationality
D-3419-1 ll
Goal Formation and Incentives ~
Goals and incentives can. simplify decisionmaking by focusing
the attention of an individual on a small part of the enterprise
and making him responsible and accountable for its success. In
terms of information flows (and therefore model structure), goals
and incentives determine what information is viewed as important,
and what is considered irrelevant at different points of the
organization.4
Authority, Culture and Style
Authority, culture and style also simplify decisionmaking,
though often in intangitle ways. They serve to transmit basic |
values and traditions of the organization to all its members.
Authority and culture permeate thinking at the decision nodes of
the enterprise, altering the premises of decisionmaking and often
introducing bias and distortion into the interpretation of
information.
For example, in an interesting case modeled by Forrester
(1968,1) the president of a company with a fast-growing new
product line insisted on maintaining strict personal control over
the approval of all capital expenditures. As a result there was
a bias in the company's decision process for capital-equipment
ordering. Considerable demand pressure, in the form of high
order backlogs, had to accumulate to justify expansion. The
model that incorporated this conservative facet of executive
D-3419-1 12
style showed that the bias in capital-equipment ordering could
cause sales to stagnate in a potentially enormous market.
Routine
Organizations are great storehouses of specialized decision
processes and routines. (Allison 1971, p. 83; Nelson and Winter
1982, pp. 96-136) Experienced employees carry around in their
heads a repertoire of standard responses to recurrent business
situations. Routines are yet another way of simplifying
decisionmaking. They predetermine the information to be used in
decisionmaking and supply rules of thumb for processing the
information. Typically, routines use small amounts of
information and simple rules of thumb. For example, a pricing
policy might be routinized to set prices for this year's product
x-percent higher than last year, where the: percentage increase is
governed, say, by inflation in costs. Routines are important
because they introduce momentum into organizational behavior. an
organization that encounters rapid change in its environment
(say, competitor prices) may find its repertoire of standard
responses (cost plus pricing) inappropriate to the new situation.
Basic Cognitive Processes
When the conditions surrounding decisionmaking have been
simplified by factoring, by goal formation and incentives,
authority, culture, and routine, there still remain limitations
on rationality imposed by basic cognitive processes.” People
D-3419-1 13
take time to collect and transmit information. They take still
more time to absorb information, process it, and arrive at a
judgment. There are limits on the amount of information they can
manipulate and how much they retain in memory. ‘These basic
cognitive processes are also a part of bounded rationality--in
fact, the basic constraint on rationality once organizational
measures to improve it have been exhausted. Cognitive processes
can introduce delay, distortion, and bias into information
channels which the modeler should try to capture.
Summary
In this section we have explored the two concepts of
objective and bounded rationality in some detail. Objective
rationality is an ideal of rationality that requires monolithic,
highly integrated thought and is rarely exhibited in real choice
problems, except perhaps trivial ones. Bounded rationality is
the rationality of normal humans in real organizations. To
portray and interpret bounded rationality in a model requires a
knowledge of features of organization used to simplify decision-
making: factoring, goal formation, incentives, authority,
culture, style, and routine.
USING RATIONALITY TO PROVIDE STEPPING STONES IN MODEL~BASED
THEORY
This section proposes two methods of analysis--premise
description and partial model testing-~for clarifying the
structure of computer simulation models and refining their
D-3419-1 14
implicit theories of behavior. ‘Both methods examine the bounded
rationality of the model's decisionmaking--first at the level of
equations, then at the level of simulation runs. The two methods
provide stepping stones to fill the (usually) large gap in logic
between the assumptions embodied in single equations of a model
and the simulated corisequences of the many equations.®
Premise Description of Decision Functions
Premise description is similar to a normal equation
description (see, for example, Forrester 1981, pp. 215-251). But
where a-normal equation description reports in a journalistic
sense how decisions are made, premise description goes further by
focusing on the simplifying conditions/organizational processes
that bound the rational adjustment of each decision function.
The modeler starts with a diagram of the model system
showing the network of interlinked decision functions. He then
presents the equations corresponding to each decision function,
drawing attention to the way factoring and local goals simplify
rational choice; how authority and culture influence the content
and interpretation of information streams; and how routine and
cognitive limitations influence the collection, processing, and
transmission of information.’ At the back of his mind the
nodeler has as a yardstick the notion of objective rationality.
This yardstick raises questions of why some information is
available in a decision function and other is not, why delay and
D-3419-1 15
distortion occur in the transinission and interpretation of
information, and why bias is present. ‘The answers to these
questions naturally point to empirically observed organizational
processes that underlie bounded rationality.
Such a model description alerts the reader to the
deficiencies present in the information network and signals the
possibility of problem behavior in the system as a whole. The
decision functions of the model are seen to be intendedly
rational within the bounds set by common organizational practice,
yet far. removed from the demanding standards set by objective
rationality.
No unique way exists of describing the “extent of
rationality" of a given decision function-or for measuring how
much a function departs from objective rationality. The
description of premises simply makes the modeler (and reader of
the model) conscious of the limitations on decisionmaking
embodied in the model.
In cases where substantial insights have been gained into
the conditions required for “optimal” decisionmaking, the
yardstick of objective rationality may be applied with more
precision. For example, in the well-studied area of production
planning and control (Bitran and Hax 1977; Holt et al. 1960),
there is considerable understanding of how aggregate production
D-3419-1 16
rates should be set to minimize inventory carrying, set-up, and
overtime costs. The information content of a heuristic
production planning decision rule might be judged against the
richer and more complex information structure of the “optimal
rule."® put, more often, it is up to the modeler to decide how
best to draw the attention of the reader to the bounded
rationality implicit in the simulation model.
Partial Model Tests.of Intended Rationality
The second method of model analysis is partial model
testing. Partial model testing has long been used in simulation
modeling to debug subsystem models prior to whole model
simulations. Here we suggest that partial tests have a much more
important role to play in model analysis. ‘They should be used to
expose the intended rationality of business decisionmaking.
There is a single assumption that justifies the new and
important role of partial model tests. It is that decisionmaking
is rational within the context of the premises supplied to the
decisionmaker and the limits of his mental computing capacity.°
This assumption enables one to decompose a complex simulation
model into small pieces and to expect simulation runs of the
pieces to reveal intuitively clear, plausible behavior. The
partial tests. should show that local. decisions are well adapted
to achieving local goals provided the organizational setting is
sufficiently simple. ‘The assumption of intended rationality does
D~3419-1 17
not imply that the behavior of the whole system is well adapted
to the many goals of the enterprise. Dysfunctional behavior of
the organization is quite possible but is a systemic problem
resulting from the coupling of decision functions--in other
words, a flaw in the structure and design of the organization as
a whole.
The analysis begins with a causal-loop diagram (Richardson
and Pugh 1981, pp. 25-30) that shows compactly the feedback loops
resulting from organizational, cognitive, behavioral, and
physical assumptions of the underlying equations. The
causal-loop diagram is then used to design a sequence of
simulation experiments to explore the behavior of pieces of the
total feedback structure. !°
The tests show how one (or perhaps a
flew) decision functions work when the premises of rational
adjustment for the functions are not seriously violated. Partial
tests are then compared with whole model tests to understand the
causes of behavior (particularly dysfunctional behavior) in the
complete system.
For example, in a model of production planning and labor
adjustment containing simple linear inventory control and labor
hiring rules (Forrester 1968,2; Holt et al. 1960, pp. 363-388),
it is instructive to consider how the inventory control rule
performs when the delay in adjusting production (caused by labor
hiring) is made small. (Lyneis 1980, pp. 185-205) Under this
D-3419-1 18
partial model test aggressive inventory management (meaning vapid
correction of inventory imbalances) always has the intuitively
correct effect of bringing inventory in line with its goal
quickly. ‘The premise of the inventory policy (eliminating
inventory discrepancies quickly) is perfectly valid if production
can respond instantaneously to requests for inventory replenish-
ment. However, in the complete model, when a labor adjustment
delay of say four to six weeks is present, rapid correction of
inventory discrepancies leads to the initially counterintuitive
result that inventory rebalancing is delayed. ‘The system becomes
quite oscillatory and takes a long time to settle into
equilibrium.
The great strength of partial model testing is most apparent
when the whole model (or some larger configuration of loops
exhibits counterintuitive and highly ineffective behavior.!+
Then it is apparent that the surprise behavior of the whole model
is a consequence of the interaction of many intendedly rational
parts. In other words, in the coupling of the many decision
functions, the premises or conditions for rational adjustment of
individual functions are violated. In these situations a system
of decision processes fails to integrate in a way that the
rationality of the parts is a close approximation to the
objective rationality required for success of the total system.
The contrast, of partial and whole model tests provides a powerful
explanatory tool for behavior analysis and theory creation.
D-3419-1 19
A BEHAVIORAL MODEL OF A SALES ORGANIZATION
In this section a simple model of a sales organization is
described and analyzed using the methods of premise description
and partial model testing. The simple model is based on a much
larger model developed to examine marketing strategy for a vendor
of advanced office equipment. ‘The larger model contained more
than 100 active equations describing some twenty interlinked
decision functions, covering customer purchasing and price
perceptions in the market, and salesman time allocation,
overtime, motivation, and objective setting in the sales
organization. ??
The structure of the larger model, in terms of
factored decision nodes, information flows, heuristics, routines,
biases, and so forth, was derived from the operating knowledge of
members of the project team and subject area experts in sales and
marketing. The information was-gleaned mostly from interviews
and roundtable discussions as described in Morecroft (1983,2).
The simple model contains just 14 equations and focuses
attention on the decision functions for overtime and objective
setting, and the behavioral function for sales force motivation.
The structure and parameterization of these functions are
virtually unchanged from the larger model. ‘The rationality of
the simple model is therefore representative of the larger,
empirically derived model. Moreover, its simulated behavior, to
be described later, has much in common with the more complex
model.
D-3419-1 20
Model Overview - Factored Decisionmaking
Figure 3 shows the policy structure of the simple model
containing six functions for sales, sales objective, performance,
overtime, motivation, and sales effort.+? prom discussions in
the sales organization it was clear that revenues and
profitability were a major concern of chief executives. But
there was obviously no monolithic, integrated decisionmaking
process for maximizing either revenue or profit. Rather, the
task of maintaining acceptable rates of revenue and profit was
factored within the sales organization among marketing managers;
staff analysts, and salesmen. Figure 3 depicts this factoring.
On the right side of the figure, sales are generated by the
sales effort of individual salesmen. Market planning managers
and their staff have the responsibility for setting challenging
sales objectives, which they do largely based on past sales
performance. The sales objective is then handed to the field
sales force that has the responsibility of deciding how much time
must be expended to meet the objective. The major decisionmaking
nodes, then, are the setting of the sales objective, which is
factored to market planning; assessment of sales performance,
which is factored between field sales managers and their
salesmen; and finally the overtime decision, which is the
personal responsibility of individual salesmen.
D-3419-1 a.
D-3419-1 22
The figure also shows functions for motivation and sales
effort. Motivation is a purely behavioral function that portrays
the response of salesmen to varying conditions of workload
(measured by overtime) and performance. Sales effort is a
behavioral/physical function that computes how much sales effort
is available from a given number of salesmen working a given
Soles amount of overtime at a particular level of motivation.
Effort Time Per
Premise Description of Decision Functions
This section describes the premises of decisionmaking for
the setting of sales objectives and overtime. The description
draws attention to the sources, uses, and interpretation of
information in the sales objective and overtime functions.24 a
documented listing of the equations of the complete model is
included in the appendix.
a). Sales Objective
ote MSO, ="SC, * (1+MASC) q)
MASC=0.05 Dimensionless ql.)
Performance \* , MSC, =MSC,_,+(1/TESC) (MS,_,-MSC,_1) (2)
Objective 7 MSCy=MSq (2.1)
~ fa A-1508
2 Months (2.2)
where MSO - Monthly Sales Objective (Units Per Month)
MSC - Monthly Sales Commitment (Units Per Month)
MASC - Margin for Achievement of Sales Commitment
(Dimensionless)
Figure 3. Policy Structure of A Simple
Sales Organization
D-3419-1 23
MS - Monthly Sales (Units Per Month
TESC - Time to Establish Sales Commitment (Units Per Month)
The sales objective is set by market planning managers and
their staff. The process is a particularly interesting example
of bounded rationality that illustrates the role of authority,
organizational routine, and cognitive limitations in forming the
premises of decision.
Equation 1 states that the monthly sales objective MSO is
based on a monthly sales commitment MSC inflated by a fixed
margin MASC of 5%. The formulation captures a political goal
formation process. Managers make a commitment to higher-level
executives to sell a certain number of units in their sales
region. Their own performance as managers-is judged on their
ability to fulfill this commitment. To build in a margin of
safety for themselves and a challenge for.the sales force, they
deliberately inflate the sales objective above their own
commitment, in this case by a margin of 5%. The margin provides
security for the market manager and, at the same time, pressures
the sales force to improve on its past sales performance. It is
a remarkably simple device by which executive pressure for
cost-effective performance can be transmitted through
middle-level managers to affect the efforts of salesmen,
D-3419-1 24
Equation 2 states that the ‘monthly sales commitment MSC-is
an exponential smooth (Forrester 1961, pp. 406-411) of past
monthly sales MS with a time constant T2sC of twelve months. At
the heart of the sales-commitment process is the routine of
commiting to sell in the future the same amount as was sold in
the recent past. It is a routine that demands little detailed
information--certainly much less than would be required by
sophisticated market forecasts or other more formal and more
“rational” approaches to commitment and planning. Yet there was
wide agreement in the organization that the real process was
heavily. dependent on recent sales.
b). Overtime
MOT, =£, (PSO,) . (3)
where MOT - Multiplier From Overtime (Dimensionless)
PSO - Performance on Sales Objective (Dimensionless,
£, - Nonlinear Decreasing Function of PSO
Equation 3 for overtime states that salesmen take
performance on the sales objective PSO as the premise for their
overtime decision, Provided salesmen are meeting or exceeding
the sales objective (PS0>1), there is no particular incentive to
put in overtime. As performance falls below the objective
($0<1), salesmen feel pressure to work harder--both to look good
on the job and to avoid loss of income from sales bonuses.
D-3419-1 25
Overtime rises sharply to a peak 40% greater than the standard
130 hours per month. (See the full equation listing in the
appendix for the exact shape of the nonlinear function f,.)
Two features of rationality deserve comment in this formu-
lation. First, the salesman's decision on how hard to work is
tied exclusively to the local sales objectives supplied by market
planning managers. The decision function does not contain a
revenue-maximizing algorithm for the whole sales organization,
which would require much more information. Moreover, the
function does not contain any explicit income-maximizing algo-
rithm for individual salesmen, The assumption is that a salesman
works overtime to achieve his sales objective, not to maximize
his personal income. (An increase in overtime usually prevents
income loss but is not precisely calculated to minimize loss.)
¢). Performance on Sales Objective
PSO,=PSO,_)+(1/TPSO) ( (MS, _1/MSO,_1)-PSO,_1) (4)
PSO9=IPSO (1)
IPSO=1/(1+MAPC) (4.2)
TPSO=3 Months (4.3)
where PSO - Performance on Sales Objective (Dimensionless.
MS - Monthly Sales (Units Per Month’
MSO - Monthly Sales Objective (Units Per Month)
IPSO - Initial Performance on Sales Objective
(Dimensionless)
TPSO - Time for Performance on Sales Objective (Months)
D-3419-1 26.
Performance on the sales objective PSO is formulated in-
equation 4 as an exponential smooth of current performance, with
a time constant TPSO of 3 months. Current performance is the
ratio of monthly sales MS to the monthly sales objective MSO.
The formulation. captures a natural cognitive smoothing process in
decisionmaking. A fall in monthly sales relative to the
objective does not immediately lead the salesman to conclude his
performance has declined. Only a drop in sales sustained for
several months will persuade the salesman he is missing target
and should take corrective action.
Description of the Behavioral Motivation Functions
This section presents a standard equation description of the
model's nonlinear motivation functions that relate the
productivity of salesmen to pressures from overtime and
performance against sales objective. These functions are not
conscious decision functions; they portray behavioral properties
of people. It is important to know how the functions are
formulated to interpret the simulation runs presented later.
EMSE,=£,(M,) (5)
MeeMyyCL/TEM) (M41 —My) (6)
M gsIMz (6.1)
TEM=3 Months (6.2)
MI,= (MIO, *MIP,) *SMI+(1-SMI) *IMI ay
sMI=0 (7.2)
D-3419-1 27
IMI=MI09*MIP) . (7.2).
MIO, =£, (MOT, ) : (8)
MIP, =£ ,(PSO,) (9)
where EMSE - Effect of Motivation on Sales Effort
(Dimensionless)
£, ~ Nonlinear Increasing Function of Motivation
M - Motivation (Dimensionless)
TEM - Time to Establish Motivation (Months)
MI - Motivation Index (Dimensionless)
SMI - Switch for Motivation Index (Dimensionless)
g
a
'
Initial Motivation Index (Dimensionless)
MIO - Motivation Index From Overtime (Dimensionless)
£, - Nonlinear Decreasing Function of MOT
MIP - Motivation Index From Performancé (Dimensionless)
fy - Nonlinear Increasing Function of MIP
Equation 5 asserts that low motivation will reduce the
Sales effort of the sales force. Motivation is defined on a
dimensionless scale from 0 to 1. Figure 4 shows the shape of the
behavioral relationship. When motivation is high (around 1), it
has little or no depressing effect on sales effort. As
motivation falls below .8, it has an increasingly depressing
effect on sales effort, reducing it by fully 35% when motivation
reaches a value of .4. It is assumed that no matter how low
motivation falls, it will not depress sales effort by more than
608.
D-3419-1
T
95,
a |
85 i
|
8 j
Effect of |
Motivation on i
Soles Effort 7
EMSE 65
6
s| 5
4 f 1
o- 2 4 6 8B |
Motivation M
(Dimensionless Index)
Figure 4, Effect of Motivation on Sales Effort
12
8
Motivation
Indices |
MIO, MIP 6
Motivotion index
Ma from Overtime MIO
Py ee)
Motivation Index
2 fom Performance MIP 3
° A 1 "
8 9 ! WW 12 13 14
Multiplier from Overtime MOT
T T r T T + a]
6 z 8 9 1 We
Performance on Soles Objective PSO
tae
Figure 5. Determinants of the Motivation Index
28
D-3419-1 29
Equations 6 to 9 assert that sales force motivation depends
on working conditions--the level of overtime and performance
against sales objective. High levels of overtime and poor
performance lower motivation, Equation 6 states that motivation
M-lags three months behind the motivation index MI--the index of
current working conditions. It takes time to become demoralized!
In equations 7, 8, and 9 the motivation index MI is defined as
the product of nonlinear functions (£3 and £,) of overtime MoT
and performance on sales objective PSO. The shape of the
functions is shown in figure 5.
Partial Model Testo of Intended Rationality
Figure 6 shows the feedback structure of the system which
forms the basis for designing partial tests of intended
rationality. It is composed of four interlocking loops.
Loop 1 contains the adjustment process of the individual
salesman. If sales volume falls, say, because each sale takes
more time, then the salesman will see his performance fall and
will compensate by putting in overtime, thereby boosting sales.
Loop 1 is, in system dynamics terminology, a goal-seeking loop,
in which the overtime decisions of salesmen are geared toward
meeting the sales objective.
Loop 2 contains the commitment process of market managers.
In loop 2, if sales fall, managers will gradually become aware of
D-3419-1
Co.»
Effective
“Motivation
Figure 6.
4
/ Effort
Overtime
\
Soles
Performonce
Soles «Time Per
Sole
Soles
Objective
Morgin
Ants,
Feedback Structure of the Sales Model
30
D-3419-1
w
s
D-3419-1 31
1200
12
the fall, and factor it into their commitments and objectives. ay
Loop 2 then works in the opposite sense to loop 1, allowing a
Performance on Soles Objective |
Se A Ee A
Soles Objective
relaxation of sales objectives when general market. conditions
1000
tighten.
Soles ond Sales Objective (Units / Month)
Performance ond Motivation (Dimensiontess Indices) <==-—
Loops 3 and 4 are formed by the behavioral motivation 8
function in the model. Because the function itself is highly
nonlinear, the loops are not active under most business es
conditions. However, when they do become active, they tend to Sr
undermine the efforts of salesmen to achieve their sales |
objective. When motivation is low, a salesman who puts in longer 8 j - : ay
hours to achieve his sates objective might end up generating ap 7? zs ‘iis sc 1%
fewer sales due to his decreased sales effectiveness. |
g ¢
8 z
Intended Rationality of Salesman Overtime Adjustment
Figure 7 shows the adjustment of the system to a 50%
unanticipated increase in the normal time.per sale, from 60 hours
75000
to 90 hours per unit.!® The adjustment is made under the <=
Effective Soles Effort
assumption that the sales objective does not change and that the ;
motivation function is neutral. We therefore see a test of the
overtime adjustment by itself, in other words, loop 1 in
Stondard Soles Effort
Overtime (Fraction)
isolation.
45000
T
a
8
The adjustment is rapid and intuitively sensible. In month
1, sales fall by 1/3. The resulting large discrepancy between
Effective and Stondord Soles Effort (Hours Per Month)
60000
30000
|
° 25 5 75
Months
monthly sales and the objective causes salesmen to increase
Figure 7, Salesman Overtime Adjustment Loop 1 in Isolation
D-3419-1 33
overtime from an initial value of 10% (of the normal 130 hours
per month) to a final value of almost 40%. Most of the
adjustment is complete in the first two months of the run. ‘The
system settles into a “stressed equilibrium,” in which salesmen
are working long hours under pressure from an unyielding sales
objective.
Intended Rationality of Commitment and Objective Setting
It is clearly unreasonable to assume that market managers
will fail to learn about major market changes and factor them
into their commitments and objectives. Figure 8 shows the
adjustment of the sales objective in response to the same 50%
increase in time per sale, under the assumption that overtime
cannot rise above 10% and that the behavioral motivation function
is neutral. The simulation is a test of the adjustment around
loop 2 in isolation, (Readers should note that the time scale in
this run has increased to 50 months by comparison with 10 months
in the previous run.)
Monthly sales fall by 1/3 in month 4, The sales objective
falls as market managers learn of the tightened market conditions
and renegotiate their sales commitment with executives. But the
fall is gradual. It takes time to be convinced that the decline
in sales is permanent and not simply the result of unusual, but
temporary, market conditions or reduced effort by salesmen. The
market manager must have a convincing story to tell executives in
D-3419-1 34
1
ie s
ais Ze ieiatn =
2 Sa2--- 7 7" Spertormence on Soles Objective
8 9 Jo
8 2
. Soles Objective do
Soles ond Soles Objective (Units / Mon!
300
Performance and Motivation (Dimensionless Indices} =-==—
8 2
St =
3 \soies
3
A 2 1 1 duu
0 10 20 30 40 507
Months
Figure 8, Adjustment of Sales Objective Loop 2 in Isolation
D-3419-1 35
order to negotiate a reduction in his sales commitment without
loss of face. Routine and authority therefore result in
considerable inertia of the sales objective. Nevertheless, the
objective does respond in a rational though cautious way,
yielding more than 60% of the sales decline in twelve months.
Bounded Rationality and Inefficiency in the Full system
The two previous simulation experiments show that overtime
and the sales objective adjust in a plausible and intuitively
obvious way to an unexpected increase in the difficulty of
selling.
Figure 9 shows the adjustment of the complete system of four
interacting loops--the two loops already examined in partial
model tests, and the two new loops (3 and-4 in figure 6) opened
by activating the behavioral motivation function.1© the
adjustment is grossly inefficient. The sales organization
becomes locked into a trap in which sales are well below
potential and effective sales effort falls below the standard
that can be achieved with no overtime.
Why should the apparently reasonable decision rules of
market managers and salesmen fail so noticeably in the more
complex environment? Why is the system incapable of adjusting to
the new but lowered market potential without first passing
through a phase of more than two years where salesmen are
operating well below potential?
D-3419-1 36
A careful scrutiny of figure 9 provides insight into the
difficulties of managing the complete system. Monthly sales fall
by 1/3 in month 4, thereby opening a large gap between sales and
the objective. Salesmen put in more overtime, increasing
effective sales.effort and so preventing further decline in their
sales performance, as shown-by the leveling off of performance
between months 4 and 6. So far the adjustment makes sense;
however, two unforeseen problems are occurring.
First, the high level of overtime coupled with low
performance causes a sharp decline in motivation. Compounding
this problem, the sales objective itself does not fall as quickly
as it did in isolation, because the efforts of salesmen are
masking the full decline in, the market. (To illustrate this
point, the figure shows superimposed’ the sales objective as it
was in isolation.)
By month 10 of the simulation run, motivation has depressed
sales effort below the effort available from a well-motivated
force working no overtime! Consider now the rationality of the
salesman and market manager decisions, The salesman, pressured
by the sales objective, continues to work long hours even though
his effective effort in the market falls. The result is a
further decline in sales. The market managers are now very
confused. They have been cautiously lowering their sales
objective as they learn about the tighter market conditions.
D-3419-1
emer?
1200
Performance on
Soles. Objective
Motivation
eet Objective from Figure 3} |
600
Sales and Soles Objective (Units / Month)
Performance and Motivation (Dimensionless Indices!
Overtime
14
12
&
=
a
3
= <
~ Ss
i 3
a Effective Lf
6 8 Soles Effort <
3 3 Stondord Soles Effort £
3
3 g
§ ©
& Sb LOST EFFORT 42
pF
M 3
Bs 4
2 1 Fl : !
a 86 10 20 30 40 50
Months
Pigure 9. A Productivity and Sales Trap in the Full system
D-3419-1 38
But, starting in month 9, sales ‘begin to decline still further.
The objective-setting process. cannot distinguish the fall in
sales caused by the market from the fall caused by lowered sales
force motivation and productivity. Cautious downward adjustment
of the objective keeps the pressure on the salesmen--usually a
sensible thing to do. But in the prevailing situation continued
pressure lowers rather than raises the effective effort of
salesmen, There has been a complete breakdown in the logic of
sales management and control process.
Sales continue to decline until month 20. The system is in
a trap. It has been managed, or rather mismanaged, into a
situation whére the productivity of each salesman is much lower
than normal and sales are below potential. A recovery occurs
gradually after month 20, when motivation -and productivity have
reached rock bottom and the sales objective falls low enough to
relieve workload and performance pressure on the salesmen. But,
as the shaded areas in the figure show, there has been a major
loss of sales and much wasted sales effort.
The feedback structure of the system, the set of four
interlocking, nonlinear loops, makes sales management a hazardous
task. When only moderate changes in market conditions occur, the
overtime and objective-setting functions work effectively
together. (A small, say, 20% increase in time per sale causes a
temporary increase in overtime and a gradual relaxation of sales
D-3419-1 39
objectives--with no hint of a productivity or sales trap.)
However, a much larger increase in time per sale activates the
nonlinear motivation loops.
When these. loops become dominant, they reverse the normal
response of salesmen to pressure from the sales objective.
Instead of increasing their effort through overtime, salesmen
work longer, but much less effectively--a result entirely in
violation of the premises of the objective-setting process.
Under these circumstances the market managers make dysfunctional
decisions. Failing to meet their sales commitment, they
(unwittingly) set objectives that guarantee a still larger
discrepancy between future sales and commitment. Their
decisionmaking, though intendedly rational, is not sufficiently
close to objectively rational to account for.the large changes in
Salesman productivity caused by the highly nonlinear motivation
loops acting in concert.
SUMMARY AND CONCLUSIONS
The previous section has shown how a description of the premises
of decisionmaking followed by partial model testing can aid the
interpretation of a system dynamics behavioral simulation model.
But what do these methods of analysis provide that normal methods
of equation description and simulation analysis cannot?
D-3419-1 40
Clarifying the Theory Implicit in a Model
Normal methods of description and analysis leave a large gap
in logic (and therefore in the theory) between the assumptions
embodied in individual equations and the simulated behavior that
results from combining the equations in a simulation model.
Premise description and partial model tests bridge this gap.
Premise description relates the information content of
decision functions to factoring, routines, traditions, and
biaSes--in other words, to known and enpirically observed
organizational processes. Premise description specifies the
bounds on rational adjustment in the model and is the first step
in exposing the model's theory of behavior. For example, in the
sales model the sales objective-setting process of market
managers was quite myopic--its "area’ of adjustment" was bounded
by the routine of sales forecasting and by executive bias
transmitted through the sales commitment. The ayopia of
objective setting, which was embodied in several equations, was
important in explaining the unnecessary loss of sales following a
hardening of the market.
Partial model testing relates the premises of decisionmaking
to simulated behavior and is the second step in exposing the
model's theory of behavior. If we accept that business decision-
making is intendedly rational, then we should expect partial
model tests to reveal behavior that is intuitively clear and
D-3419-1 al
consistent with respect to the premises of the model's decision
functions. So, as we saw in the sales model, when motivation is
held constant the overtime function always adjusts so that
salesmen put in more effort when sales, performance falls below
objective--the intuitively correct response. A comparison of
partial and whole model tests provides an explanation for why
dysfunctional or counterintuitive behavior occurs.!7
Precision of Formulation and Policy Analysis
The understanding acquired from premise description and
partial. model testing can be helpful in justifying model
formulations and selecting between alternative formulations. For
example, awareness of the myopia (and its consequences) in the
objective-setting function naturally prompts the question of why
a more “intelligent” function is not’ in use.. why don't market
managers learn more quickly about changes in market conditions
and integrate them into sales objectives?. One possible answer is
that perhaps they do learn quickly, but the process has just not
been adequately modeled. Another answer is that the information
required to improve rationality is simply not available. Yet
another is that the information is available but is ignored.
Fear of renegotiating a sales commitment may block real knowledge
about changing market conditions, In any case, the modeler is
prompted to scrutinize his basic assumptions.
D-3419-1 42
Premise description and partial model testing are also -
helpful in policy design. An. understanding of the conditions
that cause a breakdown in the rationality of a given decision
function and a subsequent problem in the system may well point to
the changes necessary to remedy the problem. For example, in the
sales model a policy change that assumes market managers know and
act on motivation information greatly reduces the likelihood of
being caught in the productivity and sales trap. Alternatively,
a policy change that assumes market managers have instantaneous
and detailed knowledge of market conditions, and use the
knowledge to renegotiate their sales commitment, avoids the trap.
In conclusion, premise description and partial model testing
provide powerful diagnostic, tools for simulation modeling that
can improve the quality of model formulation.and analysis and
help clarify the theory implicit in the model to both academic
and managerial audiences.
D-3419-1 43
" NOTES
1. A fascinating account of the failure of the Saturday Evening
Post is provided in Harvard Business School Case Study
9-373-009 (1972). An interesting model-based theory of the
collapse is provided in Hall (1976).
2. See, for example, Forrester (1961) chapters 2 and 15, Lyneis
(1980) chapter 7, and Coyle (1977) chapter 10.
3. Much of Simon's Administrative Behavior (1976) is devoted to
showing first that actual human rationality departs from
objective rationality and, second, that organizations are
intended to place their members “in a psychological
environment that will adapt their decisions to the
organization objectives, and will provide them with the
information needed to make these decisions correctly.”
Chapter 10, "The Anatomy of Organization," provides a useful
summary of the central thesis.
4, For a picture of the goal-oriented decision process, see
Forrester (1961), p- 95.
5. For a detailed account of cognitive limitations, see Hogarth
(1980).
6. For a thoughtful discussion of problems of computer
simulation and theory building, see Frijda (1967). The paper
discusses the modeling of psychological processes but makes a
number of interesting general observations about the
strengths and weaknesses of simulation as a tool of theory
creation. Bell and Senge (1980), Forrester and Senge (1980),
and Mass and Senge (1978) discuss some thought-provoking
issues in the validation of model-based theories.
7. This kind of formulation description was used in Morecroft
(1983,1), though not as an explicit descriptive aid.
8. Optimal is put in quotes here because the optimality holds
only within the bounds set by the simplifying assumptions of
the optimizing algorithm. Like any decision function, the
algorithm itself has limits to its rationality set in this
case by the assumed constraints on the availability of
capacity, labor, and overtime.
9. Decisionmaking that is rational given its premises is
intendedly rational with respect to its environment. Simon
(1976, p- xxviii) has made the following interesting
assertion on the assumption of intended rationality and its
relationship to theory creation:
D-3419-1 44
10.
ll.
12.
13.
14.
is.
16.
It is precisely in the ‘realm where human ‘behavior
is intendedly rational...that there is room for a
genuine theory of organization and administration
For another example of partial model testing used to examine
rationality, see Morecroft (1983,1).
See Mass (1981) for further discussion of the process of
Giagnosing surprise model behavior.
There were in addition 45 accounting equations, which were
not part of the feedback structure, and 35 supplementary
equations,
For a discussion of policy structure diagrams and their
relationship to other diagraming methods in system dynamics,
see Morecroft (1982).
The actual model was written in the DYNAMO simulation
language (1976). DYNAMO equations are very similar to
discrete difference equations with one important difference:
DYNAMO allows independence of the time unit of description in
the modeled system from the time unit of computation. The
description of the sales model is in terms of months, but the
simulation interval is in weeks. When the time units of
description and computation are equal, then the DYNAMO and
discrete difference equations are identical. Unfortunately,
in many situations this restriction causes integration error
in numerical computations during simulation (Forrester 1961,
pp. 403-406), In other words, the behavior of the system
becomes sensitive to the computation interval. Such
sensitivity is an undesirable and misleading feature in a
system dynamics model (though in pure discrete difference
equation models it might be a perfectly acceptable feature).
The reader should therefore treat the difference equation
format as an approximation of the DYNAMO format, and realize
that in all simulation runs to be presented, the time unit of
computation is one week, much smaller than the one-month time
unit of description.
This is a significant hardening of the market, but quite
plausible. For example, in the full-scale project model
(Morecroft 1983,2) the market was being converted from old to
new technology. A sudden hardening could occur when sales
had been made to all the easy-to-convince customers, leaving
only the die-hards in the old technology. Precisely when
such a transition would occur was very difficult to predict.
In the model: the motivation fuction is activated by setting
the switch for motivation index SMI to 1 in equation 7.1 of
the text. In the base model SMI is set to 0.
D-3419-1 45
17. For another example of how partial model tests lead to a.
clarification of model theory, see Sterman's (1983)
explanation of the causes. of the so-called Kondratieff
long-wave economic cycle.
D-3419-1
‘APPENDIX
SALES IS A SIMPLIFIED MODEL BASED ON THE
SALMOD SERIES AND USED AS THE EXAMPLE IN THE
PAPER ‘RATIONALITY AND STRUCTURE IN
BEHAVIORAL MODELS OF BUSINESS SYSTEMS
BY JOHN D.W. MORECROFT, FEBRUARY 1983
SALES
MS.K=ESE.K/TPS.K
MS - MONTHLY SALES (UNITS PER MONTH) <1>
BSB - - _EFFECTIVE SALES EFFORT (HOURS PER NONTH) <3>
TPS -- ‘TIME PER SALE (HOURS PER UNIT) <2>
TPS. K=NTPS*(1+STEP(STPS, TSTPS) )
NTPS=60
STPS=0
TSTPS=4
TPS - TIME PER SALE (HOURS PBR UNIT) <2>
NTPS - NORMAL TIME PER SALE (HOURS PER UNIT) <2>
STPS -.- STEP IN TIME PER SALE (DIMENSIONLESS) ¢2>
TSTPS - TIME FOR STPS (MONTHS) <2
SALES EFFORT AND OVERTINE
ESE. K=SSE.K*MOT.K*EMSE.K
ESE © - EFFECTIVE SALES EFFORT (HOURS PER MONTH) <3>
SSE - STANDARD SALES EFFORT (HOURS PER MONTH) <4>
MOT = MULTIPLIER FROM OVERTIME (DIMENSIONLESS) <6>
ENSE - EFFECT OF MOTIVATION ON SALES EFFORT .
(DIMENSIONLESS) <10>
SSE.K=SF.K*NHSM
SSB - STANDARD SALES EFFORT (HOURS PER MONTH) <4>
SF - SALES FORCE (MEN) <5>
WHSM = NORMAL HOURS PER SALESMAN MONTH (HOURS PER
SALESMAN PER MONTH) <5>
SF.K=1SF
ISF=400
NHSM=130
SF - SALES FORCE (MEN) <5>
IsF - INITIAL SALES FORCE (MEN) <5>
NHSM = NORMAL HOURS PBR SALESMAN MONTH (HOURS PER
SALESMAN PER MONTH) <5>
MOT. K=TABLE( THOT, PSO.K,.75,1.1, +05)
PMOT=1.4/1.4/ 1% 55/16 25/1.4/1/1/
WO? = KULTIPLIER FROM OVERTIME (DIMENSIONLESS) <6>
THOT - TABLE FOR MULTIPLIER FROM OVERTIME <6>
PSO = - PERFORMANCE OW SALES OBJECTIVE (DIMENSIONLESS)
Documented Listing of Sales Model
46
D-3419-1
PSO.
PSO=IPSO
=PSO. J+(D2/TPSO) ((MS.J/MSO.J)-PSO.J)
~ TPS0=1/(1+MASC)
TPSO=3
PSO
- PERFORMANCE ON SALES OBJECTIVE (DIMENSIONLESS)
<T>
COMPUTATION INTERVAL OF SIMULATION (MONTHS) <14>
TIME FOR PERFORMANCE ON SALES OBJECTIVE (MONTHS)
SP 29,
~- MONTHLY SALES’(UNITS PER MONTH) <1>
~ MONTHLY SALES OBJECTIVE (UNITS PER MONTH) <8>
~ INITIAL PERFORMANCE ON SALES OBJECTIVE
(DIMENSIONLESS) <7>
MARGIN FOR ACHIEVEMENT OF SALES COMMITMENT
(DIMENSIONLESS) <a>
OBJECTIVE SETTING
MSO. K=MSC.K*(1+MASC)
MASC=.05
MSO
MSC
MASC
~ MONTHLY SALES OBJECTIVE (UNITS PBR MONTH) <B>
“= MONTHLY SALES COMMITMENT (UNITS PER MONTH) <9>
- MARGIN FOR ACHIEVEMENT OF SALES COMMITMENT
(DIMENSIONLESS) <8>
MSC. K=MSC.J+(DT/TESC) (MS. J-MSC.J)
MSC=NS
TESC=12
MSC - MONTHLY SALES COMMITMENT (UNITS PER MONTH) <9>
Dr - COMPUTATION INTERVAL OF SIMULATION (MONTHS) <14>
TESC - TIME TO ESTABLISH SALES COMMITMENT (MONTHS) <g>
MS - MONTHLY SALES (UNITS PER MONTH) <1>
MOTIVATION
EMSE.K=TABLE( TENSE, M.K,0,1,.2)
TEMSE=. 4/.5/.65/.85/.95/4
EMSE
TENSE
4
- EFFECT OF MOTIVATION ON SALES EFFORT
(DIMENSIONLESS) <10>
- TABLE FOR EFPECT OF MOTIVATION ON SALES EFFORT
<10>
- MOTIVATION (DIMENSIONLESS) <11>
MW. K=M.J+(DT/TEM) (MI. J-M.3)
M=IML
TEM=3
u
pr
TEM
MI
IML
- MOTIVATION (DIMENSIONLESS) <11>
- COMPUTATION INTERVAL OF SIMULATION (MONTHS) <14>
- TIME T0 ESTABLISH MOTIVATION (MONTHS) <11>
- MOTIVATION INDEX (DIMENSIONLESS) <12>
~ INITIAL MOTIVATION INDEX (DIMENSIONLESS) <12>
Documented Listing of Sales Model (cont.)
47
4,10
T1064
D-3419-1 48
MI. K=(W1O.K*MIP. K)*SMI+(1-SMI)*IMT Ay12
SMI=0 8 C,12.1
INI=HIO*MIP Ny 12.2
MI - MOTIVATION INDEX (DIMENSIONLESS) <12>
MIO = MOTIVATION INDEX FROM OVERTIME (DIMENSIONLESS
<13>
MIP ~~ -MOPIVATION INDEX FROM PERFORMANCE
(DIMENSIONLESS) <14>
SMI - _ SWITCH FOR MOTIVATION INDEX (DIMENSIONLESS) <12>
IMI ~ INITIAL MOTIVATION INDEX (DIMENSIONLESS) <12>
MIO. K=TABLE(TMI0,MOT.K,.8,1.5,.1) AL13
ee a 9/.7/-4/ 4.3 1,131
MIO = MOBTYARTON 7 rwDEK FROM OVERTIME (DIMENSIONLESS)
13>
oMIO- TABLE FOR MOTIVATION INDEX FROM OVERTIME <13>
MOT © - MULTIPLIER FROM OVERTIME (DIMENSIONLESS) <6>
MIP. K=TABLE( TMIP, PSO.K,.5,1.2,+1) Asta
mips 4/- 45/-6/-75/.95/1/1/4 & PB, 14.4
~ MOTIVATION INDEX FROM PERFORMANCE
(DIMENSIONLESS) <14>
TMIP - TABLE FOR MOTIVATION INDEX FROM PERFORMANCE <1 4>
PSO - PERFORMANCE ON SALES OBJECTIVE (DIMENSIONLESS)
<7? 3
SPEC LENGTH=0/DT=. 25/PLTPER=1/PRTPER=0 14.4
LENGTH - LENGTH OF SIMULATION RUN (MONTHS) <14>
DT - COMPUTATION INTERVAL OF SIMULATION (MONTHS) <14>
PLIPER - PLOT PERIOD (MONTHS) <14>
PRIPER - PRINT PERIOD (MONTHS) <14>
PRINT ESE,SSE,M,MSO,MS, PSO, MOT, EMSE 14.5
PLOT ESE, SSE(30E3,90E3)/MOT(.6,1.4) 14.6
PLOT MS,MSO(400,1200)/PSO,M(-1,1) 14.7
BSE © - EFFECTIVE SALES EFFORT (HOURS PER MONTH) <3>
SSE - STANDARD SALES EFFORT (HOURS PER MONTH) <4>
x - MOTIVATION (DIMENSIONLESS) <11>
MSO —---- MONTHLY SALES OBJECTIVE (UNITS PER MONTH) <B>
us - MONTHLY SALES (UNITS PER MONTH) <1>
PSO - PERFORMANCE ON SALES OBJECTIVE (DIMENSIONLESS)
<T>
Mo? = - MULTIPLIER FROM OVERTIME (DIMENSIONLESS) <6>
EMSE - EFFECT OF MOTIVATION OW SALES EFFORT
(DIMENSIONLESS) <10>
Documented Listing of Sales Model (cont.)
D-3419-1
RUN COMPILE
LENGTH=50
SMI=1
STPS=.5
LENGTH - LENGTH OF SIMULATION RUN (MONTHS)’ <14>
SMI - SWITCH FOR MOTIVATION INDEX (DIMENSIONLESS) <12>
STPS - STEP IN TIME PER SALE (DIMENSIONLESS) <2>
RUN BASE
TESC=3
pEsc - TIME TO ESTABLISH SALES COMMITMENT (MONTHS) <9>
RUN FLEX OBJ
PMOT=1.0952/1.0952/1.0952/1.0952/1.0952/1. Carel hosert. aaaal
THOT - TABLE FOR MULTIPLIER FROM OVERTIME <6
RUN OBJ SETTING WITH MOTIVATION
SHI=0
aNOre 1-0952/1.0952/1.0952/1.0952/1.0952/1.0952/1.0952/1.0952
SMI "= SWITCH FOR MOTIVATION INDEX (DIMENSIONLESS) <12>
THOT - TABLE FOR MULTIPLIER FROM OVERTIME <6>
RUN OBJ SETTING ONLY
TESC= 1086
LENGTH=10
PLIPER=. 25
TSTPS=1 a
TESC - TIME TO ESTABLISH SALES COMMITMENT (MONTHS) <9>
LENGTH - LENGTH OF SIMULATION RUN (MONTHS): <44>
PLIPER - PLOT PERIOD (MONTHS) <14>
TSTPS - TIME FOR STPS (MONTHS) <2>
RUN OVERTIME ADJUSTMENT ONLY
Documented Listing of Sales Model (cont.)
49
D-3419-1 50
REFERENCES
Allison, G. T. (1971), Essence of Decision. Little, Brown and
Company, Boston MA, 1971, pp. 79-81
Bell, J. A. and P. M. Senge (1980), "Methods for Enhancing
Refutability in System Dynamics Modeling," System Dynamics (in
‘TIMS Studies in the Management Sciences, Vol. 14, 1980, pp.
61-73).
Bitran, G. R. and A. C. Hax (1977), "On the Design of
Hierarchical Production Planning Systems," Decision Sciences,
Vol. 8, No. 1, January 1977, pp 28-55.
Coyle, R. G. (1977), Management System Dynamics. John Wiley and
Sons, Chichester, Englands 1STTC
Cyert, R. M. and J. G. March (1983), A Behavioral Theory of the
Firm, Prentice-Hall, Englewood Cliffs NJ, 1963. aa
Forrester, Jay W. (1961), Industrial. Dynamics, The MIT Press,
Cambridge MA, 1961.
Forrester, Jay W. (1968,1), "Market Growth as Influenced by
Capital Investment," Sloan Management Review, Vol. 9, No. 2,
pp. 83-105, Winter 1968.
Forrester, Jay W. (1968,2), "Industrial Dynamics--A Response to
Ansoff and Slevin," Management Science Vol. 14, No. 9, pp.
601-618, May 1968.
Forrester, Jay W. and P, M. Senge (1980), “Tests for Building
Confidence in "System Dynamics Models," System Dynamics (in
TIMS Studies in Management Sciences, Vo y 1980, pp.
209-228.
Hall, R. I. (1976), "A System Pathology of an Organization: The
Rise and Fall of the Old Saturday Evening Post," Administrative
Science Quarterly, Vol. 21, 1976, pp. 185-211.
HBS Case Services (1972), ‘The Saturday Evening Post, Case
9-373-009, Harvard Business School, Boston MA, 1972.
Hogarth, R. M. (1980), Judgement and Choice, John Wiley and Sons,
New York, 1980.
Holt, C. C., F. Modigliani, J. F. Muth, and H. A. Simon (1960),
Planning Production Inventories and Workforce, Prentice-Hall,
Englewood Cliffs NJ, 1960.
D-3419-1 51
Lyneis, J. M. (1980), Corporate Planning and Policy Design, The
MIT Press, Cambridge MA, 1980.
Mack, Ruth P, (1967)., Information, Expectations and Inventory
Fluctuation, National Bureau of Economic Research, 1967.
Mass, N. J. (1981), "Diagnosing Surprise Model Behavior," System
Dynamics Group Working Paper D-3323, Sloan School of
Management, M.I.T., Cambridge MA, October 1981.
Mass, N. J. and P, M. Senge (1978), “Alternative Tests for the
Selection of Model Variables," IEEE Systems, Men and
Cybernetics, June 1978.
Morecroft, J. D. W. (1982,3), "A Critical Review of Diagramming
Tools for Conceptualizing Feedback System Models,” Dynamica,
Vol. 8, Part I, Summer 1982, pp. 20-29.
Morecroft, J. D. W. (1983,1), "System Dynamics: Portraying
Bounded Rationality," Omega, Vol. 11, No. 2, March 1983, pp.
131-142.
Morecroft, J. D. W. (1983,2), “Strategy Support Models," System
Dynamics Group Working Paper D-3365-1, Sloan School of
Management, M.I.T., Cambridge MA, March 1983.
Nelson, Richard R. and Sidney G. Winter (1982), An Evolutionar;
Theory of Economic Change, Belknap Press of hatvard: University
Press, Cambridge ge WR, 1982 7 :
Pugh, A. L. (1976), Dynamo Users Manual, Fifth Edition, The MIT
Press, Cambridge MA, 1976. a
Richardson, G. P. and A. L. Pugh (1981), Introduction to System
Dynamics Modeling with DYNAMO, The MIT Press, Cambridge crea
rset
Simon, H. A. (1976), Administrative Behavior, Third Edition, The
Free Press (A Division of MacMillan Publishing Co.), New York,
1976.
Simon, H. A. (1979), "Rational Decision Making in Business
Organizations," The American Economic Review, Vol. 69, No. 4,
September 197
Simon, H. A, (1982), Models of Bounded Rationality (Vol 2):
Behavioral Economics and Business Organization, The MiT Press,
Cambridge MA, 1982.
Sterman, J. D. (1983), "A Simple Model of the Economic Long
Wave," System Dynamics Group Working Paper D-3410, Sloan School
of Management, M.I.T., Cambridge MA 02139, February 1983.
