Professional Documents
Culture Documents
SYSTEMS
THINKING TOOLS
A Users Reference Guide
BY DANIEL H. KIM
T H E T O O L B O X R E P R I N T S E R I E S
ISBN 1-883823-02-1
Introduction 3
PART I: AN OVERVIEW 5
Systems Thinking as a Language by Michael R. Goodman 6
Levels of Understanding: Fire-Fighting at Multiple Levels 8
A Palette of Systems Thinking Tools 10
3
I N T R O D U C T I O N
e are in the midst of a changing of an agefrom the age of machines to the sys-
W tems age. Our past was defined by a view of the world as a machine that could be
understood by breaking it into smaller and smaller parts. In the machine age view of the
world, the parts are what is most importantby understanding each of the parts, we
build up our understanding of the larger wholes. In the systems view, it is the whole that
is most importantparts in isolation have no meaning in and of themselves. Systems
thinking embodies the idea that the interrelationships among parts relative to a common
purpose of a system are what is important.
There is nothing more powerful than an idea whose time has come. But ideas
without practical tools can take us only so far in making any meaningful changes that will
have an impact on the world. Systems thinking provides the ideas that can help us see the
world in new ways, as well as the tools that can help us take new actions that are systemic
and more effective. This booklet provides a basic introduction to the various tools of sys-
tems thinking that have been developed and used over the last 50 years.
ACKNOWLEDGMENTS
Much of this work has been developed over the years through the efforts of many system
dynamicists. Systems Thinking Tools: A Users Reference Guide, part of the Toolbox Reprint
Series, was created and compiled by Kellie Wardman OReilly.
adds to Y.
+ Level o Level
Gap Gap include in the diagram
s +
o A causal link between two variables,
_ where a change in X causes a change
Actual B Adjust- Actual B Adjust- flow regulator
Level ments Level ments
X subtracts from Y.
s +
Delay
AN OVERVIEW
SYSTEMS THINKING AS A
LANGUAGE
B Y M I C H A E L R . G O O D M A N
anguage has a subtle, yet powerful Diagrams also facilitate learning. perceptions of a problem into black-and-
L effect on the way we view the world. Studies have shown that many people learn white pictures that can reveal subtle differ-
English, like most other Western languages, best through representational images, such ences in viewpoint.
is linearits basic sentence construction, as pictures or stories. A systems diagram is Example: In one systems thinking
noun-verb-noun, translates into a world- a powerful means of communication course, a team of managers was working on
view of x causes y. This linearity predis- because it distills the essence of a problem an issue they had been wrestling with for
poses us to focus on one-way relationships into a format that can be easily remem- months. One manager was explaining his
rather than circular or mutually causative bered, yet is rich in implications and position, tracing through the loops he had
ones, where x influences y, and y in turn insights. drawn, when a team member stopped him.
influences x. Unfortunately, many of the Does that model represent your thinking
A systems diagram is a
most vexing problems confronting man- about this problem? he asked.
agers and corporations today are caused by The presenter hesitated a bit, reviewed
a web of tightly interconnected circular powerful means of his diagram, and finally answered, Yes.
communication because it
relationships. To enhance our understand- The first man, evidently relieved,
insights.
problem.
Systems thinking can be thought of as a Allows examination and inquiry.
language for communicating about com- Systems diagrams can be powerful means
plexities and interdependencies. In particu- for fostering a collective understanding of a
lar, the following qualities make systems Adds precision. The specific set of syn- problem. Once individuals have stated their
thinking a useful framework for discussing tactical rules that govern systems diagrams understanding of the problem, they can col-
and analyzing complex issues: greatly reduce the ambiguities and miscom- laborate on addressing the challenges it
Focuses on closed interdependencies. munications that can occur when we tackle poses. And by focusing the discussion on
The language of systems thinking is circu- complex issues. the diagrams, systems thinking defuses
lar rather than linear. It focuses on closed Example: In drawing out the relation- much of the defensiveness that can arise in
interdependencies, where x influences y, y ships between key aspects of a problem, a high-level debate.
influences z, and z influences x. causal links are not only indicated by Example: When carrying on a systems
Offers a visual language. Many of the arrows, but are labeled s (same) or o discussion, differing opinions are no longer
systems thinking toolscausal loop dia- (opposite) to specify how one variable viewed as human resources view of our
grams, behavior over time diagrams, sys- affects another. Such labeling makes the productivity problem or marketings
tems archetypes, and structural nature of the relationship more precise, description of decreasing customer satisfac-
diagramshave a strong visual component. ensuring only one possible interpretation. tion, but simply different structural repre-
They help clarify complex issues by sum- Forces an explicitness of mental mod- sentations of the system. This shifts the
ming up, concisely and clearly, the key ele- els. The systems thinking language trans- focus of the discussion from whether
ments involved. lates war stories and individual human resources or marketing is right, to
L E V E L S O F U N D E R S TA N D I N G :
F I R E - F I G H T I N G AT M U LT I P L E
LEVELS
ts another busy night in the four distinct levelsevents, patterns of THE SYSTEMS THINKER, V3N7). This
I hospital emergency room. Several events, systemic structure, and shared vision is consistent with our evolutionary history,
car accident victims have been rushed into (see Levels of Understanding). Events are which was geared toward responding to
surgery, one little boy is having a broken the things we encounter on a day-to-day those things that posed an immediate dan-
arm set, a drug overdose victim is being basis: a machine breaks, it rains, we eat din- ger to our well-being.
treated, and numerous other people fill the ner, see a movie, or write a report. Patterns Events require an immediate response.
chairs in the waiting room. Each night is of events are the accumulated memories of If a house is burning, we react by taking
different, and yet each one is also the same. eventswhen strung together in a series action to put out the fire. Putting out the
The doctors and nurses must act fast to over time, they reveal recurring patterns. fire is appropriate, but if it is the only action
treat the most seriously injured, while the Systemic structure can be viewed as event that is ever taken, it is inadequate from a
others wait their turn. Like an assembly generators because they are responsible for systemic perspective. Why? Although it
line of defective parts, patients are diag- producing the events. Similarly, shared solved the immediate problem (the burning
nosed, treated, and released. Each injury is vision can be viewed as systemic structure house), it has done nothing to alter the fun-
a crisis that demands immediate attention. generators because they are the guiding damental structure that caused that event
So whats wrong with this picture? force behind the creation or change of all (e.g., inadequate building codes, lack of fire
After all, isnt this what emergency rooms kinds of structures. detectors, fire prevention education). The
are meant to do? The answer depends on We live in an event-oriented world, and Levels of Understanding diagram and
the level of understanding at which we are our language is rooted at the level of events. framework can help us go beyond typical
looking at the situation. At work, we encounter a series of events, event-orientation responses and begin to
which often appear in the form of problems look for higher leverage actions.
LEVELS OF
that we must solve. Our solutions, how-
U N D E R S TA N D I N G FROM FIRE-FIGHTING TO
ever, may be short-lived, and the symptoms
There are multiple levels from which we FIRE PREVENTION
can eventually return as seemingly new
can view and understand the world. From a problems (see Using Fixes That Fail to At the event level, if a house is on fire, all
systemic perspective, we are interested in Get off the Problem-Solving Treadmill, we can do is react as quickly as possible to
put the fire out. The only mode of action
that is appropriate and available is to be
LEVELS OF UNDERSTANDING reactive. If we reacted to fires only at the
events level, we would put all of our energy
Levels of Understanding Action Mode Time Orientation Typical Questions into fighting firesand we would proba-
What are the stated or bly have a lot more fire stations than we do
Shared Vision Generative Future unstated visions that generate today.
the structures?
If we look at the problem of fires at the
What are the mental or
Systemic Structure Creative organizational structures that pattern of events level, we can begin to
create the patterns? anticipate where they are more likely to
What kinds of trends or occur. We may notice that certain neigh-
Patterns of Events Adaptive patterns of events seem to be
recurring?
borhoods seem to have more fires than oth-
ers. We are able to be adaptive by locating
What is the fastest way to
Events Reactive Present react to this event NOW? more fire stations in those areas, and
A PALETTE OF SYSTEMS
THINKING TOOLS
here is a full array of systems think- turn have sub sub-factors. Many layers of They serve as a starting point from which
T ing tools that you can think of in the nesting, however, may be a sign that one of one can build a clearer articulation of a
same way as a painter views colorsmany the sub-factors should be turned into a business story or issue. Specific archetypes
shades can be created out of three primary major factor. include: Drifting Goals, Shifting the
colors, but having a full range of ready- Burden, Limits to Success, Success to
DYNAMIC THINKING
made colors makes painting much easier. the Successful, Fixes That Fail,
TOOLS
There are at least 10 distinct types of Tragedy of the Commons, Growth and
systems thinking tools (an abbreviated sum- Behavior Over Time (BOT) Diagrams are Underinvestment, and Escalation (see
mary diagram appears on the facing page). more than simple line projectionsthey Systems Archetypes at a Glance, p. 20).
They fall under four broad categories: capture the dynamic relationships among
variables. For example, say we were trying STRUCTURAL THINKING
brainstorming tools, dynamic thinking
to project the relationship between sales, TOOLS
tools, structural thinking tools, and com-
puter-based tools. Although each of the inventory, and production. If sales jump 20 Graphical Function Diagrams, Structure-
tools is designed to stand alone, they also percent, production cannot jump instanta- Behavior Pairs, and Policy Structure
build upon one another and can be used in neously to the new sales number. In addi- Diagrams can be viewed as the building
combination to achieve deeper insights into tion, inventory must drop below its blocks for computer models. Graphical
dynamic behavior. previous level while production catches up Functions are useful for clarifying nonlin-
with sales. By sketching out the behavior of ear relationships between variables. They
BRAINSTORMING TOOLS different variables on the same graph, we are particularly helpful for quantifying the
The Double-Q (QQ) Diagram is based on can gain a more explicit understanding of effects of variables that are difficult to mea-
what is commonly known as a fishbone or how these variables interrelate. sure, such as morale or time pressure.
cause-and-effect diagram. The Qs stand for Causal Loop Diagrams (CLDs) provide a Structure-Behavior Pairs link a specific
qualitative and quantitative, and the tech- useful way to represent dynamic interrela- structure with its corresponding behavior.
nique is designed to help participants begin tionships. CLDs make explicit ones under- Policy Structure Diagrams represent the
to see the whole system. During a struc- standing of a systems structure, provide a processes that drive policies. In a sense,
tured brainstorming session with the QQ visual representation to help communicate when we use these tools we are moving
diagram, both sides of an issue remain that understanding, and capture complex from painting on canvas to sculpting three-
equally visible and properly balanced, systems in a succinct form. CLDs can be dimensional figures.
avoiding a top-heavy perspective. The combined with BOTs to form structure-
COMPUTER-BASED TOOLS
diagram also provides a visual map of the behavior pairs, which provide a rich frame-
key factors involved. Once those factors are work for describing complex dynamic This class of tools, including computer
pinpointed, Behavior Over Time Diagrams phenomena. CLDs are the systems thinkers models, management flight simulators, and
and/or Causal Loop Diagrams can be used equivalent of the painters primary colors. learning laboratories, demands the highest
to explore how they interact. Systems Archetypes is the name given to level of technical proficiency to create.
A QQ diagram begins with a heavy certain common dynamics that seem to On the other hand, very little advance
horizontal arrow that points to the issue recur in many different settings. These training is required to use them once they
being addressed. Major hard (quantita- archetypes, consisting of various combina- are developed.
tive) factors branch off along the top and tions of balancing and reinforcing loops, are
soft (qualitative) factors run along the the systems thinkers paint-by-numbers
bottom. Arrows leading off of the major setusers can take real-world examples
factors represent sub-factors, which can in and fit them into the appropriate archetype.
f(x)
A
B
Time x
Can be used to graph the behavior of Captures the way in which one variable Lets you translate all relationships
variables over time and gain insights into affects another, by plotting the relation- identified as relevant into mathematical
any interrelationships between them. ship between the two over the full range equations. You can then run policy
(BOT diagrams are also known as of relevant values. analyses through multiple simulations.
reference mode diagrams.)
COCKPIT
s B DECISION INFO
o STOCK
HIRING
STOCK
R C
B HIRING
s
A
s Time
Reflection
Experimentation
Helps you recognize common system A conceptual map of the decision-making A managers practice field. Is equivalent
behavior patterns such as Drifting process embedded in the organization. to a sports teams experience, which
Goals, Shifting the Burden, Limits to Focuses on the factors that are weighed blends active experimentation with
Growth, Fixes That Fail, and so on for each decision, and can be used to reflection and discussion. Uses all the
all the compelling, recurring stories of build a library of generic structures. systems thinking tools, from behavior
organizational dynamics. over time diagrams to MFSs.
B E H AV I O R O V E R T I M E D I A G R A M
SYSTEMS ARCHETYPES
Balancing loops try to bring a system to a desired state and keep it there. In an inventory con- Of course, most things in life cannot con-
trol system, the desired inventory is maintained by adjusting the actual inventory whenever
there is too much or too little.
tinue growing forever. There are other
forcesbalancing loopsthat resist fur-
To see how these two basic loops can com- Structure Behavior Over Time
bine to form more complex structure-
Energy s
behavior pairs, lets revisit the Level Perf.
Burnout
o Employee
employee-supervisor feedback loop. s
Performance Level
R
Clearly the employees performance will B2 R1
not improve indefinitely just because the Hours s B
Worked
supervisor is supportive. The employee Supervisors Positive Diminishing
s Supportive Reinforcement Returns
may have been putting in longer hours in Behavior Time
order to continue impressing the supervi-
Reinforcing and balancing loops can be combined to describe more complex behavior. For
sor. Over a period of time, the increased
work hours may begin to wear down the example, encouragement by the supervisor could lead the employee to work longer and longer
hours in order to continue impressing the supervisor, eventually leading to burnout and a
decrease in performance.
employees energy level (see Reinforcing
Loop Coupled with a Balancing Loop
B A L A N C I N G L O O P S W I T H D E L AY S :
T E E T E R - T O T T E R I N G O N S E E S AW S
ost of us have played on a seesaw at The goal of a seesaw ride is to always process are produced by two balancing
M one time or another and can recall keep things in a state of imbalance (it loops that try to stabilize on a particular
the up and down motion as the momen- would be pretty boring to sit on a perfectly price. But the process is complicated by
tum shifted from one end to the other. The balanced one). But the goal in the market- the presence of significant delays (see
more equal the weights of both people, the place is exactly the oppositeto bring Supply and Demand).
smoother the ride. At a very basic level, a supply in balance with demand. Unfortu-
free market economy is a lot like a seesaw B A L A N C I N G S U P P LY A N D
nately, the supply and demand balancing
DEMAND
with supply at one end and demand on the process feels a lot more like a seesaw ride
other end. Prices indicate the imbalance than a smooth adjustment to a stable equi- Tracing through the loops you can see that
between the two, like a needle positioned librium. As shown in a causal loop dia- if demand rises, price tends to go up (all
at the pivot point of the seesaw. gram, the dynamics of this adjustment else remaining the same), and as price goes
up, demand tends to go down (Beanie
S U P P LY A ND D E M A N D
Babies notwithstanding). If there is enough
inventory or capacity in the system to
Price absorb the increased demand, prices may
not go up immediately. As demand out-
Lo Hi strips supply, however, price will rise.
nd
Dema On the supply side of the seesaw, an
increase in price provides a profit incentive
for firms to produce more. Of course, it
Supply
takes time for firms to expand. The length
of the delay depends on how close they
lay
o s De already are to full capacity and how
Supply Price Demand
quickly they can add capacity to produce
B1 B2
more. Hiring new workers may take only
s o a few days, while obtaining additional cap-
lay
De ital equipment or factory floor space may
take months or even years. While firms are
making supply adjustments, the gap
between supply and demand widens and
Demand
price goes even higher. The higher price
spurs companies to increase their produc-
tion plans even more.
Supply
As supply eventually expands and
catches up with demand, price begins to
Time
fall. By this time, firms have overexpanded
A free market economy is a lot like a seesaw with supply at one end and demand on the other.
their production capacity and supply over-
The dynamics that result from trying to balance supply and demand are produced by two balanc- shoots demand, causing price to fall. When
ing loops that try to stabilize on a particular price. Due to the presence of significant delays, a
cycle of overshoot and collapse occurs.
the price falls low enough, the product
becomes more attractive again and
G U I D E L I N E S F O R D R AW I N G
CAUSAL LOOP DIAGRAMS
he old adage if the only tool you Time horizon. It is also helpful to deter- Boundary issue. How do you know
T have is a hammer, everything begins mine an appropriate time horizon for the when to stop adding to your diagram? If
to look like a nail can also apply to lan- issueone long enough to see the dynamics you dont stay focused on the issue, you
guage. If our language is linear and static, play out. For a change in corporate strategy, may quickly find yourself overwhelmed by
we will tend to view and interact with our the time horizon may span several years, the number of connections possible.
world as if it were linear and static. Taking while a change in advertising campaigns Remember, you are not trying to draw out
a complex, dynamic, and circular world may be on the order of months. the whole systemonly what is critical to
and linearizing it into a set of snapshots Time itself should not be included as a the theme being addressed. When in
may make things seem simpler, but we may causal agent, however. After a heavy rain- doubt, ask, If I were to double or halve
totally misread the very reality we were fall, a river level steadily rises over time, but this variable, would it have a significant
seeking to understand. Making such inap- we would not attribute it to the passage of effect on the issue I am mapping? If not,
propriate simplifications is like putting on time. You need to identify what is actually it probably can be omitted.
your brakes and then looking at your driving the change. In computer chips, Level of aggregation. How detailed
speedometer to see how fast you were $/MIPS (million instructions per second) should the diagram be? Again, the level
going, says Bill Isaacs of DIAlogos. decreased in a straight line in the 1990s. It should be determined by the issue itself.
would be incorrect, however, to draw a The time horizon also can help determine
A R T I C U L AT I N G R E A L I T Y
causal connection between time and how detailed the variables need to be. If the
Causal loop diagrams provide a language $/MIPS. Instead, increasing investments and time horizon is on the order of weeks (fluc-
for articulating our understanding of the learning curve effects were likely causal tuations on the production line), variables
dynamic, interconnected nature of our forces. that change slowly over a period of many
world. We can think of them as sentences Behavior over time charts. Identifying years may be assumed to be constant (such
that are constructed by linking together key and drawing out the behavior over time of as building new factories). As a rule of
variables and indicating the causal relation- key variables is an important first step thumb, the variables should not describe
ships between them. By stringing together toward articulating the current understand- specific events (a broken pump); they
several loops, we can create a coherent story ing of the system. Drawing out future should represent patterns of behavior
about a particular problem or issue. behavior means taking a riskthe risk of (pump breakdowns throughout the plant).
Following are some more general being wrong. The fact is, any projection of Significant delays. Make sure to identify
guidelines that should help lead you the future will be wrong, but by making it which (if any) links have significant delays
through the process: explicit, we can test our assumptions and relative to the rest of the diagram. Delays
Theme selection. Creating causal loop uncover inconsistencies that may otherwise are important because they are often the
diagrams is not an end unto itself, but part never get surfaced. For example, drawing source of imbalances that accumulate in
of a process of articulating and communi- projections of steady productivity growth the system. It may help to visualize pres-
cating deeper insights about complex issues. while training dollars are shrinking raises sures building up in the system by viewing
It is pointless to begin creating a causal loop the question, If training is not driving our the delay connection as a relief valve that
diagram without having selected a theme or growth, what will? The behavior over either opens slowly as pressure builds or
issue that you wish to understand better. time diagram also points out key variables opens abruptly when the pressure hits a
To understand the implications of chang- that should be included in the diagram, critical value. An example of this might be
ing from a technology-driven to a market- such as Training Budget and Productivity. a delay between long work hours and
ing-oriented strategy, for example, is a Your diagram should try to capture the burnout: After sustained periods of work-
better theme than To better understand structure that will produce the projected ing 60+ hours per week, a sudden collapse
our strategic planning process. behavior. might occur in the form of burnout.
Litigation
1 Use nouns when choosing a variable name. Avoid verbs and action phrases, because the
action is conveyed in the loops arrows. For example, Costs is better than Increasing
Costs, because a decrease in Increasing Costs is confusing. The sign of the arrow (s for same
Costs
Increasing Costs
or o for opposite) indicates whether Costs increase or decrease relative to the other variable.
Rewards
2 Use variables that represent quantities that can vary over time. It does not make sense to
say that State of Mind increases or decreases. A term like Happiness, on the other
hand, can vary.
Happiness
State of Mind
Demand
3 Whenever possible, choose the more positive sense of a variable name. For example,
the concept of Growth increasing or decreasing is clearer than an increase or decrease
in Contraction.
Growth
Contraction
s
4 Think of the possible unintended consequences as well as the expected outcomes for
every course of action included in the diagram. For example, an increase in Production
Pressure may increase Production Output, but it may also increase Stress and decrease
Production Pressure
s
Production Output
Stress
o
Quality. Quality, etc.
5 All balancing loops are goal-seeking processes. Try to make explicit the goals driving s Quality
s Quality
the loop. For example, Loop B1 may raise questions as to why increasing Quality
o
would lead to a decrease in Actions to Improve Quality. By explicitly identifying Desired B1 Desired
Quality as the goal in Loop B2, we see that the Gap in Quality is really driving improve- Quality
B2 Gap in
Quality
ment actions. Actions to Actions to s
Improve Improve
Quality
o Quality
s
LOOP CONSTRUCTION
s
6 Distinguishing between perceived and actual states, such as Perceived Quality
Dela
Actual
versus Actual Quality, is important. Perceptions often change slower than reality does, Quality s
y
and mistaking the perceived status for current reality can be misleading and create undesir- Perceived
Actions to Quality
able results. Improve
Quality B2 R1
o
s Gap in
s
Desired
Quality Quality
s
7 If a variable has multiple consequences, start by lumping them into one term while
completing the rest of the loop. For example, Coping Strategies can represent many
different ways we respond to stress (exercise, meditation, alcohol use, etc.).
s
Stress B Coping
Strategies
o
8 Actions almost always have different long-term and short-term consequences. Draw
larger loops as they progress from short- to long-term processes. Loop B1 shows the
short-term behavior of using alcohol to combat stress. Loop R2, however, draws out the long- Stress
s
B1 Alcohol
term consequences of this behavior, showing that it actually increases stress. o Use
o o
Productivity R2 Health
s
o
9 If a link between two terms requires a lot of explanation to be clear, redefine the vari-
ables or insert an intermediate term. Thus, the relationship between Demand and
Demand
s
Quality
o
GENERAL TIPS
Quality may be more obvious when Production Pressure is inserted between them. Demand
Production Quality
Pressure
o
10 A shortcut to determining whether a loop is balancing or reinforcing is to count the
number of os in the loop. An odd number of os indicates a balancing loop (i.e., an
odd number of U-turns keeps you headed in the opposite direction); an even number or no
Bank
Failures
o
Solvency R Depositors
os means it is a reinforcing loop. CAUTION: After labeling the loop, you should always Confidence
read through it to make sure the story agrees with your R or B label. o
Withdrawals o
from Banks
S Y S T E M S A R C H E T Y P E S AT A G L A N C E
ARCHETYPE DESCRIPTION GUIDELINES
Drifting Goals In a Drifting Goals archetype, a gap Drifting performance figures are usu-
o between the goal and current reality can ally indicators that the Drifting Goals
Goal
Pressure to be resolved by taking corrective action archetype is at work and that real cor-
Lower Goal
B2 (B1) or lowering the goal (B2). The rective actions are not being taken.
s
critical difference is that lowering the A critical aspect of avoiding a potential
s
o
Gap goal immediately closes the gap, whereas Drifting Goals scenario is to determine
s corrective actions usually take time. (See what drives the setting of the goals.
B1
Corrective
The Systems Thinker, October 1990.) Goals located outside the system will be
Actual
Action less susceptible to drifting goals pressures.
s lay
De
Escalation In the Escalation archetype, one To break an escalation structure, ask the
party (A) takes actions that are per- following questions:
ceived by the other as a threat. The What is the relative measure that pits
other party (B) responds in a similar one party against the other and can
s As Result Bs Result s
s o manner, increasing the threat to A and you change it?
Activity
B1 Quality of As Position B2
Activity resulting in more threatening actions What are the significant delays in the
by A Relative to Bs by B
s
by A. The reinforcing loop is traced system that may distort the true nature
s
Threat
o s
Threat out by following the outline of the fig- of the threat?
to A to B
ure-8 produced by the two balancing What are the deep-rooted assumptions
loops. (See The Systems Thinker, that lie beneath the actions taken in
November 1991.) response to the threat?
Fixes That Fail In a Fixes That Fail situation, a Breaking a Fixes that Fail cycle usu-
problem symptom cries out for resolu- ally requires acknowledging that the
s
tion. A solution is quickly imple- fix is merely alleviating a symptom,
Problem
Fix
mented that alleviates the symptom and making a commitment to solve
Symptom B1
(B1), but the unintended consequences the real problem now.
o
s of the fix exacerbate the problem A two-pronged attack of applying the
De (R2). Over time, the problem symptom fix and planning out the solution will
R2 lay
Unintended returns to its previous level or becomes help ensure that you dont get caught
Consequence s
worse. (See The Systems Thinker, in a perpetual cycle of solving yester-
November 1990.) days solutions.
Growth and Underinvestment In a Growth and Underinvestment Dig into the assumptions which drive
s s archetype, growth approaches a limit capacity investment decisions. If past
Growth that can be eliminated or pushed into performance dominates as a consider-
Effort R1 Demand B2
the future if capacity investments are ation, try to balance that perspective
s Performance
o
Performance
Standard made. Instead, performance standards with a fresh look at demand and the
s
o s
are lowered to justify underinvestment, factors that drive its growth.
Capacity B3 Perceived Need leading to lower performance which If there is potential for growth, build
to Invest
s De
further justifies underinvestment. capacity in anticipation of future
lay lay
De (See The Systems Thinker, June/July demand.
Investment s
in Capacity
1992.)
Limits to Success In a Limits of Success scenario, con- The archetype is most helpful when it is
tinued efforts initially lead to used well in advance of any problems,
improved performance. Over time, to see how the cumulative effects of
however, the system encounters a continued success might lead to future
Constraint
limit which causes the performance to problems.
s o s slow down or even decline (B2), even Use the archetype to explore questions
Efforts R1 Performance B2 Limiting as efforts continue to rise. (See The such as What kinds of pressures are
Action
s Systems Thinker, December building up in the organization as a
s
1990/January 1991.) result of the growth?
Look for ways to relieve pressures or
remove limits before an organizational
gasket blows.
Shifting the Burden/Addiction In a Shifting the Burden, a problem is Problem symptoms are usually easier to
solved by applying a symptomatic recognize than the other elements of the
Symptomatic solution (B1), which diverts attention structure.
Solution s away from more fundamental solutions If the side-effect has become the problem,
B1 (R3). (See The Systems Thinker, you may be dealing with an Addiction
s September 1990.) In an Addiction structure.
o Problem structure, a Shifting the Burden Whether a solution is symptomatic or
Symptom R3 Side-Effect
o fundamental often depends on ones per-
degrades into an addictive pattern in
Dela
y
B2 which the side-effect gets so entrenched spective. Explore the problem from a dif-
that it overwhelms the original problem fering perspective in order to come to a
Fundamental s
Solution o symptom. (See The Systems Thinker, more comprehensive understanding of
April 1992.) what the fundamental solution may be.
Success to the Successful In a Success to the Successful Look for reasons why the system was set
archetype, if one person or group (A) up to create just one winner.
is given more resources, it has a Chop off one half of the archetype by
Success s o Success higher likelihood of succeeding than focusing efforts and resources on one
of A of B B (assuming they are equally capable). group, rather than creating a winner-
s Allocation to A s
R1 Instead of B
R2 The initial success justifies devoting take-all competition.
more resources to A, and Bs success Find ways to make teams collaborators
Resources Resources
to A
s o to B diminishes, further justifying more rather than competitors.
resource allocations to A (R2). (See Identify goals or objectives that define
The Systems Thinker, March 1992.) success at a level higher than the individ-
ual players A and B.
Tragedy of the Commons In a Tragedy of the Commons struc- Effective solutions forTragedy of the
s
Net Gains
s
ture, each person pursues actions Commons scenario never lie at the
for A
which are individually beneficial (R1 individual level.
R1 and R2). If the amount of activity
B5 Ask questions such as: What are the
As o
grows too large for the system to sup- incentives for individuals to persist in
s Activity
s
Resource port, however, the commons their actions? Can the long-term col-
Limit
R3
Total Gain per becomes experiences diminishing ben- lective loss be made more real and
Individual
y
Activity
la
o s efits (B5 and B6). (See The Systems immediate to the individual actors?
De
Activity
s R4
Thinker, August 1991.) Find ways to reconcile short-term
s Bs
Activity o cumulative consequences. A governing
R2
body that is chartered with the sustain-
R2 B6
ability of the resources limit can help.
Net Gains s
for B
s
S T R U C T U R E - B E H AV I O R PA I R
tructural thinking tools can help us become even more explicit about the struc-
S tures that create the dynamic behaviors we are trying to understand. From
Causal Loops to Graphical Functions: Articulating Chaos (p. 24) and Graphical
Functions: Seeing the Full Story (p. 26) describe how graphical function diagrams
can easily represent nonlinear relationships. These relationships characterize the
nature of most interconnections in complex systems (as opposed to the simple, linear
relationships that are often assumed).
The second half of this section focuses on structure-behavior pairs. Accumulators
and flows provide a rigorous framework for representing systemic structures in a
more precise way than through causal loop diagrams alone. They can better represent
a systems nonlinearity, as well as distinguish between things that accumulate (water in
a bathtub) and things that flow (water flowing through a faucet). The articles on accu-
mulators and flows (p. 2837) show how these concepts add further precision to our
thinking and understanding about the link between structure and behavior.
tomers, and corporate reputation share a equation. And when we were done, it
Discr. Expenses
similar underlying structure (see would be hard to tell if the equation repre-
Accumulators: Bathtubs, Bathtubs sented our savings account or the number 0
Savings
20000
GRAPHICAL FUNCTIONS:
SEEING THE FULL STORY
n executive of a large automotive frame of reference for their discussion, each perspectives. Graphical functions can help
A company tells the story of two engi- had assumed the others viewpoint was wrong. us go beyond merely observing correla-
neers who were arguing about the correct tional relationships (when X happens, Y
angle of an engine mount. The two had INDIVIDUAL WORLDS happens) to exploring our understanding
been at it for more than half an hourone The story of the two engineers points out of the causal connection between two vari-
engineer swearing that the angle was 40 an age-old communication problem. Each ables (X causes Y). In constructing GFDs,
degrees while the other fumed that it was of us carries our own set of assumptions one should follow the 60 percent ruleits
50 degrees. After several civil attempts to about realityour own individual picture better to get it 60 percent right very
correct each others viewpoint, they had just of the world. Oftentimes, we mistakenly quickly and spend time modifying it than
started attacking each others intelligence, assume that our viewpoint is the only way spend a great deal of effort trying to get it
ability, and character when the executive of looking at a situation. Both engineers, 100 percent right the first time.
happened to walk by. for example, believed the other persons
What axis of reference are you position was based on the same axis as GRAPHICAL FUNCTIONS
using? he asked. their ownthey never even questioned it. V S . S C AT T E R C H A R T S
The vertical, of course! exclaimed If we dont acknowledge our assumptions Graphical functions are best described by
one engineer. at the outset of a discussion, we risk expe- first establishing what they are not.
The horizontal! said the other. riencing the same frustrations as the two Although they may look similar, graphical
Both stopped in amazement as they engineers. functions are not the same as scatter charts,
realized they had been saying the same thing! In many instances, spoken language which plot one variables data against
Because they had not established a common can be a hindrance rather than a help in anothers. If we were to look at the rela-
communicating our mental pictures of tionship between sales and delivery delay
reality because words, unlike pictures, do using a scatter chart, we would plot some
SALES VS. DELIVERY
D E L AY not force us to be explicit when explaining data points and then draw a regression line
our reasoning. Graphics, because they can through them (see Sales vs. Delivery
Sales represent ideas more clearly, can be a Delay).
($) Sales = 1000 - 25 * Delivery Delay
much more powerful and effective means From the scatter chart, we can see that
of communication (see Systems Thinking in weeks one through four, sales fall by
1000 K
as a Language, p. 6). Trite as it may $25K for each one-week delay. We can then
sound, the saying a picture is worth a extrapolate beyond the historical data to
900 K thousand words still holds true. Had the predict that a five-week delay will result in
two engineers simply drawn two axes and an additional $25K drop in sales. In general,
800 K a line, they would have saved a great many scatter diagrams answer the question,
angry words. When the issue is more com- What happened historically, was there a
plex than a single angle, the use of graph- correlation, and based on that information
0 1 2 3 4 ics can become even more important for what can I expect to happen in the future?
Delivery Delay (weeks)
reaching a shared understanding. They tend to be retrospective.
Scatter charts plot one variables data
Using graphical function diagrams A graphical function, on the other
against another and answer the question, (GFDs), it is much easier to capture how hand, is very much prospective in nature.
What happened historically? The view is
retrospective.
two variables relate in a format that is con- By including the full spectrum of possible
cise and invites others to share their own values, GFDs can help you see beyond the
ing the following questions: ery delay and sales than a Sales ($)
What do we know from the outset scatter chart based on histor- 1000 K
about the causal relationship between these ical data. The diagram helps 900 K
two variables? visualize the full range of 800 K
Are there any neutral zones where implications and minimizes
700 K
the variable on the y-axis is not affected by the danger of remaining
changes in the x-variable? myopically focused on a nar- 600 K
What are the extreme values that both row band of possible out- 500 K
variables can assume? comes. Developing the
400 K
If we looked at the sales and delivery delay diagram as a group can also
help surface differing mental 300 K
example using a GFD, we would start by
asking what we think the general nature assumptions about the poten- 200 K
0 20
of the relationship is between the two vari- tial impact of deteriorating Delivery Delay (weeks)
ablesis it flat, is it upward sloping, or is or improving delivery per-
it downward sloping? With most products, formance (remember the E F F E C T O F D E L I V E RY D E L AY
longer delivery delays mean lower sales, so engineers!). ON SALES
we can assume that the relationship would Sometimes it is helpful to
slope downward. convert the relationship into a
Effect of Delivery Delay on Sales
1.4
Using available historical data and past more general form where the Maximum Advantage
experience, we can then take a first cut at y-variable is converted to an 1.2
Neutral Zone
identifying a neutral zone where sales may effect-of variable. Instead 1.0
be insensitive to differences in the length of of Sales on the y-axis, for .8
the delay (see Delivery Delay Graphical example, we would have
.6
Function Diagram). Past experience may Effect of Delivery Delay on Maximum Disadvantage
suggest that sales will increase steadily as Sales (see graph), which .4
the delay falls below four weeks. A sam- shows that a 3.5 to 4.5 week .2
pling of customer contacts may tell us that delay has no effect, shorten-
0
there is not a whole lot of difference ing the delay nets us a maxi- 0 4 10 20
mum gain of 5 percent (1.05 Delivery Delay (weeks)
between 3.5 and 4.5 weeks. On the other
A graphical function diagram conveys a much richer descrip-
tion about the relationship between delivery delay and sales
hand, past market research also tells us that times the sales number we
(top) than a scatter chart based on historical data. Looking at
if the delay grows greater than five weeks, would have obtained if we
sales will fall dramatically. Looking fur- were in the neutral zone), the effect of delivery delay on sales (bottom) allows you to
and lengthening the delay to compare the relative effects of different variables, to see which
is the dominant driver.
ther, we realize that even in the extreme
case of a 20-week delay, $200K of sales will 20 weeks can choke sales by
STRUCTURAL THINKING:
THE WORLD ACCORDING TO
A C C U M U L AT O R S A N D F L O W S
vice president of a major U.S. man- Systems tells this story: While perusing a country and killed all the cows. What
A ufacturer once questioned whether well-known economic journal, I came would the above model predict for next
todays rapid pace of change means that all across an article which described a model years milk production? The answer would
our old tools and ways of managing are now that had been constructed to forecast U.S. most likely look a lot like the number for
inadequate. Are we doomed to keep on milk production. The model was of the last years milk production, which is clearly
throwing out our current tools and practices Y=f(Xi) form [Y = Y0 + a1X1 + a2X2 +...+ incorrect. The model must be abandoned.
as soon as the next wave of innovations anXn], where the Xis included such things Unfair, you might say. Its not that
comes along? he asked. as: last years milk production, interest rates, the model is wrong. Its just that the world
The answer is . . .it depends. It spending on cattle feed, GNP growth, and has changed dramatically since the model
depends on the underlying theory on which other macroeconomic factors. As the article was originally built and the changes must
the current tools and methods are based. If detailed, the model performed quite well as now be added. But what has really
ones management practices are based on a predictive deviceat least in terms of its changed? Yes, the cows are now dead, but
transient or situation-specific phenomena, ability to track history. The obvious thing the basic fact that milk comes from cows,
they are likely to require revision whenever about this model, that would bother both and that without cows there can be no milk,
the circumstances change. If, on the other dairy farmers and people who were partial is as true now as it was before the mass dec-
hand, they are based on a structural under- to operational specifications, is: wheres the imation. From a structural perspective, the
standing, the situation may change, but the cows?! Simply stated, if youve got no nature of the world has not changed at all.
tools will still apply. cows, youve got no milk! Crude, but true. The model was inadequate because it was
How does all this talk about cows relate based on situation-specific data that has
WHERE ARE THE COWS?
to our vice presidents question? Well, now changed.
Barry Richmond of High Performance imagine that an epidemic swept over the
STRUCTURAL THINKING
A C C U M U L AT O R S : B AT H T U B S ,
B AT H T U B S E V E RY W H E R E
hens the last time you actually took SHOWERHEAD VS. can lead to disastrous results.
W a real, honest-to-goodness bath? If B AT H T U B T H I N K I N G When Just-in-Time (JIT) manufactur-
you are like most people, it has probably Taking showers disconnects us from experi- ing first hit the U.S., for example, many
been quite a while. We live in the world of encing one of natures most basic struc- companies implemented it using a shower-
quick showers and instant breakfasts. Yet, turesaccumulators. Lakes and ponds are head perspective. The basic concept of JIT
it wasnt too long ago when taking baths accumulators of various water flows. Global is to manage a steady flow of materials
was part of our normal daily routine. The warming has been attributed to the cumula- through a factory with minimal accumula-
shift from baths to showers marked a far tive effects of burning fossil fuels. Plants are tions of inventory at each step. Many com-
more deeper change in our thinking than accumulators of energy and nutrition. panies that instituted JIT tried to
merely a change in personal hygiene Displacement, velocity, and acceleration can minimize their own accumulations by
habits. be represented in terms of accumulators. demanding that their suppliers provide
When we run the bathwater, we can That is, displacement represents the accu- them with materials just when they
visually see the water accumulating in the mulation of past velocity, and velocity is an needed them and not any sooner.
tub (see Bathtubs and Accumulators). We accumulation of past acceleration. The problem with the above approach,
know we have to keep an eye on the water If we use showerhead thinking, we are of course, is that the flow of materials has
level so it wont overflow. When we take less conscious of accumulations. Flows of to accumulate somewhere, and it was accu-
showers, however, the accumulation process materials such as water, fuel, or energy mulating in the suppliers warehouses. The
is virtually eliminated. Water flows out of simply go away somewhere. But from a JIT flow was accomplished by shifting the
the showerhead, over our bodies, and out bathtubor systemsperspective, there is accumulations to suppliers, severely strain-
the drain. Where does the water go? We no away. Everything accumulates some- ing the relationship between suppliers and
hardly give it any thought. where. Forgetting about that somewhere manufacturers. Bathtub thinking would
have highlighted the fact that unless the
B AT H T U B S AND A C C U M U L AT O R S entire flow from raw materials to final
customer worked together, there would be
undesirable accumulations for somebody
Faucet in the system.
I N V I S I B L E B AT H T U B S
Time
So how can you locate the invisible bath- flows? Both tools have their unique
R
tubs lurking in your company? For every strengths. Tools like systems archetypes o
Time
flow (action, decision, policy), try to figure capture and communicate dynamic issues Productivity
out what, if anything, is accumulating in a concise way, but they do not provide a Stressful Events
and what are the implications of those detailed representation of the structure s
Work Pressure
accumulations. producing the dynamics. s o Capacity
Workload s
to Handle
For example, as workload outstrips There are cases when tracing through Workload
capacity and work pressures become exces- a loop diagram can be confusing. For
Increasing work pressure can lead to an
increased number of stressful events, which
sively high (see Stress Accumulator), you example: Savings and interest form a
adds to the accumulation of stress.
should question whether those pressures reinforcing loop where higher savings bal-
simply come and go or whether their ance leads to higher interest payments,
effects are accumulating somehow. For
example, extra pressure may generate
FROM LOOP DIAGRAMS TO
more stressful events, which will accumu- A C C U M U L AT O R S A N D F L O W S
late into increasing levels of stress. High
stress levels will then lead to lower produc- Savings Balance
tivity, which further reduces work capacity
and leads to more stressful events. This s Savings
reinforcing loop of accumulating stress is Time
A C C U M U L AT I O N M A N A G E M E N T:
AV O I D I N G T H E PA C K R AT
SYNDROME
here is a story about a trivia pack accumulator management. A N AT O M Y O F
T rat, a man who had spent his A C C U M U L AT O R
PA C K R AT S A N D N O M A D S MANAGEMENT
entire life memorizing trivia. He knew
baseball statistics of every player in the his- Life can in some ways be viewed as a A typical accumulator management struc-
tory of the major league. He had memo- never-ending task of managing various ture (AMS) has the following elements: the
rized the titles, directors, and actors of accumulators. Our pantries, refrigerators, Accumulation, Acquisitions, Depletions,
hundreds of movies. He knew the name of checking accounts, and closets are among Desired Accumulation, and a Corrective
every television show that had ever aired. the many accumulations we manage daily. Action (see Accumulator Management).
But one day he found himself in an On one end of the accumulation man- In addition, there is almost always some
awkward predicamentno matter how agement spectrum is the pack rat who delay between the Corrective Action and the
hard he tried, he could not memorize any throws nothing away. On the other end is Acquisition, because it takes time to actually
more trivia. He had finally taxed the limits the nomad who makes a virtue of owning memorize data or clear out the closet once
of his rote memorization capacity. Although no more than what can be packed into one we have decided to do so.
he had worked hard at acquiring his stock suitcase. In between these two extremes lies The accumulator management struc-
of trivia throughout his life, he had never the majority of the population who is con- ture is a generic structure that can repre-
considered how he might go about depleting stantly struggling to maintain the right bal- sent a wide range of business settings
it. He had not learned the fundamentals of ance between acquisitions and depletions. where accumulation management is
A C C U M U L AT O R INSURANCE BUSINESS AS
MANAGEMENT A C C U M U L AT I O N M A N A G E M E N T
Desired Retention
Accumulation Accumulation
s o
s
B Gap
ay
Del
Corrective s
Action New Policies Loss Events Claims Adjusted Policies Lost
Acquisitions Depletions
Policyholders Claims Pending Claims Settled
s Accumulation
Del
ay B o
Desired
Gap s Accumulation Policies Lost
Corrective
Action Investments
s
Cash In Cash Out
Accumulator management can be viewed sim-
ply as a balancing loop with delay (top). A The insurance business can be mapped into a relatively simple diagram that highlights the major
structural diagram (bottom) reveals that the accumulators and flows. If we assign numbers next to each accumulator or flow indicating the
flows controlling the accumulation are acquisi- percentage of organizational resources devoted to it, the diagram can help reevaluate the organi-
tions and depletions. zations current emphasis.
investments. together in a structure similar to that repre- If you want to avoid the pack rat syn-
If managers assign a number next to sented in the AMS diagram (see Supply drome, you need to manage the whole
each accumulator and flow in the diagram Line and Delay in the Beer Game). Within accumulator management structure and
to represent the percentage of organiza- that team, each participant must make not just focus on one piece of it. The obser-
tional resources devoted to each, the dia- ordering decisions in order to maintain his vations about the difficulties of managing
gram can highlight which areas receive the desired level of inventory. the Beer Game suggests that you should
largest organizational focus. This exercise According to MIT professor John think through the following questions
can point out any weaknesses in the current Sterman, when participants try to manage when confronting a typical accumulator
organizational emphasisfor example, accumulations in the Beer Game they usu- management situation: (1) Where are the
spending too little time trying to retain ally run into three common problems. First, supply line delays and how are they chang-
current policyholdersand reveal ways in they typically underestimate the true length ing? (2) What factors are determining
which the company can better serve its of the delay from the time they order to what Desired Accumulation should be? (3)
customers. when they receive the beer and then overad- How do current policies and decisions feed
just their orderseven when they are given back into this system to produce the results
S U P P LY L I N E S A N D full information about the supply line delays. we have observed? The accumulation
D E L AY S They do not appear to recognize that their management structure diagram is a useful
If we had direct and immediate control ordering decisions affect the length of the starting point to begin addressing these
over all the elements in the AMS diagram, supply line delaythat is, the more they questions.
managing accumulations would be simple: order, the longer it takes to receive the beer.
We would calculate the depletion rate, set In addition, he found that when people Further Reading: Modeling Managerial Behavior:
Misperceptions of Feedback in a Dynamic Decision
our desired accumulations accordingly, find it difficult to determine their optimal Making Experiment, by John D. Sterman,
and implement actions that will immedi- inventory level, they simply anchor their Management Science, Vol. 35, No. 3, March 1989.
ately result in acquisitions. In our home desired inventory on the initial inventory
life we already pretty much follow this and adjust from there.
pattern. For example, we plan our meals, This finding high- S U P P LY L I N E A N D D E L AY
decide on an appropriate amount of food lights the more gen- IN THE BEER GAME
to have on hand, figure out how long it eral tendency people
will be before we run out of certain staples, have to anchor on s Delivery
and go to the grocery store as needed. Delay
past goals or stan-
R
Unfortunately, things are not that straight- dards rather than Orders Shipments
forward when we move into the organiza- search for better ones.
o
tional context. The third obser-
Supply Line Inventory
One of the most challenging aspects of vation is that people
s B Deliveries
managing accumulations within organiza- generally point to fac- o
Corrective Action Gap Desired
tions is captured in one worddelays. tors outside the system s Inventory
Identifying and characterizing the nature as being responsible
The structure of the inventory management system in the Beer Game is
similar to the AMS diagram. Understanding the nature and source of
and source of delays often plays a critical role for the instabilities
delays in a systemssuch as the supply line delay aboveoften plays
in managing accumulations effectively. A big they observe in the
part of the problem is that we usually have game. That is, people a critical role in managing accumulations without overcorrecting.
very little control over the supply line delay. offer open loop expla-
D E L AY S : A C C U M U L AT O R S
IN DISGUISE
THERE! s Work in
Shipments
Process
The Production Chain diagram Full R2
Utilization
does not show the complete picture o
Down Finished
of what goes on in a typical manu- s Time Goods
Efforts to expedite orders can backfire by actually lengthening the average production delay and reinforcing
facturing setting (see Expediting
the need to expedite (loops R3 and R4).
Loops). When shipments fall short
of customer orders, customers are
S - S H A P E D G R O W T H A N D T H E L AW
OF DIMINISHING RETURNS
ost of us are familiar with the story discovery hits himthe Law of Success archetype. At the heart of both
M of Sir Isaac Newton sitting under- Diminishing Returns. structures is a pair of reinforcing and bal-
neath an apple tree and discovering the ancing loops that interact to produce the S-
THE LAW OF
law of gravity when he saw an apple fall shaped pattern.
DIMINISHING RETURNS
(some say on his head). Had Newton been
The phenomena of diminishing returns DIFFUSION DYNAMICS
an entrepreneur, he might have discovered
another law. With apple salesand prof- when more effort yields fewer resultsis The basic Bass Diffusion Model is usually
itsin mind, he might have shaken the ubiquitous. Oil recovery and mining opera- given as a set of equations:
tree vigorously, causing more apples to tions exhibit this behavior. Companies expe-
f(t) = dF(t)/dt = [p+qF(t)] [1 F(t)]
drop to the ground. The harder he shook, rience rapid new product sales followed by
p, q 0
the more apples would fall, meaning more decreasing demand. At a personal level, we
p = coefficient of advertising,
sales and more profits. After a while, how- see that working longer hours, jogging
q = coefficient of interaction.
ever, each shaking would produce fewer more miles, and eating less food lead to
and fewer apples. diminishing returns in productivity gains,
Integration yields an S-shaped
We can almost picture the scene: Sir health benefits, and weight loss.
growth curve of diffusion.
Isaac wipes the sweat from his brow, then The Law of Diminishing Returns can
climbs the tree to knock down the tena- be considered the law of gravity for the Although the equations may offer an
cious few apples left. Precariously perched, business world. Launching a marketing elegant way to represent such dynamics,
he strains to reach one of the last remain- campaign, for example, is like the trajectory most of us dont view the world as a set of
ing apples. The limb gives way, and he of a cannonballthe returns climb higher equations. From an accumulator and flows
falls. As he lands on the ground, another and higher, until the force of diminishing perspective, we see diffusion dynamics as a
returns kicks in and flow of people from a pool of potential
pulls the rate of return adopters to adopters (see Bass Diffusion
B A S S D I F F U S I O N M O D E L
A S T R U C T U R A L V I E W P O I N T down. When traced out ModelA Structural Viewpoint).
over time, the cumula- Instead of ps and qs, we talk about an
tive returns of the mar- advertising effect and a word-of-mouth
keting effort produce an effect. This structural view makes the
S-shaped curve. dynamics more explicit, closer to the way we
Time
B R Time
The Law of actually think about and experience the
Potential Adopters Diminishing Returns is world.
Adopters essentially about satura- Faddish products, such as hula hoops
Time
tion effectsreaching and Cabbage Patch dolls, usually exhibit
the limits of a particular classic S-shaped growth. Remember the
Adoption Rate system. The characteris- sudden popularity of pet rocks in the late
Advertising Interaction tic S-shaped behavior of 1970s? They were just plain old rocks
COMPUTER-BASED TOOLS
COMPUTER MODEL
M A N A G E M E N T F L I G H T S I M U L AT O R
L E A R N I N G L A B O R AT O R Y
any of the systems we are charged with managing are so dynamically complex
M that they are almost incomprehensible. Complex social systems frequently exhibit
counterintuitive behavior, where actions that provide short-term relief often result in
greater long-term pain. When actions and consequences are greatly separated in space
and time, making effective decisions for the long-term well-being of a system becomes
extremely difficult.
Causal loop diagrams, archetypes, and structure-behavior pairs can help us build a
better conceptual understanding of the key relevant structures of a system and perhaps
even predict the general behavior of the system over time. Modeling for What
Purpose (p. 40), however, describes times when we need even greater precision about
the ramifications of certain actions at specific points in time.
The rest of this section describes how we can translate our pen-and-paper represen-
tations into computer-based models that can be simulated, converted into interactive
decision-making games (management flight simulators), and embedded in a rich learn-
ing environment (learning laboratories).
M O D E L I N G F O R W H AT P U R P O S E ?
B Y J A Y W . F O R R E S T E R
ystem dynamics does not impose behavior. In general, influential system ally can agree) merely because they initially
S models on people for the first time dynamics projects are those that change disagree with the dynamic conclusions that
models are already present in everything we the way people think about a system. Mere might follow.
do. One does not have a family or corpora- confirmation that current beliefs and poli- If we divide knowledge of systems into
tion or city or country in ones head. Instead, cies are correct may be satisfying but three categories, we can illustrate wherein
one has observations and assumptions about hardly necessary, unless there are differ- lie the strengths and weaknesses of mental
those systems. Such observations and ences of opinion to be resolved. Changing models and simulation models (see Three
assumptions constitute mental models, and unifying viewpoints means that the Categories of Information). The top of
which we then use as a basis for action. relevant mental models are being altered. the figure represents knowledge about
The ultimate success of a system structure and policies; that is, about the
UNIFYING KNOWLEDGE
dynamics investigation depends on a clear elementary parts of a system. This is local
initial identification of an important pur- Complex systems defy intuitive solutions. non-dynamic knowledge. It describes
pose and objective. Presumably a system Even a third-order, linear differential information available at each decision-
dynamics model will organize, clarify, and equation is unsolvable by inspection. Yet, making point. It identifies who controls
unify knowledge. The model should give important situations in management, eco- each part of a system. It reveals how pres-
people a more effective understanding nomics, medicine, and social behavior usu- sures and crises influence decisions. In
about an important system that has previ- ally lose reality if simplified to less than general, information about structure and
ously exhibited puzzling or controversial fifth-order nonlinear dynamic systems. policies is far more reliable, and is more
Attempts to deal with nonlinear often seen in the same way by different
dynamic systems using ordinary processes of people, than is generally assumed. It is only
T H R E E C AT E G O R I E S
O F I N F O R M AT I O N description and debate lead to internal necessary to dig out the information by
inconsistencies. Underlying assumptions using system dynamics insights about how
may have been left unclear and contradic- to organize structural information to
Observed
tory, and mental models are often logically address a particular set of dynamic issues.
Structure incomplete. Resulting behavior is likely to The middle of the figure represents
and Policies
a be contrary to that implied by the assump- assumptions about how the system will
a
Expectations
tions being made about underlying system behave, based on the observed structure and
About Behavior structure and governing policies. policies in the top section. These beliefs are,
b System dynamics modeling can be in effect, the assumed intuitive solutions to
b
Actual effective because it builds on the reliable the dynamic equations described by the
Behavior part of our understanding of systems while structure and policies in the top section of
compensating for the unreliable part. The the diagram. They represent the solutions,
There are three categories of information
system dynamics procedure untangles sev- arrived at by introspection and debate and
about a system: knowledge about structure
eral threads that cause confusion in ordi- compromise, to the high-order nonlinear
and policies; assumptions about how the sys- nary debate: underlying assumptions
tem will behave based on the observed struc-
system described in the top part of the fig-
ture and policies; and the actual system
(structure, policies, and parameters), and ure. In the middle lie the presumptions that
behavior as it is observed in real life. The
implied behavior. By considering assump- lead managers to change policies or lead
usual discrepancy is across the boundary a-a: tions independently from resulting behav-
expected behavior is not consistent with the
governments to change laws. Based on
known structure and policies in the system.
ior, there is less inclination for people to assumptions about how behavior is
differ on assumptions (on which they actu- expected to change, policies and laws in the
MANAGEMENT FLIGHT
S I M U L AT O R S : F L I G H T T R A I N I N G
FOR MANAGERS (PART I)
magine youre leaving on a six-hour ing, in the traditional business-school operations and focusing on the long-term
I flight from Boston to Los Angeles. sense, is the equivalent of ground-school dynamics of managerial decisions.
As the plane pulls away from the gate, the for pilots. Managers-to-be read textbooks
C R E AT I N G A F L I G H T
pilot comes on over the loudspeaker: Hi, and solve already formulated problems,
S I M U L AT O R
Im Captain Bob, and I want to thank you but they dont get much real experience
for choosing to fly with us today. . . . Just before they have to perform on-line. There are four stages involved in creating
wanted to let you know Ive recently com- a management flight simulator: (1) select-
MANAGEMENT FLIGHT ing an issue focus, (2) developing a concep-
pleted ground school training, and I have
S I M U L AT O R S tual model, (3) constructing a computer
read all the manuals, but this is my first
time in the cockpit. So sit back, relax, and Management Flight Simulators (MFSs) pro- model, and (4) translating the computer
enjoy the flight, as we learn together. . . . vide a simulated environment in which model into an interactive simulator (see
Of course this scenario is ludicrousa managers can learn from experience in a Management Flight Simulator
pilot is allowed into a cockpit only after controlled setting. The simulator captures Development Stages). These four stages
hundreds of hours of experience in a flight the interconnections between the different involve integrating many of the tools of
simulator. Then he or she spends many parts of the system under study and pro- systems thinking into a single, powerful
additional hours as a co-pilot, assisting in the vides a computer interface that allows learning tool (see A Palette of Systems
operation of an aircraft. The result of this managers to interact with the model Thinking Tools, p. 10, for a description of
careful system of education and training is through a familiar lens (reports, graphs, each of the tools).
an industry with the highest safety record of and spreadsheets). 1. Select issue focus. The first step in
any mode of transportation. Similar to a pilots flight simulator cock- designing a flight simulator is to choose an
pit, an MFS puts managers in control of a issue to explore. To select a topic, look for
FLIGHT TRAINING FOR realistic environment where they are in a problem symptom that has been around
MANAGERS charge of making key decisions similar to for a long time or a puzzling dynamic you
Imagine if we trained pilots like we do the ones they face in their actual work set- want to investigate (see The Dos and
managers; how many people would be tings. MFSs are particularly useful for get- Donts of Systems Thinking on the Job,
willing to take a flight? Managerial train- ting away from the details of day-to-day p. 52, for guidelines on identifying good
M A N A G E M E N T F L I G H T S I M U L AT O R D E V E L O P M E N T S TA G E S
Revise
MANAGEMENT FLIGHT
S I M U L AT O R S : F L I G H T T R A I N I N G
FOR MANAGERS (PART II)
management flight simulator, MANAGING VS. LEARNING A clear, real-world context provides a
A along with causal loop diagrams There are two fundamentally different real operational focus that engages line
and systems archetypes, allows you to see uses for a computer model and simula- managers in learning more about their own
more clearly the connections between your tormanaging and learning. Simulators issues.
decisions and future consequences. As sim- and models designed to support decision- Face validity: Make the MFS real
ulated months pass in a matter of minutes making in a real operational setting must enough so the simulator grounds people in
and the consequences of your actions focus on capturing the operational reality their own real-life experiences.
unfold, an MFS provides a means for precisely because operational or strategic A strong conceptual framework helps
making sense of the short-term and long- decisions will be based on those numbers. make systemic sense out of the complex
term effects of your decisions. Simulators that are designed for learn- dynamics (e.g., systems archetypes).
Management flight simulators can be ing, on the other hand, are much more Conventional and unconventional
most useful for understanding situations in concerned with surfacing the tacit mental information systems provide a familiar
which causality is distant in time and models that drive managers decision- information environment, as well as an
space. When the inherent time lag is par- making. Accuracy of specific numbers is opportunity to explore and experiment with
ticularly long (on the order of months or not as important as the relevancy of the new ones.
years) and organizational complexity is issues and concepts captured in the simula- Surface and challenge mental models
high (see Organizational Complexity), tor; in other words, simulators for learning to break through individual mental straight-
learning from experience can be fraught are idea-rich versus data-rich. jackets and corporate sacred cows and
with pitfalls. An MFS allows you to lever- There are several different simulator advance team learning.
age your ability to learn from experience in design criteria to keep in mind when
a complex environment. designing an MFS: DESIGNING MFSS AS
TRANSITIONAL OBJECTS
L E A R N I N G L A B O R AT O R I E S :
PRACTICING BETWEEN
PERFORMANCES
magine you are walking across a encounter in the actual setting. It does not mean you agree or disagree with
I tightrope stretched between the A managerial practice field should also that persons view; you simply acknowledge
world trade towers in New York City. The have its own sets of equipment and tools for the right of that person to hold that view. A
wind is blowing and the rope is shaking as making the practice sessions meaningful. second rule is to suspend ones own assump-
you inch your way forward. One of your The purpose of a learning laboratory is to tions and the other persons and hold them
teammates is sitting in the wheelbarrow you provide an environment in which managers equally in our minds, without judging ours
are balancing in front of you, while another can experiment with alternative policies, test to be superior or right. Creating such a
perches on your shoulders. There are no assumptions, and practice working through learning space also means engaging in dia-
safety nets, no harnesses. You are thinking to complex issues productively. It should allow logue rather than discussionoperating in a
yourself, One false move and the three of us managers to practice working together as a spirit of inquiry rather than advocacy.
will be taking an express elevator straight team on issues of real significance to them.
down to the street. Suddenly your trainer M A P P I N G M E N TA L
To be effective, the learning lab must pro-
MODELS
yells from the other side, Try a new move! vide (1) an environment conducive to learn-
Experiment! Take some risks! Remember, ing, (2) a way of surfacing deep-rooted Along with the proper environment, we
you are a learning team! assumptions that affect the way we think need tools for helping people surface and
Sound ludicrous? No one would be and act, (3) tools for understanding our real- share their assumptions. For example, the
crazy enough to try something new in a sit- ity in a way that highlights the interconnec- Ladder of Inference, developed by Chris
uation like that. And yet that is precisely tions and the systemic consequences of our Argyris, distinguishes between directly
what many companies expect management actions, and (4) a management flight simula- observable data, shared cultural meanings,
teams to doexperiment and learn in an tor that allows us to speed up or slow down judgments, conclusions, and values and
environment that is risky, turbulent, and time, experiment with different strategies, assumptions. Argyris uses the ladder to
unpredictable. Unlike a high-wire act or and see the long-term consequences of our illustrate the leaps of inference that occur
sports team, however, management teams actions (see A Sample Learning Laboratory when people take a little bit of observed data
do not have a practice field in which to Design). (a person walks into a 2:00 meeting at 2:15)
learn; they are always on the performance field. and go straight up the ladder to the level of
C R E AT I N G A S A F E values and assumptions (Hes late and
DESIGNING MEANINGFUL L E A R N I N G S PA C E doesnt care about the project or the other
PRACTICE FIELDS Learning usually involves making mistakes players) without even being conscious of it.
A learning lab can be viewed as a managers because we are trying things we have never The ladder provides a useful framework for
equivalent of a sports teams practice field. done before. It requires us to approach helping people walk back down the lad-
The goal of a learning lab design is to pro- things from a place of not knowing. It der to understand what is really happening
vide a real enough practice field so that involves risk. How, then, can we create a and begin managing by facts, not opinions.
the lessons are meaningful, but safe enough safe space where people feel free to learn? Systems archetypes also provide a pow-
to encourage experimentation and learning. There are some ground rules that can erful set of tools for mapping out a persons
In the tightrope example, a practice field help create such safe spaces. One ground understanding of a problem or issue in a
could be a rope stretched across two pillars rule is to hold each persons viewpoint as form that invites others to inquire and clar-
six feet off the ground. There may be mats valid. That requires taking the position that ify the picture together. Having ones
below to cushion the fall, but also a large fan If I could stand in the other persons shoes, assumptions captured in terms of archetypes
to simulate the kinds of winds you would I too could see what the other person sees. and causal loop diagrams helps depersonal-
A S A M P L E L E A R N I N G L A B O R AT O R Y D E S I G N
should help managers step out of day-to-day demands to: are introduced in a storytelling format in which partici-
reflect on their decision-making pants begin to tell systemic stories about their issues.
develop a common language Facilitators then describe a small portion of the CLDs that
learn new tools for thinking systemically were in the game model to connect the tool to the issues
discuss operational objectives and strategies in an at hand. The underlying purpose is to get people to imme-
open forum diately begin to connect each structure to corresponding
test operating assumptions patterns of behavior over time.
experiment with new policies and strategies for
managing
In small groups, the participants are asked to focus on a
4. Using the ToolsConceptualizing
have fun.
particular issue, such as one of the decision variables in
the management flight simulator, and (1) determine the
Explaining the context of the LL to participants (the history key factors that affect that variable, (2) sketch patterns of
1. The First Crucial HourBuy-In
of its development, the original intent or purpose) is critical behavior, (3) provide structural explanation (using CLDs),
for establishing a common understanding: The LL is not and (4) identify intervention points. By having the group
meant to provide the answers, but to serve as a useful explore these variables, the participants can replicate part
vehicle for illuminating and communicating issues of of the model-building process and accept the predevel-
importance. The facilitators are positioned as enablers, oped model. The overall objective in this section is to
not authority figures, and the participants are encouraged have the group cover all the major issues contained in the
to question the assumptions behind the LL design. model and have a chance to challenge and test the inter-
relations that different people within the group may pro-
pose.
This exercise helps the participants construct a group pic-
2. Current RealityWhere Are We?
storm and come up with a list of operational objectives, contains all the major variables in the model. They trace
strategies required to achieve them, and obstacles that through the major loops and explain the dynamic conse-
need to be overcome in order to reach their goals (e.g., quences of a particular action or incident, and then sketch
reduce settlement expenses by 10 percent). The idea is to a corresponding pattern of behavior and connect it back to
get everyone thinking in terms of their own operational the structure. This is followed by a hands-on introduction
reality. to the computer and simulator.
A S A M P L E L E A R N I N G L A B O R AT O R Y D E S I G N
In this section, it is best if people work in groups of two at This time, the participants are free to choose their own
6. Planned ScenariosHolding the Reins 7. Free PlaysCutting the Reins
each computer. The teams are instructed to pursue a objectives and time tables. Again, each team strategizes,
single-minded strategy where they are accountable for and then explains to the rest of the group how they plan to
meeting one particular goal (e.g., hiring freeze). Each two- achieve their goals. For the designers of the LL, this sec-
person team is responsible for doing the following: tion provides the opportunity to challenge deep-rooted
(1) plan a strategy and commit to it on paper, (2) predict norms and assumptions, address specific hot topics, or
the consequences of executing the strategy by sketching re-create various historical behavior modes for further
in behavior over time of some key variables, (3) play the exploration.
game, and (4) debrief game results and explain to the rest
of the group. This stage allows participants to begin to
address particular organizational issues within a carefully
controlled setting.
REFERENCE GUIDE
o, youve taken a systems thinking GENERAL GUIDELINES Cues that non-systemic thinking is going
S courseor maybe youve read a few DONT use systems thinking to further on: Phrases such as We need to have imme-
issues of The Systems Thinkerand now you your own agenda. Systems thinking is most diate results, We just have to do more of
want to start using systems thinking on the effective when it is used to look at a problem what we did last time, or Its just a matter
job. How do you begin? Your best bet is to in a new way, not to advocate a predeter- of trying harder.
approach this endeavor in the spirit of mined solution. Strong advocacy will create
learning to walk before you run. Here are resistanceboth to your ideas, and to sys- G E T T I N G S TA R T E D
some suggestions: tems thinking. It should be used in the spirit DONT attempt to solve a problem imme-
of inquiry, not inquisition. diately. Dont expect to represent, much less
OVERALL GUIDELINE
DO use systems thinking to sift out understand, persistent and complex systemic
The tools of systems thinking are best used major issues and factors. problems overnight. The time and concen-
as vehicles to promote team learning in the Benefit: Systems thinking can help you tration required should be proportional to
organization. Whether you are doing paper break through the clutter of everyday the difficulty and scope of the issues
and pencil models or creating full-fledged events to recognize general patterns of involved.
microworlds, the process of constructing behavior and the structures that are produc- More realistic goal: to achieve a fuller and
and using models is primarily about explor- ing them. It also helps in separating solu- wider understanding of the problem.
ing and examining our mental models tions from underlying problems. Too often DO start with smaller-scale problems.
the deeply held assumptions that influence we identify problems in terms of their solu- DONT attempt to diagram the whole
the way we think and act. tion; for example, the problem is that we systemotherwise youll quickly become
have too many ________ overwhelmed.
(fill in the blank: people, Better: Try to focus on a problem issue
I D E N T I F Y I N G A S Y S T E M S initiatives, steps in our pro- and draw the minimum variables and loops
P R O B L E M cess), or the problem is youll need to capture the problem.
that we have too little DONT work with systems thinking
The problem should have ALL of the following
__________ (resources, techniques on line under pressure, or in
1. The issue is important to me and my busi-
characteristics:
ness.
information, budget . . .). front of a group that is unprepared for or
time event.
thinking to blame individu- Additional danger: If the audience is not
diagram:
general structures that might clarify the Better: Reduce the
Add a link.
problem. information delay between
Break a link.
Advantage: Systems archetypes provide a sales and manufacturing by
Shorten a delay.
focal point or a storyline to begin the pro- creating a new information
Make a goal explicit.
cess of understanding a problem. system.
Slow down a growth process; relieve a
DO work with one or more partners. DO make an interven-
Advantage: Multiple viewpoints add tion specific, measurable, limiting process.
3. The best intervention is likely to be a combi-
nation of interventions applied gently and
richness and detail to the understanding of a and verifiable.
patiently.
problem. Example: Cut the infor-
PART I: AN OVERVIEW
Systems Thinking Tools: A Users Reference Guide is a volume in the Toolbox Reprint Series. Other volumes
include Systems Archetypes I: Diagnosing Systemic Issues and Designing High-Leverage Interventions, Systems
Archetypes II: Using Systems Archetypes to Take Effective Action, Systems Archetypes III: Understanding Patterns
of Behavior and Delay, and The Thinking in Systems Thinking: Seven Essential Skills. All volumes are avail-
able for $16.95 each. As these booklets are often used in training and introductory courses, volume
discounts are available. See the copyright page or call 1-800-272-0945 for details.
The Toolbox Reprint Series has been compiled from The Systems Thinker Newsletter, which presents a
systems perspective on current issues and provides systems tools for framing problems in new and insightful
ways. The Systems Thinker includes articles by leading systems thinkers, case studies of systems thinking
implementation, software and book reviews, and numerous other columns geared to different levels of
systems thinking ability. Individual subscription rates to The Systems Thinker are as follows: A special rate of
20 issues for $169, or 10 issues (one year) for $109. Library subscriptions are available for $189 for 10 issues
(one year). Visit www.pegasuscom.com for more information.
Anthologies
Managing the Rapids: Stories from the Forefront of the Learning Organization
Reflections on Creating Learning Organizations
The New Workplace: Transforming the Character and Culture of Our Organizations
Organizational Learning at Work: Embracing the Challenges of the New Workplace
Making It Happen: Stories from Inside the New Workplace
The Pegasus Workbook Series
Systems Archetype Basics: From Story to Structure Systems Thinking Basics: From Concepts to Causal Loops
The Learners Path: Practices for Recovering Knowers
The Billibonk Series
Billibonk & the Thorn Patch Frankls Thorn Patch Fieldbook
Billibonk & the Big Itch Frankls Big Itch Fieldbook
Learning Fables
Outlearning the Wolves: Surviving and Thriving in a Learning Organization
Shadows of the Neanderthal: Illuminating the Beliefs That Limit Our Organizations
The Lemming Dilemma: Living with Purpose, Leading with Vision
The Tip of the Iceberg: Managing the Hidden Forces That Can Make or Break Your Organization
Listening to the Volcano: Conversations That Open Our Minds to New Possibilities
Other Titles
Human Dynamics: A New Framework for Understanding People and Realizing the Potential in Our Organizations
When a Butterfly Sneezes: A Guide for Helping Kids Explore Interconnections in Our World Through Favorite Stories
The Innovations in Management Series
From Mechanistic to Social Systemic Thinking: A Digest of a Talk by Russell L. Ackoff
Applying Systems Archetypes
Toward Learning Organizations: Integrating Total Quality Control and Systems Thinking
Designing a Systems Thinking Intervention: A Strategy for Leveraging Change
The Natural Step: A Framework for Achieving Sustainability in Our Organizations
Anxiety in the Workplace: Using Systems Thinking to Deepen Understanding
The Soul of Corporate Leadership: Guidelines for Values-Centered Governance
Creating Sustainable Organizations: Meeting the Economic, Ecological, and Social Challenges of the 21st Century
Creating Value: Linking the Interests of Customers, Employees, and Investors
Relinking Life and Work: Toward a Better Future
Facing the Competition: An Organization Mobilizes for Large-Scale Change
Organizational Change at Philips Display Components: Reflections on a Learning Journey
Introduction to Systems Thinking
Rebounding, Rebuilding, Renewing at Shell Oil: A Former CEO Reflects on Large-Scale Change
Reinventing Human Resources at L.L. Bean: Lessons for Learning and Change
The Essentials of Servant-Leadership: Principles in Practice
Dialogue at Work: Skills for Leveraging Collective Intelligence