Documenti di Didattica
Documenti di Professioni
Documenti di Cultura
Ackoff's
Fables ~ ·
Russell L.
Ackoff
THE ART OF
PROBLEM SOLVING
ACCOMPAN IED BY ACKOFF'S FABLES
RUSSELL L. ACKOFF
Illustrations by
KAREN 8. ACKOFF
Bibl iography: p.
Includes index.
1. Problem solving. I. Title.
R. D. Lain g, Th e Politics
of the Family and Other Es sa ys
Vintage Books, New York, 1972
Preface
RUSSELL L. ACKOFF
Phi ladelphia , Pennsylvania
January 1978
Contents
l
r
l
CHAPTER ONE
Competence
Commun i ca tiven ess
Concern
Courage
Creativity
Figure 1.1.
That creativity can be acquired seems to follow from the (act that it
tends to get lost in the process of growing up. Adults recognize that
young children, parti cula rly preschoolers, are full of it I recall a dra-
matic illustration of this point given by an eminent student of creativ-
ity, Edward de Bono, in a lecture to an audience consisting of managers
and management scientists. He drew a picture on the blackboard of a
wheelbarrow with an elliptical wheel (Figure 1 .1) and asked the audi-
ence why it had been designed that way. There was a good deal of
squirming, murmuring, and embarrassed tittering, but no answer. De
Bono wa ited, letting the discomfort grow. He then told his audience
that he had recently asked the same quest ion of a group of children
and almost immediately one of them had rushed to the board and
drawn a squiggly line such as that shown in Figure 1.2. "The wheelbar-
row is for a bumpy road," the child had said. The audience blushed
and laughed self-consciously.
Most of us take for granted both the c reativity of children and its
subsequent loss. We do not try to understand, let alone prevent, this
loss. Yet the disappearance of creativity is not a mystery; the explana-
tion lies in a query that Jules Henry (1963), an Amer ican anthropologist,
once made : What would happen, he asked,
Figure 1.2.
• • •
• • •
• • •
Figure 1.3.
6 The Art
Figure 1.4.
The instructions below the figure said that a pen or pencil was to be
placed on one of the dots and then four straight lines were to be drawn
without lifting the pen or pencil from the paper so that all nine dots
were covered by the lines.
My daughter had tried to solve the problem, with no success. She
asked me for help, assuring me she would not claim the solution as her
own. I recognized the problem, but I was unable to recall or find its
so lut ion. Impatient to get back to the work she had interrupted, I told
her to forget about the problem. "It's not that important," I said. She
left unconvinced and with an obviously lowered opinion of my prob-
lem-solving ability.
A short while later I heard her sobbing in the next room. I went
in to see what was wrong. She told me she was ashamed to go to
school without a solution to the problem. I invited her into my study
and said that this time I would make a ''real try." She came skeptically.
I knew, for reasons considered later, that a puzzle is a problem that
we usually cannot solve because we make an incorrect assumption that
preclu des a solution. Therefore, I looked for such an assumption . The
first one that occurred to me was that the pape r had to remain flat on
a surface while the lines were drawn. Once this assumption came to
mind and I put it aside, a solution came quickly. I folded the sheet "in"
across the middle line of dots and "out" across the bottom li ne so that
the bottom dots fell on top of the dots of the top Ii ne (see Figure 1.4).
Then, using a felt-tipped pen I drew a line through the top line of dots,
ho!ding the pen against the folded edge on which the bottom dots
were loca ted . Keeping my pen on the last dot, I unfolded the paper and
flattened it. There was a line through the top and bottom rows of dots
Creativity and Constraints 7
• • •
Figure 1.5.
(see Figure 1.5). With three lines left it was easy to cover the rema in-
ing dots (see Figure 1.6).
My daughter was delighted with the solution, and her fa ith in me
was partially restored. I returned to my work with more than a l ittle
self-satisfaction.
When I returned from work the next day I could hardly wait to hear
what had happened in my daughter's class. She returned my greeting
as I entered with her usual "Hi" but nothing more. I waited a few mo-
ments and then asked, "Well, what happened in your math class?"
"It doesn't matter," she replied, not looking at me.
"Yes it does," I countered. "Now come on, tell me."
"It will only get you mad," she said.
"Maybe it will, but if it does, I will not be mad at you . So tell me."
"Well," she said, "the teacher asked the class who had solved the
problem. About five of us raised our hands. She cal led on another girl
who had her hand up and asked her to go to the board and show her
solution. She did." My daughter then drew the solu tion shown in Fi g-
ure 1.7 on a sheet of paper. It was the solution I had once know n bu t
forgotten .
"Then what happened?" I asked.
"The teacher congratulated the girl, told her to return to her seat,
Figure 1.6.
4
2
Figure 1.7.
Creativity and Constraints 9
and started to talk abo ut something else. I raised my hand . She stopped
and asked me what I wanted. r told her I had a different solu tion to the
problem, one you had given me . She was annoyed but asked me to go
lo the board and show it to the class. I told her I cou Id n' t show it on
the blackboard and needed to use the la rge pad on the easel in the
corner of the room. She told me Lo go ah ead . I drew the nine dots on
a blank sheet and started to fold it when she asked what I was doing .
I told her I was folding the paper. She to ld me I couldn't do that. I told
her that the instruc tions didn't say r couldn't. Th en she told me she
didn't care what the instructions sa id; that was what she meant. She
told me to sit down, so I never got to fin ish showing the solution"
This is how creativity is suppressed, although usually not so overtly.
The teacher made it clear to her class tha t the objective of the assign-
ment was not to find a solution to the problem, but to find the solution
she knew and could pretend to have discovered on her own. She had
no interest in any other solution.
Is it any wonder that students become more concerned with wha t a
teacher expects of them in an examination than with what are the best
answers to the questions asked?
Imagine what a teacher interested in promoting creativity · could
have done with the situation invo lvi ng my daughter. She could have
revealed the common property of both so lutio ns : they broke an as-
sumption that t/Je solver imposed on the problem. In the teacher's
solution the broken assumption was tha t the Jines drawn had to lie
within the perimeter of the square formed by the dots. She could then
have gone on to encourage the students to find other solutions . Had
she done so, one of the students might have discovered how to fold
the paper so that one line drawn with a felt-tipped pen ca n cover all
the dots (see Figure 1 .8) .
A puzzle is a problem that one cannot solve because of a self-
imposed constrain/ . Creativity is shackled by self-imposed constraints .
Therefore, the key to freeing it l ies in developing an ability to identi fy
such constraints and deliberately removing them .
It is not enough to become aware o f the fact that self-imposed con-
straints are what obstruct creative problem solving. For example, now
that you are aware of this fact, consider thi s problem .
Figure 1.8.
For examp le, the purchaser of an automobile may p lao · a li mit on how
much he is wi ll ing to spend. He may also decide that he will not buy a
used car. H is cho ices may also be constra ined by what is ava ilable at
th e time of purchase.
For example, he may get either a good car o r a "' l emonlf ote that
there must be at least two possible outcomes. If th is were not the case,
the decis ion maker's choice would have no effect on the outco me;
therefo re, his ch oice would not be a "rea l" or "mean ingfu l" on e. Fur-
thermore, the two or more poss ib le outcomes must be unequa lly de-
sirab le; their values to him must be different, otherwise i t would not
matter ro h im which outcome occurred.
To o ne ind ividual there may be no significant difference be tween
the va lue s of two automobiles of th e same mak e and mode l but of clif-
f rent co lors. To another thi s difference may be of great significance,
in fact, critical.
A decision maker tries to se lect a course of action that produces an
outcome he desires, one that is efficient relative to wha t he values.
Such courses of action are said to be eHc l iv . Effectiven s i the
product of effi cie ncy and value. One who see ks th e best, the most ef-
fectiv , course of action is said to optimize. One who seeks a solution
that is good enough is said to sa tisfice.
In summary, choice 1:xists on ly (1) when there are at leas t two pas·
sib le courses of action availab le to the dec ision ma ke r, (2) where there
are at least two possib le outcomes of unequal value to him, and (3)
where the different courses of action have d i ffe rent effectivene s. In
other words, cho ice exists when the action of the decision make
makes a diffe rence in the v.:i lue of the outcome.
Not every choice situation is a problem si tuation , but every problem
invo lves a choice. A problem arises when the deci si on maker has som
doubt about the re lative effectiven ess of th e alternative courses of ac-
tion. The solution process is directed at dispelling doubt .
It is apparent that a choice situat io n tha t presents a problem to on
person may not do so to ano ther because of a difference in thei
Creativity and Constraints 13
doubts. This is what makes for consultants, experts, advisors, and so on .
In dealing with a problematic situation, a decision maker must de-
velop a concept-a representa tion or a model-of it. He attempts to
solve the problem as he conceives it. Thus if his conception is wrong ,
lhe solution to the problem as conceived may not solve the problem
asitexists. A common example is a formulation of a problem that leads
lo the suppression of symptoms rather than the removal of th e cause
~ iU£iency thaf creates t e roblem. Because ofsuc errors of con- 1
Since con straints apply to eac h of these, they need not be treated
separately.
As stated p reviously, the pri nciples presen ted are il lustrated by a
number of cases; I ca n vouch for most of them . A few are second hand
or more, but the poss i ble inaccu racies in my account of these cases in
no way detracts from their usefu lne s as examp les of the princip les . As
one of my friends said about these examples, " If they are not true, th ey
ought to be." The reader may treat all the short illustrations as fict ional.
To assist him i n doing so I prese n t them in a form tha t one of my stu-
dents named AckoH's Fables . In these fab les I use '' Aesop ," appro -
priately I hope, as a pseudonym for the creative prob lem so lver.
Not all the fab les have happy endings ; they are not all success
stori es. Creative so lutions are often not accepted . This is not surprising
in view of the widespread resistance to change, particular ly to some-
18 The Art
thing new and unconven tio nal. Because o f such resistance the crea tive
pro blem solver is not likely to be successful unless he is al so compe-
tent, commun icative, concerned, and, most of all, cou rageous.
One fi nal note : many of the cases used here have been used in pre-
viou s writings bu t in di fferent forms and for di fferent p urposes. There-
fore, those few reade rs who may have read any of my previous books
may recognize some of the sto ries. I hope th ey w ill be greeted as old
friends rather tha n redunda nt bores.
CHAPTER TWO
Objectives
Problems are o f two types: thos involving the destruct ion, removal, or
containment of something that is presen t but not desired, and those in-
volving the acqui sition o r attainment of something that is absen t but
desired. The first type of prob lem, one tha t is negativ ly orien ted, is
concerned with eliminating a source of dissatisfaction, for exam ple, a
distractin g noise, an il lness, or a deb t. The second type, on e that is
positively oriented, is co ncern d w ith atta ining access to a sou rce of
satisfaction, for exam ple, a friend, a good book, or money.
Positive and negative-applied to objectives-appear to be rela tive
concepts. For exam ple, the desi re to get rid o f no ise may be stated as
the desire to obtai n q uiet and the desi re to cu re an illness as the desire
to obtain heal th . Howeve r, one should be wary of such gl ib equatin g.
20 The Art
If one does not want to hear the music being played on the rad io, get-
ting rid of it is a negative objective. If one wants to hear a d ifferent
type of music, this is a positive objective. Note that this positive objec-
tive implies the attainment of one that is negative, but the converse is
nol true. In most cases, getting rid of what one does not want is no!
equivalent to obtaining what one does want.
The major difference between positive and negative objectives,
however, is not logical but psychological. We explore this difference
later.
~ Although positive and negative are not completely relative con-
cepts, "ends" and "means" are. Every means can also be considered to
be an end, and every end a means. for example, buying an automobile
can be considered a means for obtaining personal transportation, an
end . Personal transportation can be considered to be a means for get·
ting to work, an end . Getting to work is a means for obtaining income,
and so on. Every less-than-ultimate end can be cons idered a means to
a more ultimate end .
The car-purchasing example also shows that every means can be
considered an end . The first means in the example, buying a car, can
be considered the end of going to the dealer. Going to the dealer can
be considered the end of borrowing a friend's car, and so on .
. That every means is an end can be seen in another way. It was
pointed out previously that decis ion makers are concerned with the
efficiencies of available means for possible outcomes . The efficiency of
a means for a desired outcome, an end, is, in most general terms, the
probability that it will produce th at end. The higher th is probability,
the more efficient the means . Thus the efficiency of a means is a meas·
ure of its instrumental or extrinsic value, its usefulness. The extrinsir
1 value of anything lies in what it can be used for, its ab ility to bring
about something else that is valued . For example, the extrinsic value of
money lies in the value of what it ca n buy. for most of us money has
virtually no other value. For coin collectors, however, certain types o
money have values that are unrelated to what they can be used for, bu:
ar~ related to what they are in themselves.
If means (courses of action) and instruments used in them-such ai
money, tables, automobiles, language, and arithmetic-had only extrin-
sic value, decision makers would be indifferen t to choices involving
them when they are equally efficient for the same ends, but this is fre-
quently not the case. For example, although each of a number of di/·
ferently co lo red shirts, identical in every other respect, has the saJllf
efficiency as clothing as any of the others, we may nevertheless \mfi
preferences among them. We like some colors better than others; th
Objectives 21
are more satisfying. Thus the desire for a particular co lor shir t may be
said to be an "end in itself." The value we p lace on the col or is intrin-
sic, noninstrumental. Similarly, we p refer one symphony to another
even though neither may be useful for anything. Intrinsic val ue derives
from immediate satisfact ion, extrinsic value from anticipated conse-
quences.
The intrinsic values of means are seldom explicit ly formulated. They
are difficult to formulate because, among other things, a large number
of them may be relevant in any situation. Fu rthermore, they are very
personal values that are difficu lt or impossible to defend. Th e impor-
tance of such values and th e difficu l ty of formulat ing th em in advance
are illustrated in the fol lowing fable .
A small company that manufactured hand tools for carpe ntry was
owned by three men who served as its president, vice president and
financial officer, and v ice president of operations. The former two were
also presidents of two banks in the small community in which the co m-
pany was located. They spent little time at the company. The v ice presi -
dent of operations was there more than the others but wa s se ldom
there more than half the time. The absence of the owner-executives was
not due to lack of concern but to their recogn ition of the fact that the
company could run itself quite well without their involvement.
The vice president of operations, in an effort to do someth ing use-
ful, began to play w ith some relevant numbers. He found th at although
the company's share of the market in which it ope rated had been in-
creasing regularly for many years, the tota l vo lume of industry sales
was declining. By looking ahead he coul d see that unless there was a
significant change in demand for hand tools, the growth of h is com-
pany could not continue.
He discussed th is apprehens ion w ith his two fellow executives. To-
gether they decided to play it safe by diversify ing, by add ing a product
line that would enable the company to maintain its growth even if de-
mand for carpentry tools continued to decrease . They call ed on Aesop
and his colleagues for assistance in this effo rt.
Aesop and his team began by ana lyzing the co mpany' s operations
to familiarize themselves w i th its product ion technology and modes of
distribution and sales. Once th is wa s comp leted they began a search
for other products that required simi lar techno logy an d mod es of dis-
22 The Art
tribution and sales. The search was directed at finding a product that
"fit" the company and whose sales were increasing.
The first one they found was obvious: tools used by plumbers. They
presented this finding to the three executives hesitantly because they
doubted that it had not already been cons idered . The executives said
that, indeed, they had considered it but they had hoped for something
"bette r." Aesop asked what "better" meant. The president said that he
could not possibly define it but that when Aesop came up with an
alternative that was "better" the executives would recognize it easily.
He asked that the search continue.
Aesop took the search into more exotic tools and came up with
those used in aircraft maintenance. When this was presented to the
executives they expressed more, but not great, interest, and again they
suggested that the search continue . When Aesop pressed for specifica-
tion of the criteria they were using, he was put off again in the same
way as before. The search was continued with several more repetitions
of proposal and rejection .
On his way to a subsequent meeting with the executives, Aesop,
l istening to his car radio, heard a discussion of the significance of the
then-recently-developed transistor. This gave him an outrageous idea.
At the meeting with the executives he suggested that they go into the
transistor business. The response was enthusiastic and was immediately
followed by the question, " Can you tell us what a transistor is?"
In the subsequent discussion Aesop learned that the executives
were looking for a challenge and an opportunity to get reinvolved in
their business by the introduction of a new, no t a famil iar, technology.
As one of the executives pu t it, "We'd like to get more fun out of the
business and fee l more a part of it."
Eventually, at the suggestion of Aesop and his co lleagues, the com-
pany went into the hydraulic-coupling and valve business. The thret
executives lived happily ever after.
Objectives 29
In 1957 Aesop spent some time in India at the invi tation of its govern -
ment to review its national development-p lanning p rocedures. Wh il e
there he met a num ber of foreigne rs who were trying to introduce
fam ily planning to India in the hope of bringing its popu lation explo-
sion under control. Most of these serious and dedicated people were
Fable 2.4. Well read can well mean dead. Moral: Whal appears lo be
black on white lo Whiles may appear to be all white lo Blacks.
Aesop devo ted his effort to try ing to fin d a way to ma ke the chi
dren want to r ad . H e fou nd s vera l ways o f do in g so . In one, a num-
ber of si lent films that had b en made for ch ildren's matinees in the
nin eteen twenti es were show n co ntinually in th e school 's audi torium.
Any child was perm itted to watch them w henever he or she wanted tf!
To understa nd the m, however, it w as necessary to b ab le to read the
subtitles. Thi s provided the necessary mo tiva tion to many of the chi~
dren, and they bega n to learn how to read.
Objectives 37
MORAL: What appears to be black on white to
Whites may appear to be all white to
Blacks.
Much of th is book w as wri tten w hile I was spending a sab bat ical year
doing research an d tea ching in Mex ico. In Mex ico I ran ac ro ss !he fol-
lowing exam ple of " we-they'' in trac u ltural misun derstan d ing.
The cultural gap that keeps p roblem sol ve rs from un derstand ing the
objectives of othe rs can occur even w ith in relat ively sma ll and appa r-
ently homoge neous organ izations. A n examp le is pro vide d in Fable 3.4 .
In problems the s lut ions to which invol ve the reac tions of others,
38 The Art
Fable 2.5. Those who give a dam may not be well received. Moral: The
reason for one person's giving may nol be the reason for anolher's re-
ceiving.
One party is in conflict with another if the action of the first party has
an effect on the outcome of the action of the second, and this effect is
to reduce the value of the outcome to the second party. If, on the
other hand, the effect of the action of the first party is an increase in
the value of the outcome of the second's behavior, the fi rst cooperates
with the second. If the first party's beh avior has no effect on that of
the second, the second is independent of the fi rst.
These relationships are not necessarily symmetrical; for example,
one party may cooperate with a second while the second confl icts with
the first. The master-slave relationship is an extreme case of this.
Where conflict or cooperation is not symmetrical, the one who bene-
fits most or suffers least is said to exploit the othe r. Exploitation need
not be deliberate or conscious.
Competition is frequently defined as "conflict according to rules''
or "constrained conflict." These definitions do not cap tu re the essen-
tial property of competition. In competition two or more pa rties are in
conflict with each other relative to one or more objectives of each, but
this conflict efficiently serves another objective of either the conflict-
ing parties or another party. For example, the opponents in a te nnis
match are in conflict with respect to winning, but this conflict is effici en t
relative to the recreational objective they share. They cooperate with
respect to recreation . In economic competition the companies in-
volved are in confl ict with respect to their desire for sales, but such
conflict is supposed to serve efficiently the interests of consumers by
providing them with better services at a low r price than they would
otherwise enjoy.
Laws, regulations, and rules that govern competition are intended
to ensure the effectiveness of the conflict involved for the "coopera-
tive" objective. In the American economic system , for example, price
frxing is precluded because this is believed to be detrimental to the in-
terests of the consumers. Similarly, the rules of tennis, or any other
game, are intended to ensure their recreationa l value for either the
participants taken collectively, or their spectators, or both .
There are three ways of dea ling with confl ict, and these correspond
to the ways of dealing with problems in general: solution, resolution,
and dissolution .
To attempt to solve a conflict of which one is a pa rt-whether or not
it is imbedded in competition-is to accept the conditions that produce
the conflict and to seek a way of getting what one wants, whatever the
40 Th e Art
cost to the opponent. There fore, to try to sol ve a confl ict is to try to
win it. This may be done with o r wi thout the use of force To try to win
without the use of fore has come to be called "gamesm anship." To
use for e is to fight. In a fight one or bot h opponen ts try to eli minate,
re move, in ca pacitate, o r inactiva te the o th er-to overcome him. A figh1
may be avo ided or discontinued by submi ssion or surrender of one ol
the parties. Fights may be prevented or stopped by the forceful in ter-
vention, or threat of such interven ti on, by a th ird party, usually stronger
tha n the opponents.
Efforts to solve J conflict genera lly intensify it until one party de-
feats the other. Therefore, solu tions frequently create condi ti ons tha1
b reed future co nfl i t, and subsequen t conflicts are often mo re intense.
Co nflicts th at ar deterred by force or th reats of force al so tend to in-
tensify over time.
To attem pt to resolvl' a conflict is to accept the condition thal
create it and to ek a compromise, a distribut ion of gains and/or losses
that is acceptab le to th e participants. Each party usually gives up some-
th ing it w anted but obtains someth ing it might not o therwise have ob-
tained . Agreement on a settl ment is normall y reached when the pa ·
tic ipants feel that the proposed distrib ut ion of ga ins and losses is fair-
that is, it elim inates or red uces exploitation.
Negoti ation , barga ining, and arbit ration are com mon processes b
which reso lu tions to onfl icts are sough t.
The th eory of gam es is a body of sci entific know ! dge developed
since World W ar II that is ap plicabl t the search fo r settlements. Re-
solvin g settlem nts usually stabilize con fli ct, bu t they s ldom deinte ·
sify them as dissolution can.
To di ·o/ve ,u:0nflict, the co ndition tha t produce it are changed so
th at it disappea rs. This ca n be don e by changi ng either th en iron·
mentor the oppon en ts.
Many co nfl icts are a con seq uenc of scarc ity. Two children want lo
play wi th th P same toy, the only one of its ki nd in the environment
Two adu lts wan t the last remain in g pa rkin g pace in a lo t. Such co
fl icts can some times be d is Iv d by rem oving th scarcity, for ex·
am ple, ma king another toy or parking space availa ble .
Some co nflicts are based on an invasion of the sen es. For exa mple
one person wants to read in a quiet room in which another wa nts 10
listen to bl aring music. Such conflicts can sometimes be dissolved b1
" divid ing" the environment. For exam ple, a pai r of earp hones mig'
all ow one to hear music at whatever vo lume he desires wh il e the othe;
read s in peace. A physical divisio n of the environment can sometime!
accom plish the same resu lt, for exa mple, the u e of a curtain to blod
Objectives 41
annoying light turned on by another, or clo si ng a door to shut out a
ise. In dissolut ion o f conflict neither party loses; both ge t what th ey
nt without compromise.
Not all conflicts result from a clash of objectives that have been
en freely, particu larly conflicts that arise between parts of the
organization . Such conflicts are often imposed on the opponents
a third party who has some contro l over both of them. Conflicts of
· type are often unintentiona l. Here is a case in point.
isisa story about a cha in of department stores that bought and sold
wide variety of products. Since the number of i tems handled has no
evance to the story I tel l it as though only one product were involved.
The two basic operations of department stores are, of cou rse, buy-
from suppliers and selling to customers. These were hand led in this
by a purchasing department and a merchand izing department
objective given to the purchasing departm ent by the chain's execu-
officer was to minimize the average size of inventory w hi le meet-
expected dem and . Therefore, the most critical variables control led
the purchasing department were the size and frequency of pur-
ases.
The objective assigned to the merchandizing department was to
imize gross pron t: (number of i terns sold) times (sel I ing price mi nus
rchase price). Its principal contro ll ed variable was selling price.
The merchand izing department had a stat istica l sec tio n that plotted
number of items so ld against sel ling pri ce in each period. Thi s
ded a price-de mand curve such as the one shown in Figure 2.1. The
''optimistic" line represented the largest numbe r of the i t m that
store could reasonably expect to sell. The middle " realistic" line
resented the "expected" or "average" amount sold at each price.
lower "pessimistic" line represented th e smallest amuunt the store
Id expect to sell. (The upper and lower lines were drawn so th at the
bability of going above the top or below th e bottom was O 05. )
Themerchandizing manager-let's ca ll him Mike- would decide what
rice to charge for the item some time in advance. Suppose he se lected
·ce Pi shown in Figure 2.2. He then had to decide how many of the
he would lik to have in stock at the appropriate time. Using Fig-
2.2 he selected quantity Q, to be sure that he would not run out
stock. If he ran o ut of stock, his measure of performance would
42 The Art
I \
\
\ \
\ \
\ \
\ \
\
' \
\
\ ''
\
''
' ........
....... ...... _
' ......
.............. ..... _ - - - -
-
-
Op tim istic
Reali me
Pessim;stlc
Price
Figure 2.1.
\
Q,
1,. \
I\\
'h. ,
I ,
' ' ',
......
, .......
............
' ...........
---- --- - -
- - - - Pessimist ic
Opt imistic
R al ist ic
Pr ice
Figure 2.2.
Objectives 43
suffer. Having made this decisio n he notified Pat, the purchasing man-
ager,of his desire for Qi of the item at the specified time.
Now Pat was the former assistant merchandizing manager ; he also
had a copy of Figure 2.2, and he knew how Mike had selected Qi. Pat
consulted the figure and decided to order only quan ti ty Q2, the ex-
pected number of sales at price P1 . He wanted to avoid being over-
stocked because this would hurt his performance.
Mike, however, knew that Pat wou ld do this so 'he really intended
to charge price Pz which would maximize his expected gross pro fit if
only Qi were available. Pat, on the other hand, knew Mike w o u ld go to
price P2; therefore he intended to order only Q J, the expected n umber
of sales at price P2. But M ike knew that Pat w ou ld do th is . . . and so
on. The limit of this process would have been reached when no thi ng
was bought and therefore noth ing was sold . The limit was not reached
because the two managers were not allowed to communicate with
each other. Each had to guess what the other would try to do. Altho ugh
lhishardly made for optimal operations, it made survival possible.
2. Once each participant can state the other's pos 1t1on to the
other's satisfaction, each should formulate the factual and/or
moral conditions under which he believes the other's position
would be valid.
They may agree on these conditions. For example, they may agree
to examine four classes of states over the last ten years: (1) states that
have used capital punishment continually over this period, (2) those
which have changed from no capital punishment to its use, {3) those
which have not used it over this period, and (4) those which once used
i t but subsequently abo li shed it. They may also agree that if the average
percentage increase in capita l crimes per year in states that chang
from no capital punishment to its use is less than the corresponding
average in each of the other three categories, capita l punishment is an
effective deterrent ; othe rwise it is not.
If the opponents cannot agree on how to determine the match be-
tween actual and dissolving conditions, this disagreement should be
made the subject of a debate that follows the procedure described
here. Once this issue is resolved the original debate can continue.
Presumably, the relevant facts will settle the iss ue. If such a determ·
nation is not possible or not feasible, the participants can proceed
the nex t step.
Justifying Conditions
opponents agree, they should accept the pos1t1on that has the ·
least serious error associated with it. If they do not agree, this be-
comes the issue they should take back to the first step.
SUMMARY
Controllable Variables
After World Wa r 11, as affluence increased in the British Isles, its in·
habitants becam e less committed to discomfort : cen tral heating and
Controllable Variables 51
.
~:
Fible 3.1. A fishy story. Moral: One is often moved deeply by another's
hunger.
The manager of a large office building had bee n rece1v 1ng an inc reas-
ing number of complaints about th e building's elevator service, particu-
larly du ring rush hou rs. When several of hi la rge r ten an ts threatened
lo move out unless this se rvice was improved, the manage r d cided to
look into the problem .
He called on a group of consulting engineers who specialized in the
design of eleva tor systems . After examin ing the situation, they iden-
54 The Art
tified three possib le courses of action : (1) add elevators, (2) replace
some or al l of the existing el valors with faster on es, or (3) add a cen-
tra l computerized contro l system so tha t the elevators cou ld be
" rou ted " to yield faster service.
The engineers then conducted cost-benefit anal yses of these alter-
natives. They found that on ly adding or replacing elevato rs cou ld yield
a large enough improvement of service, but the cost of doing eithe:
was not justified by the earn in gs of the build ing. In effect, none of the
alternatives was acceptable. They left the manager with this dilemma.
The manager then did what a manager seldom does when he is any-
thing less than desp rate: he consulted his subordinates. He called a
meeting of his staff and presented the problem to them in the formal
of what he called a "brain-storming" session . Many suggest ions were
made, but each was demolished. The discussion slowed down. During
a lull the new young assistant in the personne l department. who had
been qu iet up to this point, t imidly made a suggest ion. It was immedi-
ately embraced by eve ryone p resent. A (ew weeks later, after a rela-
tively small expenditure, the pmblem had d isappeared .
Fu ll-length mirrors had been installed on all the walls of the elevator
lo bbies on each f loor.
The young personnel psycho logist had rea oned that the complaintl
originated from the boredom of waiting for elevato rs. The actual wait-
ing time was quite small. but it seemed long because of th e lack of any-
thing to do while wait ing. He gave people som ething to do : look al
themse lves and others (pa rt icularly of the opposite s x) w ithout ap-
pearing to do so. This kept them pleasantly occupied.
Our natural inclinat ion is to try to find the cause of the de fic iency thal
gives ris e ro a problem and to remove or suppress it. We did this un-
successful ly in the case of Prohibition and we are doing it again with
respect to th e misuse of narcotics. Unfortunately, even if we get rid of
what we do not want, we do not necessar ily obtain what we do wanl
When I change a television channel to remove a d istastefu l p rogram, 1
seldom get one that is sat isfacto ry. Healt h is more tha n the absence of
d isease, even thou gh many doctors act as though they are equ ivalent
Efforts to remove defic iencies are no t without success, but there is an
alternative approach to prob lem solving tha t should also be considered
Controllable Variables 55
use it is often more effective : adding something to, rather than
btracting something from, a problematic situation to convert a ma le-
orinto a benefactor. Cons ider, for example, this fable.
The Port Authority of a major American city cont rol s, among other
things, all bridges and tunnels into its metropolitan area . Like most
Controllable Variables 57
alternative. He called for help from an externally based research group
that was working on another unrelated problem in his plant.
The researchers found that most of the women were married and
their husbands were employed and earned enough money to provide
their families with necessities. The women were w9rking to increase
the amount of money available for discretionary goods and serv ices,
things they wanted but did not really need . The women did not want
to earn as much as their husbands because they be li eved this would
threaten their husbands' self-i mage as breadwinners . (Th is was before
the feminist movement got unde,· way.) Thus they were not anx -
ious to earn more than they were alredy earning. Furthermore, working
carelessly at a le isurely pace enabled the women to socia li ze with co -
workers while doing their jobs, thereby getting some relief from their
dull and repetitive task .
More important, the researchers discovered that most of the women
had children in school and felt very guilty about not be ing at home to
receive their children when they returned from school. The ch ildren
either had to take care of themselves, which was a cause of anxiety for
their mothers, or they were car d for by others on whom the mothers
felt they were imposing. In either case the women held the company
responsible for creat ing th is uncomfortable and gu ilt-ridden situation .
When the researchers learned th is, they des igned a new incentive
system. A " fair day's work" -the number of items correctly inspected-
was specified . It was set at the highest leve l o( output the women had
previously attained. However, the women were permitted to leave
work and return home whenever they finished the speci fied output,
or they could continue on a piece-rate basis for as much additional
time as they des ired and product ion requirements permitted . The
women accepted this proposal enthusiastically . Their inspection rate
more than doubled, and they were out of the plant in plenty of time to
meet their children returning from schoo l. Errors decreased and satis-
faction increased.
One of the opera tin g companies of the Bel l System asked Aesop to try
loreduce the amount of stoc k carl'ied in trucks driven by service men
who called on subs ribers to repair, replace, in stall, or remove te le-
phones. In a no rm al d ay a servicema n ma de o nly a relatively sma ll
number of ca lls. evertheles , beca use he did no t know what parts he
would requ ire and because his com pe nsa tion w as based on the num-
berof calls h completed per day, h filled hi tru k w ith eve ry part he
could get into it.
60 The Art
Aesop des igned a new compensation system to replace the old one
He made a repa irma n's compensation a function of both the num~
of calls he completed and the value of the stock he carried in his truck
the la rger the stock, the less the compensation. This yielded very largt
reduction s of in -t ransit inventories, with only a very small reduction·,
the average number of calls completed per day.
The system involved in each of the previous two fables p rovid ed incen-
tives that produced the behavior that cau sed the problem . Of courw
they were not intended to do so. They were designed to serve one o~
jective in situations in which more than one object ive was operating
An incent ive that serves one object ive may we ll serve another ve~
badly.
The design of an effective incentive system clearly requires undei-
standing the objectives of those to be "in cented." Th is is obvioui
What is not so obvious is that design of an effective incentive syste
also requires the designer to have a clear understanding of his m1•
objectives, as the next fable shows.
At approximately the same time two very diffe rent schemes were df.
veloped to reduce automotive congestio n in New York City and Lon-
don. The New York proposa l cons isted of placing electronic sensi~
devices along streets in the city. These devices were to read magnetizer
numbers placed on the sides of automo biles and transmit them 101
cent ral computer that wou ld co ll ect and use them to prepare char
to car owners for the use of city streets. It was intended that t~
amount billed per read ing would depend on the time at which
reading was made---the day of the week and the time of day. Charg!
wou ld be greatest when congestion was greatest. It was hoped that
this way the use of automobiles during peak hours would be disCOli'·
aged. Needless to say, the sen sing-commun icat ion-compu ting syst
requ ired by this design was both very complex and ve ry costly.
The scheme developed for London was based on instal ling metei,
in each car so that i ts mileage in the city could be determined perioo-
cally, as, fo r example, al car-i nspection time o r when leaving the ci,
Charges would be made on a mileage basis . This was the simpler of
Controllable Variables 61
two schemes but its rate structure was not adaptive to traffic conditions
aswas the scheme developed for New York .
Aesop was called on to evaluate these alternatives and select the
better of the two. He o bserved that both schemes charged automobiles
for "going." Congestion, however, is not measured by how much cars
go; rather it is measured by how much they stop. Therefore, Aesop
a,gued, what is needed is a disincentive to stopping, not going. This, he
showed, could be accomplished by the use of a very simple metering
device that would count the number of stops a car made and charge a
fixed rate per stop. Suc:1 a system would automatically take account of
baffic conditions. If a trip were taken at peak hours on a work day, its
cost would be high because of the number of stops required. The same
trip taken at off hours would be much less expensive. Furthermore,
such a scheme wou ld encourage the use of less-traveled routes, and it
would encourage drivers to try to avoid stopping at liehts by slowing
down when approaching one that is red . This would increase through -
put at lights where start- up time of vehicles at rest currently consumes
much of the "go time."
It is easy to fall into the trap into which the designers of the disincen-
tive system involved in the last fable fell. Subtlety and sophistication in
incentives are generally not as effective as sim plicity and d irectness.
The intentions of an incentive system shou ld be obvious to those who
are supposed to be affected by them.
It is important to get one's objectives straight not only before de-
signing an incentive system , but also before designing any system that
might be perceived by others as providing an incentive, however un-
intended. For example, consider the way property is taxed in most of
the United States. Such taxes are generally proportional to the assessed
value of the property. Therefore, they provide an incentive that en -
courages allowing property to deteriorate. That such taxes ope rate in
this way is apparen t in urban ghettos . No wonder we have so much
rapidly deteriorating housing in most American cities .
If, in addition to taxing, the value of property, we also taxed the re-
duction of its value over time, we would provide an incentive for its
maintenance and improvement. Some countr ies offe r tax exemptions
tor the maintenance and improveme nt of residential property, much as
we provide them for business properties .
What is wrong in a problematic situation is often far from apparent,
62 The Art
F.ible 3.9. It may hurl to bend a little. Moral: It is easy to blame others
for our own mistakes, but it is hard to correct them by doing so.
Ashort time later the manager called Aesop and said that the room
was ready. They met with mock ceremony and entered the room.
Neither said a word ; they ju st looked . W ithi n a minute or so the man-
ageradmitted that Aesop's assertion about consumer rational ity held .
What they had observed was that every one of the succe sful ap pli -
11ces could be used without bending or climbing, whereas every o ne
oithe unsuccessful ones req uired be nd ing or climbing. The dishwasher,
llthat time, required load in g by putt ing dishes into it from the front;
· could not be loaded without squatti ng. On the other hand, the
64 The Art
bui lt-in oven could be used without bending, whereas the one in tlw
old-fashioned range could not.
Before Aesop and the manager left the room, the manager initiate(
work on modifying both appliances. First, he asked that a d ishwasher
be designed with pull -out drawers so that it could be loaded withOll
squatting and, second, that a cooking range be designed with the ows
and broiler above, not below, the cook ing unit, using the vacaec
space below the burners for sto rage.
Despite this fable, consumers are not always rational, but in my expen,
ence I have found them to be so more often than producers. Coo,
sumers have more chances to learn by tria l and error than produc~
Many producers devote the ir marketing efforts, particularly advertising
to trying to convince consumers that significant prod;.;ct differen~
exist where they do not. The curious thing is that such effo rts succe«
in convinci ng the producer more often than the consumer. Produc
often come to believe the decept ions they pay to propogate.
Gasoline is a case in point. It is apparent that if we filled identia
makes and models of automobile with different brands of the sa
grade of gasoline, not even the presidents of oil companies could ider,
tify the b rands by their perfo rmances . Yet their advertis in g tries to COll-
vin ce consumers that there are significant performdnce differences be-
tween brands.
Many years ago some of my colleagues conducted resea rch for
oil company to determi ne what led consumers to the se lection of tli
service station they norma ll y used. They found that one variable, na
brand, had a greater effect than all the other variab les put together.
Fur th ermore, this variab le made it clear that consumers were behavi111
ra tionally, given no di fference in brands. Th is variable and its discove11
are revea led in Fab le 5.4 .
Where there are no significant differences between alternatives oi,
fered by producers, it is more effective in general to invest in creatint
such differences than to try to convince others that they exist whei
they do not. Here is a case in point.
and many others have argued that the accelerating rate of technological
and social change is now th e source of our major societa l and organi-
zational difficu lt ies . Many of these changes are occurring more rapidly
than we adapt to them, hence the dislocations and lags from which we
SU ff er.
Altho ugh a rapid rate of change crea tes many current problems, it
also creates many opportun ities. These can be anticipated by the use of
what I call "re ference project ions." These are extrapolations from the
past into the future under an assumption we know to be false : that
things wi ll continue to be done an d to happen much as they have in
the past , that is, without significant interventions by decision makers
or the environment. Because reference projections are based on a false
assumption, they do not yield predictions of th ings to come ; they do
predict what will not o r is not likely to happen. They dq this by show-
ing how a sys tem would break down if it were to continue to operate
in the ( u tu re as it has in the past. Such breakd owns are not I ike ly to oc-
cur because there will be interventions. By revea li ng these points of
future b rea kdown now, effective interve ntions can be planned before
a future crisis forces a rapid and Iess effective intervention.
For example, in a projection of the paper work performed by the
Federal Reserve Bank of Cleve la nd, it was found that the amount of
space required to do the amount of work p rojected to the year 2000
wou ld e>"ceed the amount of commercia l space available in Cleveland
at that tin-e. Thi s "reve lation" helped spur the development of an Elec-
tronic Funds Trans fer System which would greatly reduce the space re-
quirement and has a number of other advantages as well.
In a p ro jection of the number of automobiles in the United Stales
in the year 2000, it was found that there would not be eno ugh surface
in metropo li tan areas on which to place them. Such a projection, of
course, is not a forecast of what wi ll happen but, as noted, of what will
not. What will ha ppen depends on decisions to be made between now
and then . The recognition of a iuture cri is can induc u to th ink now
about alterna tive cou rses of action , crcativ on<"s that might ave rt the
crisis. These often invo lve the fundamental r·edesign of the system in-
volved, as is the case involving the automobile that is presented in
Chapter 7.
Reference projections are not so much a way of fi nding so luti ons to
problems as they are a way of identifying the problems that ought to
be solved . Recall that formulating the right problem is at leas t as impor-
tant as gett ing the right solut ion . Such projec ti ons also reveal what
controllab le aspects of a system or a situation require change.
There is one aspect of reference p rojections that requires caution:
Controllable Variables 67
lhe changing meaning of conce pts. For examp le, in the 1950s a projec-
lion was made that showed that ev ryon e in th e United States wou ld
soon be a scien tist if cur ren t trend co ntin ued . O n the face of i t this
projection seemed sil ly, but it was not. Suppose that at the time of
Newton a similar projectio n h,1d been made of the number of mathe-
maticians in England today. In all likelih ood it would have shown tha t
all Englishm en wou ld be mat hematicians. In a ense th is would have
been close to correct. Many, if no t most, peop le in England today know
as much mathem atics as Newton did (without necessari ly unde rstand-
ing it as he did), but those now called mathemat icians know a great
deal more. The requirements for being a math ematicia n have also
changed.
Therefore, in re ference projecting one should be carefu l to deter-
mine wheth er th e meaning of the u:oJ.ts being pro j cted are subject to
change over the period covered .
To this point we have dealt only with constraints imp li citly. Now let
us consider th m expl icity- first, thos se lf-imposed co nst raints of
which we are not wa re.
As previou sly mentioned, our selec tion of th e variab les to man ipu-
lale in any given situation is cond itioned by ou r previous expe ri ence
and our education. As a result we unconsciously precl ude from con-
sideration certain types of co ntrollable variahles and confine ou r atten-
lion to certain tradi ti on al ones. Co nsider the nex t example.
Over time several th efts of money and other valuab le had been com-
mitted by employees of a large financia l institution d es pite an exten -
sive security system. For examp le, th ere we re check-in and check-out
points to cover the en try and exit of everyone, numerous armed guards
posted through ou t th e fa cility, cl o d-circu it t levi si on surveilla nce of
all work areas, multi ple doors to valua b les each I eked w i th keys held
by separate guards, an d so on . The cost of su ch a system was very la rge,
yet it was "defeated" with regu larity.
The thefts were almost always comm i tted by sm al l tea ms of em-
ployees because the tasks in the institution were so d ivided, and so
many internal checks were made that it was almost impossible for a
single individual to cover a theft long enough to get out of the facilities
or, if he did, to avo id rapid identi fication .
Research was initiated to find a way of tightening the institution's
security. The al ternat ives conside red were the obvious ones. These in -
68 The Art
eluded such things as the number, locat ion, and rotation of guards,
more extensive use of electronic detection devices, and the rotation of
work assignments at random within the bank at frequent enough in-
teNals to make th e formu lation of crimina l coalitions less li kely.
Aesop was asked to review this project. After the presentation, he
asked how the thieves removed what th ey stole from the bank. He was
told that the stolen valuables were usual ly concealed on the person or
in his or her persona l effects. Bill s of large denomination, for example,
do not require much space . Why, Aesop asked, is there not a more
thorough search at the po int of ex it ? He was told that, beca use of the
number of employees, th is would be very time co nsum ing and would
not necessarily be effec tive unless a person was exam ined in the nude.
Aesop asked why they d id not work in the nude. Everyone laughed, but
Aesop continued: "I know this is not feasible, but it suggests that all
employees be required to wear a uniform changed in to and out of at
work. This cou ld make inspection in the nude 'natura l' and unobtru-
sive." Aesop went on to explain that uniforms cou ld be designed to
facilitate work, save on personal cloth ing, and be attractive.
The solu tion suggested by Aesop was not accepted. Instead , survei l-
lance was sign ificantly increased. The thefts co nt inu ed.
cannot help but reca ll the old story of the worker at a large factory
who left work each night pushing a whee lbarrow fu ll of waste. The
guard at th e gate noted this repeated event and even inspec ted the
waste but cou ld fi nd nothin g of value in it. It w as only much la ter thal
he discovered that the worker had been stea ling wheelbarrows.
Now co nside r a case in which the suggestion that someth ing com-
monplace and trad ilional not be taken for granted was taken seriously.
h is analyses and the co nclusion he had dra wn from them . The vice
president was not surprised by these results. He asked Ae o p if h had
considered who were the purchasers of the unprofitab le p roducts.
Aesop admitted that he had not. We ll, the vice pres ident said, if Aesop
had done so he would have found these purchasers to be the major
consumers of the company's profitable products . He added that he
could not afford to run the risk of losing these cus tome rs by discon-
tinuing the unprofitabl e products. Aeso p asked how he knew that any
risk was involved. The vice pres ident admitted tha t he had no such
know ledge, but he was unwill ing to drop products to find out. Aesop
asked if he would consider dropp ing products if a risk less way of do ing
so could be found . The vice p resident sa id he w ould be wil ling to do
so but he did not think there was any such way. Aesop asked for some
time to th ink it over; he was down but not yet out.
In thinking about the problem Aes op reca lled that the company's
salesmen were paid a fixed sa lary plus a commission based on the
dollar value of their sales. This meant they earned as much fo r sell ing
a dollar's worth of unprofitab le products as for selling the sam e amount
of profitable products . Because the unprofitable products were " under-
priced " and therefore a good buy, Aesop suspected that they received
a disproportional amount of the salesmen's attention . (Salesmen had
no information about each p rodu ct's profitability.) This led Ae op to
consider a compensat ion plan for salesmen w ith th ese charac:eristics :
The profit-based incentive system revea led that a constraint that was
apparently imposed by customr-rs was actually imposed by the market-
ing manager.
There is an old story the moral of which is not unrelated to that of
the previous fable. A drunk was on his hands and knees craw li ng about
under a street li ght on a da rk night, obviously looking for something.
Apasserby stopped to asK him what he was looking for. He said he wa s
looking for his keys. Afte r a moment he added th at he had lost them
back in a nearby alley. The passerby asked h im why, then, was he look-
ing for them where he was. He answered, "Beca use it's light here."
Our conceptions of what can be done in problematic situatio ns are
often limited by constraints att ri buted to technology. We frequently
forget or overlook the fact that technology and its use are controllab le.
What appears to be a technologica l constra in t to one person may not
appear so to another.
I have heard, for example, that the makers of Scotch Tape d id not
envision most of the uses to whi ch their pro duct was put. Inven tive
consumers discovered these uses, but the company was resourceful
enough to adapt th eir product to these uses .
Problems are often caused by the misuse or abuse of technology.
We often try to so lve p roblems by proh ibiting or suppre sing such mis-
use or abuse. The intended results are often more easily obta ined by
minor modifications in the technology involved . For example . consider
photocopying equipment.
Anyone who has access to such equipment can rep roduce copy-
righted materia l without paying the roya lti es to w hich the copyrigh t
holder is legall y en titled . evertheless such practice is common. It is
virtually impossi ble to enforce a law that prohibi ts this practice o r to
collect royalties from those who engage in it.
The issue raised by current copy ing practices are being hotly de-
bated in Congress and among autho rs, publishers, lib rarians, educators,
and students, all of whom are significa n tly affected by the is sue . More
72 The Art
restrictive laws are not likely to be enforceable as long as photocopy ing
machines can be used in private. At best such laws can on ly be en-
forced in lib raries tha t are open to the p ubl ic and at commercia l repro-
duction services, b ut this woul d be costly and would cover only a small
portion o f current infractions.
Publishers have also sought a technol ogica l solut ion, for example,
an ink that could not be copied by the infernal machines. However, no
one has been able to find a practical w ay of prevent ing the abuse.
Recen tl y a research gro up at my university ca me up w ith a di fferent
app roach. It has yet to be developed in det ail, bu t it open s a new set of
altern at ive solutio ns.
Ph otocopying machi nes are of two types : machines that can copy
pages of bo und volumes and ones that cannot. To use th e latter, in -
dividual pages are usually fed through roll ers. The former usual ly in-
vo lves p lacing th e page to be copied, boun d o r un bound, face down
on a tran sparent p late. The roller-fed mach ine cannot be used to copy
pages in a bound document. Un bind ing the document is both costly
and tim e cons um ing.
Therefore, the research group proposed a tax per copy made on
non-roller-fed machines, w ith th ese exceptions : lib raries and commer-
cial rep ro duction services would col lect royalties fo r copies of copy-
righted material made on th eir mach ines an d forward them to the
appropriate places. This much cou ld be enforced.
The tax cou ld be based on perio dic rea dings o f the me ters on non-
single-sheet machin es. Most such machines are cu rre ntly rented and
cha rged fo r a per-copy ba sis. The number o f co p ies is determined by
reading a meter on the mach ine. Th erefore, the tax could be collected
by the machine sup plier and forwarded to the government. Once the
cost of collect io n and admin istrati on of the tax w ere deducted from re-
ceipts, the remai nder could be d ivided amo ng publ ishers propor-
tionally to the doll ar value of their sales as repo rted on their tax form s.
The payments cou ld be made in the form of co rporate tax allowances.
The proposed procedure would encourage the use of ro ll er- fed
copiers for no rmal use because no tax w ou ld be involved. It would also
pressure p ubl ishe rs to price their p roducts competitive ly w ith " repro-
duct ion-plu s-tax" costs and th erefore to seek new technology that
would red uce the cost of pri nting. Such pressures might even resu lt in
publ ishers using photoco pying eq ui pment for th is p urpose. For ex-
ample, document produc ti on migh t eventu all y be d ecentral ized and
copies be made to o rde r from a maste r co py suppl ied by the publisher.
Th is would remove one of th e largest problems currently faci ng pub-
lishers: decid ing how many copies of a documen t to make per print ing.
In most attempts to sol ve the illicit reprod uction of copyrighted
Controllable Variables 73
material, consideration has not been given to the use of a tax or the
difference between types of copying machines. Ono~ these are raised
to consciousness, a large number of possibilities is opened for con-
sideration.
Sometimes the technology required to sofve a problem lies in front
of our noses without our being aware o( it. We miss it because of our
preconception of where technology is to be found: among those who
are technologically educated. This is not always the case. He re are two
examples.
The more pride one takes in what one knows, the harder it is to learn
from others, particularly unlikely others, as the next fable shows.
Fable 3.13. The smart jackass. Moral: The easiest way to do something i,
often learned the hard way.
of the furnace. They were doing this on the shop floor whil the old
operator of the furnace, who had received no exp lanation of their ac·
ti vity, stood watching them . After severa l unsuccessful tries, the over·
heated and frustrated technicians took a break.
The old man then asked them what they were trying to do. They in·
dulged him and explained. He diffidently said that he cou ld tell them
how to do it without instrum ents. They indulged him aga in and asked
how.
Controllable Variables 75
The old man, w ho was chewing tobacco, xpectorated, placing a
large wad of spittle onto the side of the furnace and then consulted h is
wri t watch . In a few seconds he "a nnounced" the temperature. When
asked for an explanation, he sa id h is predecessor had taught him th at
by timin g the evaporation of spittle as he had done and mul ti p lying it
b7•a nu mber he had bee n giv n, h wou ld get the tem perature at th e
center of the furnace. The techn icians smil ed indulgently.
Days later when they had found a way to instrum ent the furnace,
they decided to have fun w ith the old man and test him. They did so
much to th eir regret. He was ri ght.
SUMMARY
terms, particu larly about organizations, and our searches for soluti on s
tend to be restricted by the dimensiona lity of our men ta l pictures. Such
dimensionality is no t inherent in th e system and is the refore subject to
our con trol. Increasing the number of d imensions in which we think
abo ut p roblems can often reveal new and more effecti ve sol utions.
In general, we try to si mplify our problem s by reducing the nu mber
of alternative solutions we consider. Sim pl ifying prob lems ca n p re-
clude so lutions better than those we consider. The consequen ces of
every simplifying assumption should therefore :.:ie seriously eval ua ted.
Simplification is o ften simple minded.
CHAPTER FOUR
Uncontrolled Variables
Back in the 1950s a team of resea rc hers of which Aesop was a part was
trying to determine fo r a ma jor domestic airline how often it should
hire girls, conduct classes to train them to be stewardesses, and how
~rge these classes should be. The attrition rate among stewardesses
was high. Most of them left the airline within two years, usually for
marriage. At that time stewardesses were not allowed to be married .
Those familiar with production - management problems will recog-
nize this school-scheduling problem as a bizarre insta nce o f the eco-
nomic-lot-size (or production-run-size) problem. Stewardesses can be
thought of as the output of a production process, the training sessions.
Thus the problem could be stated as : How many stewardesses should
bemade how often?
Asolution was found by Aesop and hi s colleagues that promised
significant but not dramatic savings to the airline . Jn the process of
finding the solution, however, Aesop and his associates learned that
~e average number of hours the stewardesses were in the air per
month was considerably les s than the max imum allowed under the
relevant un io n contract. By questioning others they found out why few
could fly the maximum allowable number of hours .
first, suppose all flights were eight hours long and that th e maxi-
mum allowable flying time was one hundred hours. Then a stewardess
82 The Art
could only be assi gned to twelve fli ghts w h ich , if they took no more
time than scheduled, would ta ke (twel ve X eight) ninety-six hours.
Four hours would necessa rily be lost; th e wo ul d be " unavoidable
scrap."
Second, supp se all flights were te n hou rs long and that ten of these
were assigned to a stewardess. If any on e of her firs t n in e fl ights was
delayed by more than a spe ifi ed amount of time, a com mon occur-
rence, she would no t be perm itted to take her last scheduled fli ht. A
reserve stewardess would have to be used . Therefore, som~ allowance
(b uffer) for the uncertainty of flight ti mes was requ ired .
Aesop and his colleagues then asked wh at w as the maxi mum aver-
age flying time that could be obtained from stewa rdesses under these
conditions. This was a very difficult question, but they were able to
determine that about a fifty percent increase was th oretically possible.
This encouraged them to try to find a way of combining flights into
monthly stewardess assignments that, in practi ce, w ould realize most
of the theoretically possible improvement. They fou nd such a way, and
it promised dramati sa vings to the airline.
With considerable self-satisfaction Aesop p resented these results at
a meeting of the ai rline's department heads. II bu t one was enthusi-
astic about Aesop's unsolicited effo rt. The h ldout was the director of
personnel , who said that the p rocedure p resen ted cou ld no t be used
because it would make the steward esses leave the ir jobs earlier than
they would normally, and this woul d increase recr ui ti ng and training
costs by more than Aesop's ass ig nment procedu re would save. He said
that most stewardesses woul d r fuse to fly as many ho urs as the pro-
posed procedure wo uld requi re o f them.
Aesop was not pre pared for this ob jection. The possib ili ty of ac-
celerated attrition had not occurred to him o r h is colleague . There-
fore, he felt obliged to save face by asking the directo r of personnel
how he knew the steward esses wo uld rea ct as he ha d sa id . The Direc-
tor's reply was: "My God ! Everybody knows thi s." Aeso p p ressed and
asked him how he knew this . The director blu rted ou t, "It's obvious to
anyone who deals with tewardesses."
This was Aesop's call to arms . He asked fo r perm ission to investigate
the truthfulness of this " fact," poi nting out that it w ould take little time
and effort. He was given permission to do so by all bu t the director of
personnel, who said it would be a waste o f time and money. Th e ma-
jority's permission prevai led, and Aesop went ah ad.
At each of the ai rport s that served as a stewardess base, an admin is-
trator used the next mo nth's fl ight schedul e to com bine il ights into
stewardess assignments fo r the month. Th e e "b ids," as they were
Uncontrolled Variables 83
called, were typed on cards that were posted on a bulletin board. The
most senior stewardess at the base had first choice. The next most
senior stewardess had second choice, and so on. Thus it was possible
to obtain records of the assignments and the choices made at each
base. Using these data Aesop and h is colleagues plotte·d the flying
hours of the assignments se lected against the seniority of the stew-
ardess (see Figure 4.1).
If the directo r of personnel had been right, the pl o t would have
followed a curve something like A, with the more sen ior girls selecting
the assignments with the sma llest number of flyin g hours. If he had
been right I would not be telling this story. The data actually followed
a curve like B. This was a surprise even to Aesop ; he had expected to
find no relationship. Therefore, he thought it desirable to find an ex-
planation of what had been observed. An analysis of the differences
between longer and shorter assignme nts revea led that the schedules
with more flying hours provided greater reg ularity of days oH and more
off-duty time at the stewardess' s home base. Aesop's team then revi sed
their scheduling procedure to provide complete regularity of days off
and to increase th e amount of free time that cou ld be spen t in the base
city. This was done with only a small reduction of the average flying
hours obtained by the first assignment procedure. The resu lts were
much better than those yielded by the procedure then in use .
.r:::
c
0
E
Least Most
Seniority
Figure 4 .1 .
84 The Art
Aesop prese nted the revised proced ure and th resu lts to both the
stewardesses and the managers. Both enthusiastically acce pted it, in-
cluding the contrite director of personnel. Tne procedure was imple-
mented successful ly .
.'.koRAL:
~;1···
There is nothing so deceptive as an ap-
parent trulh .
*
A self-evident "fact" may not only prevent our seeeing an effective
course of action that is available, but, as in this fable, it may preclude
the consideration of a course of action that is known to be available.
The more obvious a statement appears to be, the more thorough ly its
truthfulness should be investi ga ted. Unfo rtunate ly, the more c,bvious a
"fact" appears to be, the mor r sis tance th re usually is to efforts to
test it, as the next fable shows.
A large publi c util ity, most of whos operations are located in major
cities, asked Aesop to work on the following probl em. The p roportion
of its workers who \•vere Black had recently passed forty percent and
was increasing rapidly. The company expected that before th e end of
the 1970s a majority of its employees would b Black. On th other
hand, less than on e percen t fi ts supervis ory and man ageri al personnel
were Black, and this p r entage w as not increasin g. Th erefore. so me its
managers saw the com pany b coming what they c lied a "corpo rate
plantation ."
The d irec tor of personnel wa nted to increase the number of Blacks
in supervision and management. To this end he ask d Aesop to de ign
an educationa l program for the company's Black employees that would
increase their upward mobility.
Aesop bega n by ask ing for acce to the personnel records of five
hundred rand omly selected Bla ck employees and the same number of
White employees drawn from those in the same types o f job as the
Blacks held . The person nel d irector xp lained that this w ould be very
difficult be ause ach employee's records we re in th r e pa rts in three
different locations; they would be very di ffi cult to co llect and collate.
Aesop explained that he could not design an educational program for
Black employees without knowing in detail th eir education al deficien-
cies, and he knew o f no othe r feasible w ay of ob tain ing this info rma-
tion . The director of person n l to ld Aesop that he could safely a ume
Uncontrolled Variables 85
Fable 4.2. The black problem is white. Moral: Facts are easily colored by
one's perspective.
the Blacks were deficient in every relevant way. This was an assumption
Aesop was not willing to make, and , he explained , it had to be put into
quantitative terms anyhow. Therefore, he pressed for access to the rec-
ords. It was refu sed . Aeso p th en exercised his contractual right to dis-
continue th e work.
Word of thi s d ispute reached o ne of the parent company's execu-
tives, who was outraged. Becau se of press ure he brought to bear in the
right places, Aesop was invited ba ck and the records were made avail-
86 The Art
able to him. When he analyzed them, he found that Black female em-
ployees had an average of one and a half years more equivalent educa-
tion than Wh ite female employees in the same types of job. Black
ma les had an average of three-quarters of a year more equivalent edu-
cation than White males in the same types of job .
On the bas is o f these find ings, Aesop told the company that its
problem was not one of educating Black employees for upward mobil-
ity, but of educating White supervisors and managers so they would
permit Blacks to move up.
Aesop's suggestion was summarily rejected , and Aesop once again
term ina ted his relationsh ip with the company, th is time permanently.
More than a year later a federal court found the company guilty of
discriminatory practices against Black emp loyees who had brought suit
against the company.
The chief executive o f a company that produced dairy products and de-
livered them t the hom e by the familiar but vanish ing milkman found
the company's costs of deli very rising rapidly. As he increased prices
to cover these costs, the company became less competitive with retail
stores, particularly supermarkets. Fewer and fewer households were
willing to pay as much as was required for the convenience of home
delivery. ales d ecreased , and delivery costs escalated further.
It was no secret that most bus inesses that depended on regular
home delivery were in similar straits. This suggested two possible
solutions to the problem. One in volved the cooperation of competing
compani es in formin g a kind of "United Home Delivery Service" so
lhatone mil kman cou ld deliver many brands, but to customers cl ose to
each other. This would decrease delivery costs considerably, but the
company rejected this suggestion because it considered its drivers to
be salesm en whom it wanted to control.
The second alternative was accepted. It involved extending the
product line to incl ude fru it drinks , baked goods, and a selection of
staple fo ods so tha t the average size of a delivery could be increased.
Each of these products taken separately could not make it with home
delivery, but co llecti vely they could and d id .
Fable 4.5. The upside-down story. Moral : It's a good idea to know which
side one's door is bulled on.
others w ith too many left-hand doors. Thi s led to both lost sales and
exce sive inve ntories.
Aesop was called in to develop better forecasting o f the sales of the
two versions . It immed iatel y occu rred to him that if a refrigerator could
be opened from ei ther side, th e problem w ould disappear. He was
quickly informed that another man ufacturer had bu ilt such a door,
hinged in the center of its top and bo tto m, and it had flopped . It was
expensive, and ambidexterity had li ttl e or no val ue to the custome r.
Uncontrolled Variables 93
Nevertheless, Aesop felt the idea was worth pursuing. One day he
noted that if the door of the refrigerator made by the company was
turned upside down, its outer side looked exactly the same except that
the handle was on the other side. He learned that this was deliberate;
right- and left-hand doors were the same except for their hinging. Fur-
thermore, the door and the shelves mounted into the back of the door
were so designed that the shelves could also be inserted upside down.
This led Aesop to design a refrigerator door frame that had hinge re-
ceptacles on both sides. Those on the side not in use could be covered
with a snap-on plate. The hinges themselves were designed to snap
into the receptacles. This meant that the same door could be mounted
on either side; furthermore, it could be changed at any time before or
aher purchase. The additional cost of production was very small com-
pared lo the cost of additional inventory and lost sales resulting from
the traditiona l design. Aesop's design also offered the customer con-
vertibility, should it be required by a move or remodelling.
This solution was developed and eve ntu all y implemented in several
cities. On re fl ection it becomes apparent that th is type of solut ion is
already in use in subway sys tems.
A symptom may appea r in any part of a system even thou gh its
source may be in any ot her part or in the ir interactions. We do not
treat headach es by performing brain surgery .
We may fail to solve a problem not only because we take as irrele-
vant or uncontrollab le a variable tha t we can ac tu al ly control, but also
because we ass ume others cannot control one o r more variables that
they actually can. Here is an examp le.
In the 1950s one urban un iversity, like many others , had a se rious park-
ing problem on its campus . There w as not enough space on its parking
lots to accommodate all the cars driven to the campus by faculty, stu -
dents, and service personnel. The un iversi ty was firmly comm itted to
providing free parking because of the lack of adequate public trans-
portation .
Many latecomers parked " ill egal ly" in the aisles o f the parking lots
rather than park off campus and walk a considerable distance. This pre -
vented others from leavin g the campus when they wanted to . Com-
plaints mounted in number and intensity, forcing a respon se from the
university's admini strators. A fter cons idering sever a I a I tern ative sys-
tems, they selected one and ha d it insta ll ed . It required the regi stration
of cars with on-campus parking rights . Since there w as not enough
space for all , a p riority system wa s used .
96 The Art
Fable 4.7. When you can't go forward, back up. Moral: It is often harder
lo solve a problem created by others than ii is for them lo solve the
problem created by lhe solution.
The way a person who has a problem thinks about it is usually quite
different from the way that those who cause the problem think about
it. For this reason military systems designers have long used what are
called " countermeasure groups. " These are teams of highly competent
researchers who are asked to play the " enemy, " but, unlike the enemy,
they are given all the information they want. Their objective is to find
out how to beat the system . When they find a way to do so, th ey reveal
it to the system designers, who then alter their design to counter the
countermeasure. Then the countermeasure group goes at it again . This
process is continued until a design is produced that will accomplish its
mission despite the efforts of the countermeasure group. Although this
procedure does not guarantee success with a "real enemy," i t greatly
increases chances for it.
If the park ing lot problem described in the last fable had been given
98 The Art
to a student countermeasure group, it wou ld have revealed how the
system could be beaten before its ins tallation rather than afterward ·.
The behavior of others is often the least yield ing type o f uncon-
trolled variable, but this is so because we usually do not provide them
with proper incenti ves. We already considered (in Chapter 3) how in-
centives can and should be brought under contro l. They are a poten-
tially powerful means of bringing the uncontrol led beh avior of others
under control, even when the others a,e autonomous. The next fab le
illus trates this point.
This fable has a number of po ints, but the one of central interest here
is that it invo lves bringing an uncontrolled "external " variable-cus-
tomer purchasing behavior- under partial control by the use of an in-
ternal ly generated incentive for customers.
The ability to use incentives to modify the behavior of oth ers de-
rends critical ly on how well th ei r behav ior is understood . In genera l,
we do not und ers tand the ir behavior as w ell as we th in k we do . This
often precl ud es creative prob l m solving. A case in poin t is developed
in Chapter 11. It is rather compl ex, bu t it ill us tra tes well both the diffi-
100 TheArt
culty of gaining understand ing of human behavior and its value once
obtained .
CONCLUSION
The various morals and points made in this chapter can be summarized
in three genera l observations.
First, in every problematic situation there is a set of relevant facts oi
the case. Some of th ese usually appear to be obvious. The more obvi-
ous such ;i fact appears to be, the more intensely its truth should be
investigated. We are more likely to be wrong in what we accept with-
out evidence, no matter how obvious it may be, than in what we accept
with evidence, no matter how doubtfu l it may be . When an obvious
fact .of the case is topp led, it almost inva riably opens opportunities for
finding more creative and effective solutio ns to the prob lem in hand. It
does so by converting what appeared to be an uncontrolled variable
into one that is pe rceived to be controllab le.
Second, uncontrolled variables often both create the problem con-
fronted and constrain the actions that the decision maker can take to
solve it. It is often possible to remove the problem or the constraints by
enlarging the system taken to be relevant. This can convert uncon-
troll ed into controlled variables. One uncontrolled variab le can often
be used to cancel the harmful effect of another. This is the principle
behind the vertical and horizontal integ ra tion of organizations.
Third, many uncontrolled variables are not intrinsically uncontrolla-
ble. lack of control over them does not lie in the nature of the variable
but in our lack of know ledge and understanding of it. Therefore, re-
search directed at producing relevant knowledge and understanding
can often bring an uncontrolled variab le under control. Part of the art
of proble m so lving lies in knowing when and how to use such research.
CHAPTER FIVE
Relations
The way that a cou rse of action affects the outcome of a problematic
situation depends on how the relevant variables are interrelated and
how they relate to the outcome. The bel ief that a variable is signifi-
cantly ielated to the outcome of what we do is what makes us consider
it relevant. Therefore, the selection of the variables that we attempt to
manipulate and the way we attempt to manipulate them are dete r-
mined by what we believe to be the nature of these relationships .
Causality is the most important type of relationship involved in
problem so lving. We seek to do things that will bring about a desired
state. To "bring about" omething is to cause it in either a stro ng
sense- that is, to do something that completely determines the ou t-
come-or a weak sense-to do something that may or may not affect or
influence the outcome. A weak (probabilistic or nondeterministic) cause
is sometimes referred to as a producer and its effect as a product. A
producer (for examp le, an acorn) ha s some probability of producing its
product (an oa k) but is not certain to do so. In most " real" situations
we deal with weak causal relat ionships. Strong causality usually re-
quires laboratory or laboratory- like conditions.
Our ability to solve problems de pends cri tically on how well we
conceptualize the causal connections betwf>en what w e do and what
we want. Many of our problem -solving failures derive from assuming a
causal connection where it does not exist or incorrect ly characterizing
acausal connection where it does exist. Perhaps the most common sin-
gle cause of failure in problem solving derives from incorrectly assum-
ing a causal relationship between variables that have only been dem -
102 The Art
onstrated to be associated. Variables that tend to change togethe r, in
the same or opposite directions, are associated .
For example, the weight and height of persons are clearly asso-
ciated. This means that if we obtain the weight and height of each of
a sample of peop le and plot them on a graph, the points tend to in-
crease together. Such an association is said to be positive . If an increase
in one variable is associated with a decrease in the other-fo r example,
income and illness- the association is said to be negative.
When two variab les are associated, either can be used to predict the
valu e of th e other. For example, we can use the association of we ight
and height to p redict one if we know the othe r. However, we cannot
infer from their association that a change in weight will produce a
change in height.
I have used weight and height because the error of causal inference
is so apparent. Th is type of error is not always so apparent, and is there-
fore made over and over again . For example, suppose we p lot a com-
pany's annual sales for a number of years against its annua l advertising
expenditures over the same years. Such plots often show a positive
association between these two variables: sales and advertising expendi-
tures tend to increase together. It does not fol low, however, that an
increase in advertising will produce an increase in sa les . Advertising
may have an effect on sales, as shown in Chapter 10, but even if it does,
a positive association between sales and advertising does not prove it.
Payroll and sales are usua lly positively associated; neverthe less it would
be incorrect to infer that an increase in payroll wil l produce an increase
in sales.
If there is no association between two variables, we can justifiably
infer that they are not causally related under the conditions in which
the observations are made, but not under ot her conditions. However,
if there is an association, we should not infe r that they are causally con-
nected, that either is the cause of the other. The most we can infer is
that they may be so related . Therefore, we can use association to filter
a large number of variab les to determine which should be studied fur-
ther for causal relations.
The most commonly used measures of association are correlation
and regression coefficients. The nature of these measures is not impor-
tant here, but the cautious problem solver wi ll remember that they are
measures of association and therefore cannot legitimately be used for
inferring causality. Caution is required because there are a number of
widely used sophisticated statistica l techniques such as factor and clus-
ter analysis in which the fact that they are based on corre lation or re-
gression is very well concealed, often even from those who use them.
We can see th e danger of inferring causa lity from association by
Relations 103
returning to the advertising example . Suppose we plotted annual sales
against advertising expenditures in the following year. In most cases in
which I have done just this, I obtained a stronger positive association
than that obtained when the same year was used . Clearly, we cannot
infer that an increase in next year's advertising will increase this year's
sales. In fact, if there is a causal connection between these variables it
may well run in the opposite direction .
Many companies set their advertising budgets as a relatively fixed
percentage of forecasted sales. Most sales- forecasting procedures are
Sl!ch that if there is an increase in sales one year, a further increase is
likely to be forecast for the following year. Now we can see the causa l
sequence· if sales went up last year the company probably forecast an
increase in sales this year. If it forecast such an increase, it probably in -
creased advertising expenditures. Therefore, what this shows is that an
increase in sales one year tends to produce an increase in advertising in
the following year.
One cannot overestimate the frequency with which the erroneous
inference of causality from association is made. Here are two examples
of such errors made in problem areas that are currently considered to
be very critica I.
Early in the war against cancer the medical profession's battle against
smoking began. Numerous studies were published showing that smok -
ing and lung cancer were positively associated . This could not bE: con-
tradicted, but the inference drawn from such stud ies- that smoking
causes cancer-could be. Again , smoking may be a cause of lung can -
cer, but their correl~tion is not an adequate basis for asserting that it is.
One study published in a prominent medical journal showed a
strong positive correlation between per capita consumption of tobacco
and the incidence of lung cancer over a number of countries. A causal
connection was incorrectly inferred. To show that this was the case,
Aesop used the same data on per capita consumption of tobacco for
the same countries but substituted the incidence rate of cholera. He
obtained a negative correlation that was stronger than the positive cor-
relation revealed in the article. Using the same logic as that which ap-
peared in the original article, Aesop prepared another article almost
identical to the original except for the conclusion; he concluded that
smoking prevents cholera. He submitted this article to the same medi-
cal journal in which the original article had appeared . It was rejected
because, according to the referees, it was facetious. Aesop wrote to the
104 The Art
editor admitting that he had been facetious, but then, was this not true
of the original article? Why, he asked, had it been pub lished ? He re-
ceived no rep ly.
types of dieta ry d fici ency indu ce tuber ulosis. Therefo re, those who
live in high soot fa ll areas are more likely to get tuberculosis than
others, but soot fa ll is not th e cause.
Few peopl e, if any, are imm une to inferring guilt by associa tion. In the
fable that follows my own lack of im munity is appa rent.
106 The Art
Fable 5.3. SUGAR IS SWEETER BUT PROTEIN
IS NEATER.
Fable 5.4. The fill-up stations that were not full up. Moral : The less we
understand something, the more variables we require to explain it.
Si nce there are four ways to enter and four ways to leave an unre-
stricted intersection, there are sixteen routes through it, including go ing
back in the d irecti on one came from, for example, after getting gas.
Aesop selected a sample of 200 stations and counted the number of
cars passing each stat ion in each o f the sixteen ro utes and the num ber
in each ro ute that stopped for service. The data thus obtained revealed
that four routes yielded most of the customers, for example, the ro ute
that involved a righ t-hand turn arou nd the sta tion and the one tha t was
straight through with the station on the car side at the far corner of the
intersection.
110 The Art
It is widely assumed that the maximum price people are willing to pay
for a product is affected by the value they impute to it, and there is
plenty of evidence to support this assumption. What is not so apparent
112 The Art
is that the price of a product also affects the value it is believed to have.
Price and the perception of valu e are not independent variables; the
interact. The discovery of !he nature of this interaction provided an in-
genious solution to a marketing problem involving a food product.
The principal product of a food company was the highest priced in
its field, but it en joyed the largest share of its market because it wai
perceived as a quality product that was "worth the difference." As was
to be expected, competitors were devoted to toppling the product
from its pinnacle. They tried to do so by decreasing their prices. A!
first thi s had no effect, but with continued price reductions by com,
petitors the lead ing brand began to lose its share of the market. Its
producer called in a research group to help find a pricing policy that
would protect the product's share of th e market with as little loss of
profit as possible.
The researchers learned that the price of their sponsor' s product had
remained constant over many years during which competitive prices
had come down. Only after there was a considerable price difference
had there been any significant impact on the sales of the sponsor's
product. Therefore, the researchers decided to find out how and why
the changes in competitive pricing affected the sales of the produd
they were trying to protect. They looked for a relationship between
"how much more the lead ing product costs than competitive prod-
ucts" and " how much more value consumers bel ieve it to have." The-;
focused their research on the consumers' perceptions of prices and
product quality and how these percep tions were related.
First, th e researchers took a variety of product types in which there
was a large spread of prices among products of the same type. They
conducted surveys, asking consumers at what price they believed each
brand was current ly sold. When the results were plotted, they revealed
a curve such as the one shown in Figure 5.1. The prices of products
with the lowest prices were underestimated ; those with the highest
prices were overestimated, and those in the middle were correct\)
estimated .
Nex t, th e team too k classes of products the quality of which could
be measured objectively. They measured their quality and then asked
consumers to est ima te these qual ities. The results yielded a curve such
as the one shown in Fi gure 5.2. The estimates of the quality of producti
·with low quality were high , estimates of those with high qual ity were
low, and estimates of those in between were accurate .
From these two sets of observations, the team concluded that the
highest priced product in a class would generally be perceived to be
higher priced and of lower quality than was actually the case. As the
Relations 113
Actual price
Figure 5.1.
spread of prices between the highest p riced product and the next high-
est priced increased, the most expens ive appeared to increase in rel a-
tive cost and decrease in relative quality.
A reduction of the price of the sponsor's product would have re-
duced its loss of share of market or might poss ibly have increased it,
but this wo uld have requ ired an actua l reductio n o f its qual ity if its
profitability were to be ma intained . The company was reluctant to re-
duce its product's quality.
Because the resea rch team und erstood the un derly ing relati onships
in this problematic situation, it saw an alternative course of actio n th at
oth erwise would have been missed. It recommended that the co m pany
bring out a significa ntly higher priced pro duct o f even higher qu ali ty
than its current product. The te m argued that th i wou ld co rrec t the
co nsumers' perception of the leading brand' s price and quali ty and
thu enable it to recla im its previous share of the marke t. The ompany
fol lowed th i advice, and it worked .
Many effects are the result of interacting causes. We know, for ex-
ample, that water boi ls at a certai n temperatu re but that the boil ing
11 4 The Art
Figure 5.2.
point varies w ith atmos ph eric pressure. For this reason it takes longer
to prepare a hard-bo iled egg in M exico Ci ty th an in New York .
Like so many o bvio us th ings, the intera tion of causa l va riables i
often overl oo ked. We not o nly respo nd to variab les in ombination,
but, even when we respo nd to on ly o ne, w e often have a ch oice as to
w hich to respon d to, as the next sto ry shows.
In a conv rsation abou t logical thinking am on g four professors, one
asked the o thers: If two chimney sweeps cl imbed down a chi mney and
one came out wi th a di rty face and th e other w ith a clean face, which
one woul d go to wash his face?
The en gi nee r in the group immediately answered, " Th e fe ll ow wi th
the d irty face."
The scien tist in the grou p d isagreed. He sa id, "The fel low with the
cl ean face. He wou ld see that the other fe llow 's face w as di rty and
wo uld assume his w as di rty also, so he w ould go to wash hi face. But
the fe llow wi th the di rty face w oul d see that th e other's w s clean and
wo uld assume that his was clean also. So h would not go to wash hi!
face."
Th e phi losopher in the group came to the defense of the engineer.
He argued, " W hen the fellow with the clean face started for the wash-
room, as the scientist co rrec tly reason ed, the fellow w ith the dirty face
Relations 115
would ask h im wh ere he was going. The fellow with the clean face
would say, 'To wash my face.' 'Why?' the other would ask, 'your face is
clean.' Th e fellow with the clean fa ce wou ld then say, 'But yours is
dirty.' So the fe llow with th e dirty face w ou ld go to wash his face."
A studen t w ho listened to this exchange could co ntain himself no
longer and burst ou t with the question: "How co uld two fellows climb
down a d irty chi mney an d one come o ut with a clean face and the
other with a dirty face?''
One can infer different effects from the same cause. Furthermore,
one can move from the same cause to the same effect by different
routes that are not equall y effi cie nt. This is il lustrated by the story of a
physics professor who asked members of an introductory class how
they wou ld dete rmine the height of a tall building using a barometer.
One eager scie nce student blu rted out that he would measure the
atmospheric p ressure at ground level and at the top of the building,
and convert the difference into the building's height using a we ll-
known physical equ ation.
A second science student said he would do it differently. He would
drop the barometer from the top o f th e b uild in g and measure the
length of time it took to reach the ground. Then he would convert th is
time into the building's height usi ng an equally we ll-known phys ical
equation .
The professor then asked a.shy business administration student how
he wou ld do it. The student reluctantly repli ed that he would give the
baromet r to the bu ilding's h ief caretaker and ask h im how high the
build ing was.
It is apparen t th at not all effects are linearly related to their causes ;
that is, the amo unt of response is proporti on al to the amount of stimu-
lus. Few stimu li have responses that are related in this way. Some of a
drug may be beneficial, but increases in quantity may not be accom-
panied by increases in ben fit. In fac t, as the dosage increases, harmful
or letha l effects are possible.
Increases in advertisi ng or sales calls on customers may, after a cer-
tain level (the saturation point) , fail to yield any add itional response
from customers. Beyond a still higher level (the supersaturation point)
they may prod uce decreased respon ses. A st ud y show ing that this was
the case in on e company' s advert isi ng is described in deta il in Chap-
ter 10.
Now consider the type of re lat io nsh ip th at holds betwee n the symp-
tom s of a deficiency and its cause. Symptom are part of what was
previou sly referred to (in Chap ter 4) as fa ct s o f the case .
When a patient consults a doctor, he begins by giving him ce rtain
116 The Art
facts of his case : symptoms of il l healt h. The n th e doctor usua lly makes
some observations of his own; he ta kes the patient's temperature.
pulse, blood pressure, and so on, and asks ad d itiona l questions such
as : " Does th is hurt?" or "Have you been sleepi ng well?" In this war
the doctor fi nds additional symptoms that are also facts of the case.
Once a set of symptoms has been col lected, the doctor goes through
a diagnostic process . In this process he develops a hypothesis or a con,
clusion about the cause of the deficiency whose symptoms he has ob-
served . Once he is convinced of the va lidity of the diagnosis- that is,
the cause of the illness-he prescribes in a way he beli eves w ill remove
the cause . If the cause is removed , the symptoms should disappear.
Unfortunately, it does not always work this way . For examp le, some
patients who complain of stomach pains are eventually diagnosed as
having duodena l ulcers. One way of treating such ulcers is by surgery
that disconnects the duodenum from the stomach. In many cases after
such an operation the pat ient develops a peptic ulcer. Thus the "cause"
of an ulcer that was eli minated wa s not ult ima te enough. !n some such
cases, a doctor presc ribes medica tion that eli mi nates the patient's
stomach pains, the reby removing the symptoms bu t not the cause. A
duodenal ulce r can itself be considered to be a symptom of a psycho,
somatic disorder.
We have a paralle l case in the way alcoholism is treated . It is usual~
treated as a problem to be removed rather than as the alcoho lic's so/u·
tion to a serious personal problem . 1 o wonder that when his "solu-
tion" is den ied him he often returns to it as soon as he can or develops
other " solutions" that are equal ly damaging. The alcohol ic's original
problem is usually left un solved when he is denied alcohol or is in·
duced to abstain .
Problem solving is often directed at the removal of symptoms rather
than cau ses . This is particularly true where society is involved . For ex·
ample, we define crime as a social problem, and we attempt to solve
it by removing the crim ina l from society . However, crime continues to
rise, and mounting evidence shows that prisons make convicts more
likely to commit further crimes when they are released . Thus it is in-
creasingly apparen t that cr ime is a symptom, not a di sease, and that we
have not been treati ng its causes adequate ly . In fact , the treatment i1
w idely recog nized as a major cause of the continuing problem .
We tend to look for simple causes of even the most complex prob-
lems . This derives to some extent from the success that modern medi·
ci ne had in its early days with what ha s come to be known as the
"ge rm theory" of illness . Medical researchers looked for some foreign
" bug" that was completely responsible for an ailment and, having
Relations 117
found it, tried to rem ove or su p p ress it in so me way. Th e su ccess of
this approach bred a disposition toward simpl e ca uses not only in
medicine but in any field in which the subj ect of study could be
thought of as an organism or organism-like.
For example, we consi der d rug addi ction to be a disease ca used by
drugs. We usually t reat it by denying addicts access to the respons ible
drugs. That we have not succeeded w ith such treatment is apparent,
but our beli ef in this simple-m inded methodology is strong enough to
overcome evidence tha t su ggests that anoth er approa ch is bette r.
A number of studies have shown th at add icts are characteri stically
disenchanted with and aliena ted from society. They hold so ci ety re-
sponsibl e for the ir p roble ms and seek to "thu mb the ir noses at it." By
making drugs illegal we increase the ir att ractiveness to addicts and
others who are ali enated from society precisely becau se the use of
something ill ega l is a way o f th umb ing on e' s nose at the soci ety that
proscribes its use. Hence our treatm ent exacerbates th e probl em; it
does not solve it. By legal izing the use of addic tive dru gs an d di spens-
ing them in a cont rol led w ay, En gland has had m uch more success than
America in treat ing addic tion, and it has avo ided the huge crimina l
problem created by the demand for drugs that are illegal.
Most of us thin k we are im mu ne to such errors, but th is is not the
case. Much of what we thi nk is problem so lving is sym ptom supp res-
sion, and many of the problems we face are the products of solutions
we have app lied to p revi ou s problems.
For exampl e, in systems involving large am ou nts of simi lar eq uip-
ment, regular maintenance is often used to prevent b reakdown s. In
many such cases ma intenance is th e princi pal cause of subsequ ent
breakdown s. Th is is no t unfam il ia r to the o w ners of automobiles who
have frequent ly been victimized by th e w hi te-fro cked d iagnostician in
an automotive repai r shop. In one study done for a branch of th e mi li-
tary, it was found th at steam cl eanin g of moto r veh icles wa s respon-
sible not only for many breakd owns but also for the redu ced life of the
veh icles. Another study do ne on rai lroad locomotives sh owed that fre -
quen t changes of thei r motor oi l accelerated the w ear of eng ines. In
th is con nec tion "Too much of a good th ing is bad" is a useful tautol-
ogy to keep in mi nd.
SUMMARY
To help provide an idea of the content of idea lized design and a feel-
ing of th e sp irit behind it a very brief d iges t of such a design that is the
subject of a recent ly published book (Ackoff et al. , 1976) is provided
here. first, some of the background of the design may be helpfu l.
The large comp lex o f interacting individuals and inst i tut ions that are
involved in producing, d is tr ibuti ng, and marketing scient ifi c and tech-
nologica l informa tion is o ften referred to as a " system ." Th is complex,
however, is clearly not an organized system ; it is an agg rega ti on of in-
teracting but independ ent ly contro lled and uncoordinated parts . For
convenience I refer to it as the " SCA TT System ," SCA TT being an acro-
nym derived from Scient ific Commu nicatio n and Techno lo gy Transfer.
Numerous stud ies have been d irected at improving the efficiency or
effectiveness of variou s pa rts of th is sys tem. Their cumu lat ive effect has
not been impressi ve ; th ey have not p roduced si gnificant imp rovements
in the system as a whole . Users stil l comp lain abou t the rap idly increas-
ing overload of information, much of which is ei the r redundant or of
little va lue, and about the difficu l ty and excessive time required to find
and acquire at leas t some of the info rmat ion they want. Autho rs, pub-
lishers, librar ian s, and othe rs invo lved in information serv ices have
their own frequent ly expressed complaints and frustrations . Su ch dis-
satisfaction wi th the curre nt system has led to a large number of
122 Applications
changes in its pa rts, but these have not had a significant overall effed
because they, like the system itself, have neith er been coordinated nor
integrated by the use of any well -defin ed systemi c goals o r ob jectives.
The current system has many vi rtues : it is far from the wors t such
system in the world. However, as alm ost everybody wou ld agree, it is
far from th e bes t that is poss ibl e. To agree th at this is the case is not
necessarily to agree on what i th e best system possible.
The deficiencies of our current SCA n system are exa cerbated by a
situation that D r. Jordan Baruch, a prominent exp ert in the field, de-
scribed thus:
Unlike the fields of phys ics and chem i try, in fo rm ation sci ence
has been d eterred in its develo p men t by th lack of an app ro-
priate environment in which to perfo rm real is tic experiments.
. . . In Information science, co ntro l has been achieved almost al-
ways at the sacrifice of reality while realistic xperiments, by and
large, had to be conducted in a relat ively uncon tro lled environ-
m ent. (From a private comm unication w ith permi sion .)
THE DESIGN
Population Percentage
Year (in millions) over 20
Data on past usage are shown in Tab le 7.3 which also inc ludes data on
ve hic les other than automobiles.
SMSAs
Central Cities 22 9
Fringe 66 75
O utside SMSA 12 16
Total U.S.A. 100 100
1960 9,446
1980 9,759
2000 10,072
The res ults of such calcu la tions for each of the two growth rates of
number of automobil es per adult are shown in Tab le 7.6.
During the past fifteen years app roximately fifty percent of the miles
travelled in a given year was in urban areas, as indicated in Table 7.7.
Using different projection methods, W ilbur Smith and Associates (1 966,
p. 36) esti m ated that by 1980 abo ut sixty percent of the total vehicle
miles travelled will be on urban roads and that by 2000 this will in-
crease to abo ut sixty-five percent. Table 7.7 shows that the historical
percentage dist ribution of urban-rural miles for automobiles and all
vehicles has rema ined practica lly constant at abo ut fifty percent for the
past two decades.
Continuing our policy of being conservative, we assume that urban
vehicle m iles will account for fifty percent of the automobile vehicle
miles in 1980 and 2000 even though the es timates made by Wilbur
Smith and Associates are significant ly higher.
By applying these percentages to the data shown in Table 7.6 we can
estimate unconstrained total urban automobi le mi les per year. The re-
sults are shown in Table 7.8.
Table 7.8. Urban Automobile Miles per Year (in Millions) for
Various Ownership Growth Rates
Low Medium
Percentage Percentage
Year Miles of1960 Miles of1960
Table 7.9. Past and Extrapolated Vehicle Miles · (in Millions) for Buses
and Trucks and Ratios to Automobile Vehicle Miles
Table 7.10. Relation Between Highway Usage and Capacity of the Fed-
eral-Aid Primary System
Growth Rate
Year Low Medium
movements having origin or destina tion within one area. These are the
roads that carry a heavy traffic load and provide access to residenti~
locations, to the central b usiness district, to industrial areas, to periph-
eral business areas, and so on , that is, to the main destinations in urban
areas. When traffic congestion increases in this system, it also increases
on other urban streets and highways not included in the system. For
this rea son a major shift of urban traffic from the Federal -aid primary
system to other urban roads is not expected.
Using the low growth rate of automobile ownership, about 55,00J
(76,690-20,076) additional miles of urban highways will be required in
year 2000 to maintain the 1960 level of congestion . At the very con-
servative estimate of an average cost of $10 m ill ion per mile of standard
four-lane urban highway (Lyle Fitch and Associates, 1964, p. 14), the
total investment required over the next thirty years would be approxi-
mately $550 bill ion, or an average of $18.3 bi II ion per year. This con-
sti tu tes more than a threefold in crease in the total expenditure for
transportation facilities in 1967 (approximately $5 .35 billion) . The
amount required ($18.3 billion) is more than ten limes the amounl
spent on urban highways in 1967 ($1.4 billion). Such an inc rease in
expenditure is vi rt ually impossible, but it is not the only obstacle lo
constrained growth of automobile usage .
The amo unt of land that can be allocated to roads , highways, and
parking spaces in urban areas, particularly in the central business dis-
trict, also limits such growth . Lyle Fitch and Associates (1964) quote
Senator Harrison Williams on th is subject:
In most cities the proportion of land devoted to streets and park ing
in downtown areas already exceeds 40% of the total land available.
Table 7.12 shows the relevant percentages for five metropolitan area s.
Clearly, these percentages cannot be increased 3.82 times. Many
cities, of course, have smaller 11ercentages, and additional highways
(for example, ones bypassing the CBD) need not generate additiona l
parking requirements. Nevertheless it is clea r that the space constraint
would be reached for increases less than tw ice the current allocations.
Proposa ls have been made to use two-level highways to avoid the
space problem . The costs associated with this are much higher than
those for surface-level highways. Thus, if construction of the "required "
conventional road is economica lly in feasib le, as we have demonstrated ,
build ing elevated highways would be even less feasible.
Therefore, it does not appear to be practical to expand the existing
urban road and highway system to cope with the unconstra ined growth
of traffic volume over the next thirty years because of the expenditures
of money and amounts of space such expansion would require .
We have not considered such socia l costs as might arise from in -
creased accident rates, increased air pollut ion, and decreased attrac-
tiveness of the environment. The considerat ion of these costs would
provide additional support to the conclusion that we will not be able
138 Applications
to solve the urban tran spo rtation prob lem by expand ing th e highwa
and road system.
(The study of which the reference projection presented above ii
part went on to con sider other changes in the city that might reduct
the projected requirement for highways. It showed that these cannOI
be exp ected to sign ifica nt ly affec t th e requirements. It also showed that
it is unlikely that mass transit and highway and veh icle technology nO'tl
under development will significantly reduce th is requirement. The studv
did find several possib le direction s in which possible solutions migh'
be found. I include only one of these here. It is sufficient to show how
reference projections can be used to revea l both the nature of a future
p roblem and creati ve so lutions to it.)
There are no signs that increasing re lian ce on the automobile will bf
affected by increas ing traffic congestion o r the inconveniences derived
from the widespread use of the au tomob ile. On the contra ry, thereti
some evidence that urbanites wou ld prefer to relocate their jobs or
residen ces ra ther than switch from the automobile to another modeoi
transporta tion. The Department of Housing and Urban Transportatior.
(1968b) noted that :
The present design of the automobile, the five to six passenger fam-
il y car, is a compromise intended to satisfy a wide variety of nee<h
Automobiles are used for inter- and intracity trave l, to and from worl:
recreation, shopping, and so o n.
Of considerable importance is the fact that the number of two-a
families increased from seven percent in 1950 to twenty -five percentir
1966. Th is and other facts we consider la ter suggest a functional differ,
entiation between an intra- and intercity automobile. Families that 011·
or use more than one car wou ld obtain distinct advantages by usint
special-purpose automobiles better suited for specific needs, for exam-
ple, cars better suited to the characteristics of center-city traffic.
A major improvement in automotive systems is suggested by the fr
ures on autom ob ile occupancy in urban areas. For example, avera1
occupancy rates in metropo litan Philadelphia are approximately F
passe ngers per car, ranging from 1.2 for commuting trips to 1.6 fo:
nonwork trips (Penn -Jersey Transportation Study, 1964, p. 91). Tilt
average capacity of an automobile, on the other hand, is about fill
Transportation without a Future 139
people. It is apparent that substantial reduction in automobile conges-
tion cou ld be obtained if the average occupancy of automobiles, par-
ticularly for work trips, were increased. Ca r pooling, however, reduces
the advantages o f door-to-d oor travel by automobile. A less incon -
veniencing altern ative would involve the use of small urban autom o-
biles, what have been referred to by some as "urmobiles."
This alternative has been explored in several studies (for example,
Cars for Cities and Department of Housing and Urban Development,
1968b). It is generally acknowledged, however, that at higher speeds
and in free -flowi ng traffic the effect of reduced vehicle length on con-
gestion is very small. For example,.at forty miles per hou r the majority
of the road space can be said to be occupied by safety space between
vehicles, and, according to McClenehan and Simkowitz (1969) the ef-
fect of reducing ca r length by half on expressway traffic would be an
increase in flow of no more than ten to fifteen percen t. Greater in-
creases would occu r on heavi ly used city streets; as much as a seventy
percent increase in flow would be achived when congestion reached
the not-uncommon level of fifteen vehicles per light. If only a fraction
of long cars is replaced by small ones, the resulting flow is approxi-
mately a linear interpolation between the two extremes.
Relatively little is known about the effect of car w idth on traffic flow.
Experiments carried out by the Ministry of Tra nsport in England (Cars
for Cities, p. 13) showed that a lane width two and one-half to three
feet wider than the car itself represents a reasonable m inimum for
safety purposes. They also showed that in mixed traffic condit io ns,
when small and large cars travel together, the small ones usually travel
behind the larger ones using the same road space.
A considerable increase in passenger den sity co uld be ob tained by
the use of short (less than ten feet) and narrow (three and one-ha lf feet)
two -passen ger vehicles, one passenger seated behind the other. If traf-
fic were made up exclusively of such vehi cles, an increase of at least
2.2 (2.0 X 1.1•) on expressways (two vehicl es per normal lane and a
ten percent increase in flow due to sho rter ca rs) and 3.4 (2.0 X 1.7•) on
city streets cou ld be obtained. If shou lde rs of four- lane expressways
were used for one lane of such vehicles, their capacity would in crease
by 2.7 (2.2 + 0.5), and the additional lane wo uld be used for one nar-
row car which would have ha lf the width of a no rmal car. Following a
similar line of reasoning, city streets with two moving lan es for current
automobiles and one for parking, the increase would be 5.1 (3.4 +
1.7).
These ca lculations do not take trucks and buses into account, but they
• Using McClenehan and Simkowitz figures for the increase in traffic flow with ,hort
cars.
140 Applications
show it is possible to deal even with the 4.28 medium forecasted irr
crease in requ i rements for the year 2000 if the smal l car we have de-
scribed w ere generally adopted .
Parking req ui rements wou ld also be greatly reduced . For example
a normal car ta kes mo re than twenty feet along the sidewa lks. Three
ten-foot -long cars cou ld be pa rked in th e linear space required for two
norma l cars, and add itional road space would be left free for vehicular
tra ffic. Park ing space requirements could be further reduced if the door
or doo rs w ere either slid ing on the side or placed at the front or back.
The latter wou ld permit face-in parking w ith very high density.
A la rge var iety of smal l automobiles is under development no11
(Mechanics Illustrated, October 1969, p. 76 and Life, December 11,
1970) . In man y cases the new des igns incorporate changes that wi ll re-
duce the polluting effec ts of cars, and, because of the ir decreased
weight and lower speeds, they woul d greatly reduce fuel con sumption
They ca n be made to hook on to each other in tra in-like fashion to
facil itate towing o r tak ing the fa mily along on a trip (using nonmo-
tor ized cabs, for example). Their reduced maximum speed and accel-
eration capab ili ties increase their safety.
The advanta ges of usi ng smal l cars for intrac ity traffic depend on the
restrict ions imposed on the use of large vehicl es. Durin g a transition~
period, vehi cles of different sizes can mi x together. Eventu ally the use
of city streets and highways co uld be lim ited to sma ll cars from, say
7:00 a.m. to 7:00 p.m. on week d ays. In som e cities trucks are al reacfy
kept off CBD streets during these hours.
M any benefits coul d de ri ve from publ icly or privately own ed fleets
of small cars tha t wo uld be ava ilable as drive-it-yo urse lf taxis. Their
p ick- up and drop-off points co uld be widely dispersed over the ci~
(see M inicar Tran sit System) .
It is clear that a change to small urban automobil es ca n be accom-
p lished in at least a decade. Furthermore, such a change would require
little public cost an d w ould yield economies to the ind ividua l withoul
the loss of con ven ien ce o r comfo rt. Most important, it cou ld reduce
congestion significantly and pe rm it less restricted use of automobiles
than w ould otherwise be possible.
CHAPTER EIGHT
A Multidimensional
Organizational Structure ·
Organ izations are purposefu l systems some of whose parts are pur-
posefu l individuals among whom there is a functional division of labor.
Their purposefulness entails cho ice of ends (desired outcomes : objec-
tives and goa ls) and means (courses of action). Courses of action neces-
sarily involve the use of resources (inputs) to produce goods or services
(outputs) that are intended to be of greater value to those who con-
sume them than are th e inputs. The resources consumed include man-
power, materials and energy, plant and equipment, and/or money. This
is as true of organizations tha t are nonprofit as for those that are for
profit, and as true for those that are publicly owned as for those that
are privately owned .
Trad itiona lly, organizat ional structure is taken to deal with two
types of re lationship: (1) responsibility, who is responsible for what
and (2) authority, who reports to whom . Structure so conceived lends
itsel f to rep resentation by a two-dimensional tree in which boxes rep-
resent responsibi li ties and altitude and lines represent the loci and flow
of au thority.
The conventio nal representati on of an orga nization's structure does
not show either wh at inputs flow into what outputs or what means are
used for what ends . An input-output matrix is seldom used to describe
an organization's structure; a means-ends matrix virtua lly never i~
However, these matrices can revea l new and more flexible ways of
structuring organizations. In developing th is po int I use for il lustration
a typica l privately owned corporation engaged in producing goods.
O the r types of organization , public and private, are used in subsequent
i II us tra ti o ns .
A Multidimensional Organizational Structure 143
An organizat ion's outputs can be used to define its ends, for exam-
ple, outputs classified by product typ e or brands. An organizational
unit whose responsibi lity is to provide a product or service to con-
sumers outside the organization is refer red to as a program , and such
units are represented by P1, P2, . . , Pk.
Activities are the means employed by programs. They can usually
be divided into operations, activities that d irectly affect the nature or
availability of the company·s output, and services, activities that are re -
quired to support programs or operations. Typica l operations (01 , 0 2,
... , Om) are raw material purchasing, transportation , prod uction, dis-
tribution, and marketing. Typical services (51 , 52, . . . , Sn) are pro-
vided by accounting, data processing, maintenance, in dustrial relatio ns,
finance, personnel, and legal units.
A precise and completely general distinction between o pe rations
and services is difficult to make, but doing so is not essential to the
concept of organ izational structure developed here .
The way activities feed into prog rams and into activit ies themselves
can be displayed in such tables as those shown in Figures 8.1 and 8.2 .
The output of each activity may be consumed by programs and other
activities, additionally by the executive function (discussed below), by
Programs
P, P1 ... p.
o,
~
C o,
-~.;
~
a.
0 ..
~
om
·!
:~
u
c( s,
it:
s,
u
..
·~
Jl
s.
Figure 8.1.
144 Applications
Act ivities as consumers
Opera1 ions Services
o,
"'C 01
t'.' 0
·~
~ ..
.!!'
Q.
0. o,,,
~
"''"
"'
.!!' s,
S2
"'u"'
·~
"'
Ill
..
s.
Figure 8.2.
- - 0,
C:
-
--
.. 0
02
. ·-..
C ·-
·--0
..
-
- C
.. "C
8- 0
Ofl
- - om
- - s,
C:
-
-
0
w ·;;.
S2
..
(.J"'
~
-~ .':'
-
.. "C
(/) 0
0
(.J
- --- Sn
Figure 8.3.
Now cons ider how a mult idimensional (MD) organ izatio n can be
operated to maxim ize its flexibili ty an d responsi veness to ch anging in-
ternal and ex terna l cond ition s.
PROGRAMS
1. Program s sho uld be fo rmula ted so that they require more than
one so u rce of inp ut. Th ese sources may be intern al or external. A
program that requires on ly one such so urce should be combined
with another to meet this condition . As wi ll be seen, th e more ac-
tiviti es a p rog ram requ ires, the more flex ibility it can be managed .
The co rresponding requirement sho uld be imposed on activities .
They sh ould be designed to be useful by more than one pro-
gram and, in the case of services, by more than one line activity.
2. The ob jective of each pro gram should be defined so that the
benefits that can be derived from it can be measu red. Fu rther, be-
146 Applications
cau se its costs can be measured, a specified funct io n of its bene-
fi ts and costs, such as p rofi t, ca n and shoul d be used to meas ure
its performance.
3. Pro grams should be provi ded with ma nagers and their personal
staffs, but no o ther person nel. Th ey should not be organ izational
un its in the usual sense. They sh ou ld be managements whose
function is to specify, acqui re, coord ina te, and control suc h input
activit ies of oth ers as are required to attain their objectives. They
should have no authority over those who supply these services,
b ut they shou ld be able to co ntrol them in a way revealed in para-
graph 6.
4. Programs should not be p rovided with p lant o r equ ipment.
They should involve no investm ent .
5. Programs require and sh ould be provided with ope rating cap i-
tal. Th ey sho uld rece ive at least as much of the income they gen-
erate as required to cover their operat ing cos ts. If they do not
generate sufficien t cap ital for this p urpose, th ey can either bor-
row money to be paid back subsequen tly or be subs id ized . They
shou ld be subsidized only wh en the ir actua l or expected benefits
exceed th eir costs and thei r costs exceed their income.
6. Programs can buy goods and services from eit her internal ac-
ti vity uni ts or external supp liers as th e y see fit. It is through their
purchasing power tha t they contro l and co ordinate thei r sup pl iers.
For practical reasons it may be necessary to co nstrain this fr eedom
of ch oice, for example, because propri etary information is in-
volved . The fo rmulation or technology of a product, or its pro-
duction processes, may be " classifi ed" information ; hen ce the
company may be unwilli ng to have it p roduced externa lly. Wher-
ever such a con st raint is imposed on a program by the executive
fu nc tion, th e saving, if any, that could be obtai ned by b reaking
the constraint should be determined. This sh ould be pai d to the
program by the execut ive function . If the con straint is self-
imposed by the progra m, the re shou ld be no su ch paymen t It is
possib le, o f cou rse, tha t th ere is no su itabl e extern al source, leav-
ing th e pro gram no cho ice .
ACTIVITY UNITS
Figure 8.4.
150 Applications
for programs or acti vities. They should serve as representatives (advo-
ca tes) of those w ho are affected by the outputs and activities of the
organizatio n as a whole. They serve external stakeho lders. They can
evaluate programs and activities in each region from the point of view
of those outside th e organi zati on who are affected by them . These
evalua tion s can be fed back to the organization's executive, coordina-
tors, an d uni t directo rs.
By cal ling toget her the regiona l rep resentat ives, any program direc-
tor can obtain an overall picture of his p rogram's performance across
the area it se rves and in each reg ion. This en ables him to al locate his
re~ources across regions more effectively.
Geography is not the only criterion th at can be used to classify ex-
te rna l stakeholders. Many other criteria can be used . For example, a
co mpany th at supplies a va riety of different industries-say, with lubri-
cants-may find it useful to have advocates for industries rather than
regions. They might be des ig na ted as automot ive, aerospace, machine-
tool, and so on. A public service orga nizat ion may define its advocates'
responsibility by the socioecono mi c characteristics of the rec ipients of
the services .
SOME VARIATIONS
The MD orga nizational design presented here is, as previou sly noted, a
theme around w hich many variations can be written, variations that are
respon sive to the unique characterist ics of an organization's mission or
environ ment. Only two such variations are descr ibed here.
A recently prepared design of the national hea lth and welfare serv-
ices in Iran invo lves a three-dimensional organizationa l structure (pro-
grams, operations, and services) for ea ch of the reg ions into which the
country is divided. The sma ller regions are not provided with all the
operations and services they require, but they can buy these from
larger neigh boring regions. The region al executives report directly to
the minister. This des ign, original ly sti mulated by a desire to decen·
trali ze management an d be mo re responsive to vary ing local needs,
also provides the flexibility required in a rapid ly changing society.
Another variant is to be found in th e organizat ion of the Industrial
Management Institute of Teh ran wh ich designed the hea lth and wel-
fare system descri bed above. The ir)stitute is divided into two parts,
each a two-dimensio nal matrix of programs and operation s. They share
common services (see Figu re 8.5) . Operations can also be used as a
common thi rd dim ension to any number-not necessarily two- of two-
dimensiona l matrices.
A Multidimensional Organizational Structure 151
Figure 8.5.
SOME EXAMPLES
A Government Agency
1. In formation services
2. Legal services
3. Accou nting and con tro l
4. Research and development
5. Personnel services
6. Purchasing (supplies and services), personnel, transportation, and
mai ntena nce
7. Public re la tions
8. Construct ion an d p roduction of fac ili ties and equ ipment
9. General se rvices
Universities
ON DIVIDED RESPONSIBILITY
CONCLUSION
1. 1 1.2 1.3
,------
1
I
I
r - - - - - - - +-I
I
L.J"" -
Board
--,
1
,-
I Board
- -,
J
I r~- __J
I
I
I
L
~
~
1.1
I
I
I
I
I
I r - - -1 r - - -, , - - -,
I
r
I
I
1-r1 r Tl 3
I Board
...
I
I I Soard
'"
I
I I Board
L _ _ _J
I
I
In this case study, the rela tionship between the amount spent on ad·
vertising, its timing, media usage, and sales of a consumer product,
beer, is expl ored experimentally in some depth . The case revea ls how
research can be used to il lum inate such relat ionships and how complex
and counteri ntuitive they m ay be. Th e simple re lat iona l assumpt ions
that go in to ma ny advertis ing decisions can be very costly and reduce
their effectiveness significantly.
Just before mid -1961 , Mr. August A. Busch , Jr., then Pres ident and
Chairman of th e Board of Anheuser-Busch , Inc. , asked my co ll eagues
and I if we could evaluate an advert ising decisio n he was about to
make. In th at year Budw ei se r, the l argest se lli ng beer in the United
States, was budgeted to receive about $15 m illion worth o f advertising.
Mr. Busch had been approached by the vice p res ident of marketing
w i th a reques t for an additiona l $1, 200,000 to be spent on advertising
in twe lve mark eting areas. The v ice president had defended his pro-
posal on the basis o f a pro j ected increase in sa les that he bel ieved it
would produce. Mr. Busch exp lained that he was confronted w ith such
a proposa l every year and that he always acce pted it. He intended to
do the same again, but, he asked, " Is there som e way I ca n fi nd out at
the end of the year whether I got what I pa id for? '' We said we would
think about it and come back with some suggestions.
The proposal we presented to Mr. Bus ch shortly thereafter cons isted
• Thi s case is adapt ed from Ack off and Emshoff (W inter, 1975) .
The Effect of Advertising on Sales 163
of allowing the marketing d epartmen t to sele ct any six of the twelve
areas it wa nt d and giving it $600,000 fo r add ition al advertising. The
remaining six areas w ould not be tou ched and w ould be used as " con-
trols." This bi ased select ion proced ure was intended to overcome
some of the oppos iti on that the marketi ng de par tment had to any effort
to evaluate its p roposal.
Earlier we had developed an eq uation for forecast ing monthly sal es
in each marketing area. Our plan was to mea sure the deviation of
actual monthly sal es from the forecast for each marketing area in the
test. Using the statistical chara cte ris tics o f the forecasts, we es timated
that we had a ninety-five percent chance of detecting a four percent
increase in sales in the areas with add iti on al advertising. Since the in-
crease pred icted by the marketing depart ment wa s in excess of this
amount, M r. Busch authorized th e test, an d it was initiated.
The test was conducted over the last six months of 1961 yie ld in g
seventy-two (12 X 6) obse rvations. The analysis of th ese data failed to
reveal any significan t d iffere nce betwe en th e tes t and control areas .
Nevertheless the co ntrol area s did better on average than was forecast.
Therefore, we assumed that al l the sales above those forecast were at-
tributab le to the increased ad vertisi ng and eva lu ated the results accord-
ingly. Even under this ass umption th e increased amount of advert ising
was not ju stified by the deliberately overestim ated increase in sal es at -
tributed to it.
Encouraged by these results, Mr. Bu sch asked us to design research
directed at determ ini ng w hat amoun t sho uld b spent on adve rt isi ng.
He wanted to p roceed wi th caution, however, because he believed that
much of Budwe ise r' succ ss-i t w as leading the bee r market with a
share of 8.1 4 percen t in 1962-w as due to its qual ity and the effect ive-
ness with whi ch this was com mun icated through its advertising. When
we suggested research invo lving xperimenta tion w ith marketi ng areas,
he author ized us of fi ftee n such areas pro vided they did not include
any of th e company' s ma jor markets.
Constra ined in th is way, we so ught an exper imental design that
would maxi m ize learn ing abo ut adve rtis ing expe nditures. Our design
effort was guided by two methodological princip les. First, we knew
that the compa ny advertised fo r on ly on e reason : to incr ease sales.
Therefore, we were de term ined to meas ure the effect of advertising on
sales and no t, as is usually done, on one or mo re easily measured in-
tervening vari bles such as reca ll of m s ages or attitudes toward the
product. Fo r th is reaso n w e decided to con tinue to use deviations of
actual from forecast sales as th e variable to be observed. This all owed
us to ca ncel out much of th effect on sal e o f factors other than ad-
164 Applications
vertising. The refore, efforts to improve forecasting of month ly market-
ing-area sa les were co ntinuous.
Second, we we re comm itted to an attempt to explain the causal
effect of advertising on consumer purchases, not mere ly to find statisti·
cal co rrelations between them. Our search of the marketing '.iterature
for su ch an expl anat ion w as futile; all it uncove red were correlations
and regressions (associations) between advert ising and sales. These
usually showed that increases (or decreases) in the former were asso-
ciated with increases (or decreases) in the latter. From such associations
it was almost universa lly inferr~ d, incorrectly, that increases in advertis-
ing produce increases in sales almost w ith out limit. We believed that
these ana lyses really showed tha t most companies forecast next year's
sales qu ite accurately and that they set their advert isi ng budgets as a
fixed percentage of p red icted sales. Put another way: forecasts of in-
creased sales p roduce increased advertising.
Our commitmen t to experimentation derived from a determination
to find a ca usal connection between advertising and sa les, not merely
an association between them : to develop an abi lity to manipulate ad-
vertising to produce the desired effects on sales that could be observed
Since we had no tested theo ry to go on, we fabricated our own. OIH
hunch was that adve rtising could be considered to be a stimulus and
sales a response to it. A great deal is known about the general nature
of sti mulus- response fun ction s. They usually take the form shown in
Figure 10.1. Therefore, _w e formulated the following hypothesis:
In an earlier study we had d one for the Lamp Division of the Gen,
eral Electric Company (Waid, Clar k, and Ackoff, 1956) w e found such a
relationship between the frequ ency of sales calls (stimuli) and pur-
chases (responses). In the sales -call context, the idea of supersaturation
is not as shocking as it is in advert isi ng. Cl early there is an amountof
a salesman's presence that is into lerable to a buyer. Beyond this, one
would expect the buyer to try to get rid of the salesman by discontinu,
ing his purchases. Similarly, we fe lt reasonably sure that i f, for example,
The Effect of Advertising on Sales 165
Su persatu ra tion
- - ---p~oint
:ll
C
..Sl-
0
0:
St imul us
Figure 10.1. A typical stimulus-response function.
all television advertising were for one product, the public wou ld react
negative ly.
A minima l experiment would have involved applying the same per-
centage change in advertising expenditures to each of the fifteen mar-
keting areas allotted to us an d comparing the results obtained from
them with those obtained from an equ al number of control (un-
changed ) areas. However, we needed only nine areas to obtain the
level of accuracy set as our target : to be ab le to detect a two percent
difference in sa les ninety-five percent of the time . To introduce two
different treatments-one involving an increase and the other a de-
crease in advertising expenditures-req uired eighteen test areas, three
more than were ava il able. However, even an experiment w ith two dif-
ferent treatments wou ld yield only th ree points : the average effect of
each treatment and that of the control group. The difficulty this pre-
sented derived from the fact that every con fi gu ration of three points
except one, V-shaped, could be fi tted to the re lationship we wanted to
test. Therefore, there was a very low probab ility that a three-leve l ex-
periment would disconfirm our hypothetical relationsh ip; thus it was a
very poor test of the valid ity of th is re lationship.
For these reaso ns we decided to ask for three different treatments
and a contro l group, even though this wou ld requ ire twenty-seven
marketing areas plus n ine under control. Four experimental po ints
could disconfirm our theory as easily as it cou ld confirm it and hence
would have provided a reasonab le test of it.
We had noth ing to go on b ut o ur intuition in select ing the experi-
mental treatment leve ls: a fif ty percent reduction and a fifty and 100
166 Applications
percent increase in budgeted levels of advertis ing. We w anted to make
changes large enou gh to produce observab le effects on sa les, assumin1
such changes had any effect; thus, if there w ere no observable effects,
this fac t cou ld no t be di smissed because the changes were believed to
be too small. Two increases rath er th an decreases were selected lo
ma ke the experimen t more pa latable to the market ing depa rtment.
Wh en this four-level design was presented , it was rejected because
it invo lved the use of too many market ing areas. Mr. Busch agreed,
however, to o ur use of eighteen (rather th an fi fteen) areas provided
that we changed the reduction in adve rt isin g from fifty to twenty-fi't1!
percent. He felt tha t a fifty percen t reduction might ir reparably dama~
the areas so treated . This lef t us with a three-level experiment: minus
twenty-five, zero, an d fif ty percent changes from budget.
We were not complete ly happy wi th this outcome, because it d'~
not provide an adequate test o f our theory, but we were pleased th.at
w e had the opportuni ty to co nduct even a limited experiment. We
were reasona bly sure that if it produced "inte resting" resu lts, restric·
tio ns on futu re ex peri ments would be lifted .
A 3 X 3 X 3 fac to riall y design ed experiment wa s prepared in which
two other important marketing variables were exp licitly controlled: the
amount spen t on sales effort (on sa lesmen) and lhe amount spent OIi
point-of-sa fes ma terials (displays, si gns, etc.) (see Figure 10.2). We
w oul d also have liked to con tro l p ricing, bu t this wa s p recluded.
M arket in g areas w ere selected random ly from the " permissable lisl"
and randoml y assigned to the twenty-seven treatments . The use of lh~
list cou ld obviousl y b ias our res ults, but aga in our hope was that the
results w oul d j usti fy fu rther experiments and that they wou ld not beso
restri cted.
The experim ent was carried o ut over twe lve months, yield ing twelve
o bservations of each ma rketing area. We were ab le to reach a conclu-
sion at the end of six mon ths, but th e experiment was continued lo
build up con fi dence in the res ults. This did not work, however, because
the resu lts were too much at variance with expectations within the
com pa ny and its advertis ing agency. Th e three points shown in Figure
10.3 fel l into the onl y configu ration, V-shap ed, that was inconsistent
wi th o ur hypothesi s becau se the relatio nship being tested had no Vin
it. In add ition, w e found no sign ifica nt interaction between advertisin~
sa les effort, an d poin t-of- sales expenditures- a surprising but not un-
accep tabl e res ult- and tha t cu rrent levels of sales effort and point-of.
sa les expend itures w ere close to optimal. The la st resu lt was readily
accepted .
No one foun d m uch difficulty in believing that a fifty percent in·
Figure 10.2. A 3 X 3 X 3 factorially designed experiment.
15
i
;: •
.!:
10
8,
C
2
•
.,"
!!, 5
..~
Cl.
0
•
- 50 0 +SO
Percen1 change in advertising level
.,.,,/
,--.....',_ ,' / ,.--
Total popul ation
...... ,
'
Stimulus
~ 20
.s 15
1j,
C
~ 10
"'
"'
!!! 5
.,C
e
~ 0
"'""
.,
~
~
- 100 - 50 0 +50 +100 +150 +200
Percen tage change in advertising
II Ill IV
Spring X X 0 X
Summer X 0 X X
Fall X X 0 0
Winter X 0 X 0
No national TV 5 5 5 5
National TV 5 5 5 5
Social. These are o f either sex but are generally younger than
reparative drinkers . They have not ye t atta ined the leve ls of aspiration,
but believe they wi ll and that doing so requ ires approval and support
of others . They are dr iven by ambition, the desire to get ahead .
Socia l dr ink ing is heaviest on weekends, holidays, and vacat ions. It
usua lly takes place in large groups consisting of acquaintances in social
settings. The social is a control led drin ker but less so than the repara -
tive. His or her drinking is associated with friendl iness and acceptance
of and by othe rs . Alcoholic beverages are seen as lubricants of socia l
situation s.
Indulgent. These are of any age and either sex. They have not at -
tained the levels of asp iration and never expect to . They consider th em-
selves to be irretrievable fa il u res. They view life as tragic.
The indu lgent drinks most heavily when subj ected to pressure to
achieve. He or she drinks to escape such pressure. The indulgent is the
least controlled drinker and is the most li ke ly to get high or drunk and
become an alcoholic.
1.0 . - - - - - - - - - - - - - - - - - ------,,/
/
/
/
/
/
/
/
/
/
;_p<:- /
.~re /,o
e,'-'/ ,,;
o"O ~~.....
·~/
/. ;;'<>~
/ G:J
/
/
/
/
/
/
/
/
0 I/
0 1.0
l nrens i tv of stimulus
-~
..
Q.
/
/
'<,+
'iii
:i
/
E
:,
/
u /
/
/
/
/
oV
0 1. 0
In tensny of effect
Figure 11.2. The environmental effect function.
1.0
C
.g
.; Objective
Obj ect ive
-~ externa lizer
u. internalizer
(extrovert)
..
~
C
E
"E
0
C
e
·;:
/ Centraversio n
.
C
0
ti
~ C
w 0
..
-~
-~
Subjective
internali zer
Subjective
externalizer
~ (i ntrovert!
"E
:,
u,
0
0 1.0
Internalizat ion Externali,at ion
Effect on Rnv ironm ent
C
Q
"'
! Aeparative Ocean ic
...u
C: E
.. 0
E
C
e
..
·;
C
a
.~ C
0
"' ·e..
~
Indu l gent Social
u
...
'E
~
""'
0
0 1.0
Internal ization Externa li zat ,on
Efleet on environment
Figure 11.4. The matching hypothesis.
Reparative 30 5 7 1
Social 4 31 1 3
Oceanic 0 1 10 0
Indulgent 1 0 1 5
Having established this link between the drinking types and the per-
sonality types, we began a more intens ive research program to estab-
lish the causal links between our th eory and the purposes served by
the consumption of alcohol among the types . Before we cou ld seek
such an explanation, however, we had to develop and va lidate instru-
ments for measuring the dimensions of our personality types . Although
the clinical interviewing procedures had be en demonstrated to have
sufficient reliability, their cost was too great to employ them on larger
samples drawn from mult iple locations. Each o f the 125 interviews cost
approximately $300, and these were condu cted in 1968. Even if we
were to cut these costs in half, they would have been too la rge to en-
able us to deal with large samples in differen t locations . Consequently,
we turned our resea rch to the development of less expensive typing
procedures . One direction we took involves the use of a behavioral
laboratory in which we hoped to be able to reduce or el iminate clini -
cal interviewing costs by collecting data on an individual's personality
by observing his behavior in specially designed situat ions. The other
direction we took involved the design of self-adm inistered paper-and -
pencil tests that a subject could take without supervision .
The behavioral situat ion we developed worked wel l; it provided
rich and reliab le readings on both person ality scales, but the cost and
the time required were too great, particularl y when we had to operate
in different cities . Nevertheless we found that the information obtained
from this laboratory situation enabled us to design a substantially more
efficient and shortened interviewing p rocedure to be used by the cl ini-
cians. Eventually we w ere able to obta in an accurate and re liable cl ini-
cal typing of a subject for about $35.
The development of an accept able self-adm inistered paper-and -
pencil test took more than two years. Two cr iteria of acceptabi lity were
imposed on this effort First, there had to be at least a seventy-five per-
184 Applications
cent agreement betwee n the res ults of the test and clinica l eva luations.
Second, there was to be no bias in the class ification of those subjects
about whom there was disagreement between test and clinic. Th~
would assure us of acceptab le "col lective" accuracy when we went
into the fie ld and used la rge numbe rs of subjects. Fortunately, A-B's
management understood the methodological issues and did not press
us into going in to the field prematurely.
Once an acceptable test was developed, research could be directed
to testi ng the hypo theses fo rmula ted to explain drinking behavior. The
hypothesis fo rmul uted to do so was suggested by the earlier hypothes~
about the two types of alcohol ics and by the fact tha t most alcoholia
came from th e pure rather than the mixed types . In addi tion, it made
use of the fol lowi ng interm edia te " maturation hypothesis":
• These results showed that in the short run only drinke rs cou ld be indu ced to perceive
differences that were not the re. The ex periment d id no t show that drin kers co ul d be
induced not to perceive differences that were there. Bo th Emery and we found distinc-
tive taste preferences in each personality type. This meant that if a beer we re to appeal
to a particula r type it had to have certain physi cal characteristics and not have others.
186 Applications
..
-~
u
.,
E
0
..
.?
u
"'::,
E
"'
did in fact fall into these pe rceived clusters. The personalities asso-
ciated with each brand not surprisingly corresponded to those of the
people usually shown in the ir commercials. The personality segments
found to be associated wi th four major brands of beer are shown in
Figure 11.5.
Using this knowledge, it was possible to determine by further sur-
veys what personality types drank each of the three brands A-B pro-
duces: Mi chelob, Budweiser, and Busch. Althou gh these were found to
have some overlap, in th e main they were found to appeal to differ·
ent personality segments of the market. This enabled us to determine
which segments each brand was and was not reaching . From these
surveys we were also able to estimate what portion of the beer con·
sumed was consumed by each type. These portions varied significantly.
Using this info rm ation, we were then able to specify target market seg-
ments to be reached either by exi sti ng brands or by new ones, and now
we knew what kind of advertising messages would be most effective in
doing so.
For example, although it was found that Michelob had most of its
market in one part of the personality space, a smal l portion of its con-
sumers were drawn from a different part of this space, a part in which
the largest percentage of beer drinkers fell. This led to an advertising
campaign directed at the second segment. The campaign succeeded in
significantly increasing th is brand's share of the target market.
Why People Drink 167
The implica tions of these results to the p reparation of advertising
messages and the design of products is clear, but the typology an d the
theory based on it could be used in another less apparent way. We hy-
pothesized d ifferent usage of different media by different persona lity
types. For example, we pred icted that reparative drinkers (objective in -
ternalizers) wou ld watch more television than socia l drinkers (subjective
externalizers), because reparatives are more likely to observe others and
socials are more likely to be doing things, particip~ting. The chances
that a reparative will watch at least six hours of television on a non -
work day, for example, are more than two and a half times greater than
for a social. This ratio is even higher for indulgents, as one would ex-
pect. ft is also possible to predict which magazine appeals to each type.
For example, Playboy is almost twice as likely to be read by an oceanic
as by an indulgen t; the reverse is true for Readers Digest. Objective
types (oceanics and reparatives) read newspapers more regularly and
thoroughly than subjectives.
By combining information on the segmentation of the market with
the information gained abou t the use of media, it became possible to
combine messages and media in such a way as to direct advertising
messages at particular market segments in a more effective way.
Our typology and personality theory has enab led A-B to gain in-
sights into marketing phenomena tha t do not involve advertising. For
example, research was undertaken for the company to determine what
happens in a market when a new competitor enters it. In particular, the
company was interested in identifying the persona lity characteristics of
those who are most likely to try a new brand when it is introduced,
those who switch to it as their regular brand, and the way others subse-
quently learn about the brand . Research was conducted in a number of
markets that had experienced new brand entries in the re latively recent
past. This work revealed that different personality types have signifi-
cantly different likel ihoods of trying new products. One of the types
purchases new products thirty percent faster than the overall average.
A second type does so slightly more than average, a third type slightly
less, and the fourth type forty percent less than average. Furthermore,
we found that those who are identified as early triers are influenced by
advertising differently from those who switch to the product after it
becomes established . The understanding thus gained enabled Anheuser-
Busch to develop more effective advertising and other marketing strat-
egies at appropriate times before, d uring , and after the introduction of
new products.
Knowledge is power, and understanding is contro l ; they are double-
edged swords. To obtain knowledge and understanding of human be-
havior is to gain the ability to control the behavior of others more
188 Applications
effectively either in one's own or their interests . Like any instrument,
knowledge an d understanding can be used for either good o r evil-for
exam ple, an axe can be used to free a person from a burning building
or to murder him. The use of every instru ment necessari ly involves
ethica l and moral j udgm ents w hich canno t be avoided by the manager
who uses the knowl ed ge and understanding produced by research or
by the resea rcher who produces them.
The understanding of dri nki ng behavi or developed in the research
desc ribed here can be used to either in tensify or amel iorate the " drink-
ing problem ." Fo rtunate ly, Anh euser-Bu sch believes it is in the long-
run interests of th e producers of alcoho lic beverages as well as of their
consumers to ameliorate it. One of the principa l ways the com pany is
using th is understanding is in the development of more ffective ways
of preventing and t reat:ng al coholism.
CHAPTE R TWELVE
Few p rob lems, once solved, stay that way. Changing conditions tend to
unsolve problems that previously have been solved.
A number of years ago a group of us at th e Un ive rsity had worked
long and hard to determ ine whether a major oil company should lease
tankers or build its own . We studied the curre nt and future availab il ity
of oi l and leasable tankers, the costs o f building, operating, and lea sing
tankers, and the future demand for oil. From this we concluded that,
because of a virtu ally certai n oversupply of leasable tankers, it would
be less expensive to lease them than build new ones . The company ac-
cepted our recommendations and began their imp lementation.
A short whi le later th e Egyptian-Israeli co nfli ct brok e out, and the
Suez Canal wa s closed. This immediate ly created a world--wide shortage
of tankers because they had to circumnavigate Africa when bringing
oil from the Midd le East to Europe or America , and this lengthened
the ir trips considerably.
The clo si ng of th e Su ez Canal was a consp icuou s ch ange of circum-
stances that made it immediately apparent to the company's dec is ion
makers that the solut ion they had accepted an d started 10 implement
was no longer a good on e. It wa s cha nged quick ly, adapting to the
new se t of environmenta l condit ions.
The degradat ion of so lutions is seldom as apparent or sudden as it
was in th is case, but it is common. Thus th e w ise problem so lver con -
stantly mon itors solutions to past problem s to be sure they are meeting
his expectations . If they are not, he fi nd s out why and modifies them .
The problems th at manage to stay solved often create new prob-
190 Applications
lems . For example, one who has difficulty getting to work may solve
th is problem by purchasing an automobil e, but the possession of a car
creates alt sorts of new problems such as obtaining insurance, mainte-
nan ce, finding parking places, and so on . Therefore, th e wise problem
solver not only monitors previous solutions, but he also keeps alert and
constant ly scans hi s hori zon to identify new problems that have arisen
or will a rise unless preventive action is taken .
Because p rob lems do not stay solved and their solutions create new
problems, a problem-solving system requires more than an ability to
solve problems ; it also req uires the abil ity to maintain or control solu-
tions that have been imp lemented and an ability to ident ify problems
when or before they arise. In addition to th ese fun ctions, a problem-
solving system sho uld be able to provide the informat ion that the per-
formance of these functions requires.
Problem solving requires a system, because the three primary func-
tions-solving problems, controlling solutions, and identifying and an-
ticipating prob lems-together with the supporti ng information function,
are very interdependent. Therefore, the more effectively thei r interac-
tion s are designed and o rganized, the more effective problem solving
ca n be.
The prob lem-solving system of large organizations may involve a
number of differen t groups, each responsible for a different function or
subfunction, but even in the case of an individua l problem solver each
of the functions shoul d be present.
A great deal has been learned about how these fu nctions can be
organized and carried out. Here I try to provide a summary of this
know ledge. First I consider the ove ra 11 design of a problem -solving sys-
tem, then I take up the individual functions tha t are necessary parts of it.
Figure 12.1 provides a diagra mmatic representation of the system I
describe. The num be rs on the fl ows in the diagram and the letters in
the boxes help tie the description and the diagram toge ther.
As we have seen, every problem invo lves one or more controlled
variables and one or more that are uncontrolled . The set of uncon-
tro ll ed variables constitutes the relevant environmenl (A) and the set
of contro ll ed variables is th e individ ual o r system, effective manipula-
tion of which is the objective of problem solving . I ca ll this the prob-
lem-object (B) or simply ob;ect. The object of an individual 's problem-
solving efforts may, o f course, be or include himself ; for example, he
may not feel wel l or be dissati sfied with his appearance.
To know that one has a prob lem or to know the nature of a problem
that one has requires information about th e prob lem -object and its en-
vironment. Therefore, observation of this object and its environment
16) Decision Record : Expectations. assumptions. infofmat ion, and process u5ed
Environment (A l
(41 (11
Inqu ire Data
~ (1 11 Performance
131 Inqu iries ind icators
(21 In formation lnforma1 ion
(Bl Assump t ion- re levant in formation
17) Instructions on informat ion requ ired subwstem ICI
and actual performance
Jones
Fable 12.1. The expectations that drowned in a pool. Moral: Contrary to
what is commonly bel ieved, memory can improve with age.
Memory erases mistakes and thus prevents our learn ing from them .
Since learning is an importa nt objective of the solut ion-con trol sub-
system, an inactive memory is required for the storage of decision-
records.
Th e decision-record shou ld be used to inform th e information sub-
system what information is needed to monitor the solution (7). The
assumptions should be check ed perio dica lly to see if they still hold,
and the actual effects of the solution (8) should be observed and
brought together with the decision-record. The actual and assumed
conditions and the actual and expected effects sho uld be compared
(E).
If nothing more than this were done, significant imp rovemen ts in
decision making are possible. This was d rama tica lly shown in an ex-
periment conducted in one of our major corporations by one of its
management scientists. He had developed a set of equations tha t were
able to reproduce accurately the past performance of his company
given the executive decisions that had been made and the environmen-
tal conditions that had pertained. He then put these equations in to a
computer and set up the following experimental game.
Using managers in the company, he organized teams to act as the
corporation's executive office. These teams were told that the objective
of the game was to see how much better they could do than the corpo-
ration had actually done. Each team was required to make executive
decisions for forty consecutive quarters . Once they did so for a particu-
196 Applications
lar quarter, their decis ions were fed into the computer, and they re-
ceived in return a profit-and- loss statement and a balance sheet. Their
quarterly performances were plotted sequentially. Every team 's per·
fo rmance improved over time, as expected.
Then a second set of an equal number of teams was set up. These
teams had exact ly th e same assignment as the teams in the first set, but
there w as one p(ocedural difference . Once a team in the second set
had ma de its decis ions, it was requ ired to prepare a P&L statement and
a balance sh eet expressing th e expected eHecis of its decisio ns. Its de-
cisions were then fe d into the computer wh ich provided its output.
These teams were then able to compare their explicitly formulated ex-
pectations with "actual performance."
The average rates of improvement of the first and second sets of
teams were then compa red. The second set learned significantly more
rapidly than the first. This showed the va lue to decision makers of
doing nothing more th an explicitly formulating expectations and com-
paring them with actual performance. Th e problem-so lving system
bein g designed he re, however, goes much further than this.
Where assu med and actual conditions, or expected and actual ef-
fects, agree, no th ing need be done. A ll that needs to be noted are
deviations (9) of actual from assumed conditions and actua l from ex-
pected effects. Such deviati ons indicate that someth ing has gone
wrong. To determine wh at has and what should be done about it re-
quires diagnosis (F) .
The diagnostic function is to find the causes of deviations and to
prescribe corrective action. Although the causes may be difficult to
identify, th ey are only of four types.
P = f(C, U1)
in which
My student and colleague Elsa Vergara has recently reviewed the extensive
literature on creativity. Out of her reading she selected eight books that she
considers to be the most useful to one in terested in the subject. Here they are
with her annotations.
Adams, James, Concept ua l Block.busting, San Francisco Book Co., San Fran-
cisco, 1976.
Adams concentrates on what he calls "con ceptual blocks" that are " men-
tal walls which block the prob lem solver from correctly perceiving a
problem or conceiving its solution." Throughout the book he provides a
series of exercises and techniques designed to overcome these blocks:
de Bono, Edward, Lateral Thinking, Harper, New York, 1973.
de Bono introduces the concept of latera l thinking that is concerned with
breaking old patterns of thought and creating new ones . The book pro-
vides a series of exercises that help to develop one's potential for lateral
thinking. It is a pleasant and interesting how-to-do-it book .
Gordon, William, Synectics, Harper and Row, New York, 1962.
Gordon describes the technique used by the Synectic Research Group in
Cambridge, Massachusetts and the experience in group problem solving
in a number of organizations. This technique is usually app lied to techn i-
cal problems and is based on the utilization of metaphor. Emphasis is
placed on problem definition.
Koestler, Arthur, Th e Act of Creation, Dell Publishing Co ., New York, 1973.
An excellent book , full of innovative and challenging ideas . Koest ler su c-
ceeds in synthesiz ing the points of view of different discip lines . He in tro-
duces the concept of bisociation which he defines as " the perceiving of
a si tuation or idea in two self-cons istent but habitually incompatible
frames of reference. " Humor, art, and scien ce se rve as the context of his
discussion . This book is fu l l of enl ightening examples. An extensive bibli-
ography is included.
Osborn, Alex, Your Crea!ive Power, Charles Scribner's Sons, New York, 1949.
This book contains a discussio n o f brains!Orming, a techn ique useful in
group problem solving. It is based on a distinction between the produc-
tion and evaluation of ideas. Osborn provide s a checklist to help in gen-
erating new ideas. It contains questions that can be used when one is
faced with a problem .
206 Some Suggested Readings
Parnes, Sidney J. and Harold F. Hard ing, A Source Book for Crea!ive Thinking,
Charles Scribner's Sons, New York, 1962. (See Taylor below.)
Seissge-Krenke, Inge, Probleme und Ergebnisse der Krea tivitaets forshung, Hans
Hube r, Bern, 1974.
A comprehen si ve review of the literatu re o n creativi ty. It provides an
overview of the resea rch and train ing me thods relevant to the subject. It
includes a descrip tion o f an experiment o n the effects of envi ronmental
factors on the creativity o f chi ldren.
Taylor, Irving A ., Perspectives in Creativity, A ldine, Chicago, 1975.
This and the Parnes book provide a co llection of articles by the best-
known researchers on creativity. They p rovide a good way of obtaining
an overview of the ideas that preva i l in the fi eld.
REFERENCES
Ackoff, Russell L., " Management Misinformation Systems, " Management Sci-
ence, 14 (1967), B-1 47- B-1 56 .
- - , Redesigning the Future, John Wiley & Sons, New York, 1974.
- - - , "Towards Flexible Organizat ions: A Multidimensional Design,"
OMEGA, 5 (1977), 649- 662.
- - -, T. A. Cowa n, P. Davis, M. Elton, J. C. Emery, M . Meditz, and W. Sachs,
D esign in g a National Scientific and Technological Commu nication Sys-
tem : Th e SCA TT Report, University of Pennsylvan ia Press, Philadelphia,
1976.
- - , and J. R. Emsh off, " Adverti si ng Re sea rch at Anheuser-Busch, Inc.
(1963-68)," Sloan Manag em en t Review, 16 (Winter 1975), 1-15.
- - - , "Adve rti si ng Research at Anheuser-B usch, Inc. (1968- 74)," Sloan Man -
agement Review , 16 (Sp ring 1975), 1-16.
Allport, G. W. , and H. S. Odbert, ' 'Trait- Names : A Psycholexical Study, " Psy-
chological Monographs, No. 211, 1936.
Bo ttiny, Walter, "Trends in Automob ile Ownership and Indicators of Satura -
tio n," Highway Resea rch Record, 106 (1 966) .
Bureau of Pub lic Roads , Highway Stati tics, Summary to 7965, Washington ,
D .C., 1967.
Cars for Cities, Report of the Steeri ng Group and Working Group Appointed
by the Ministry of Transport, Her Majesty's Sta ti one ry Office, London ,
1967.
Davis, Stanley M ., and Pau l R. Lawrence, Matrix, Addison-Wesley Publishing
Co., Rea ding, Mass., 1977.
de Bo no, Edward, New Think, Basic Books, New York, 1967.
Department of Ho usi ng and Urban Development, Studies in New Systems of
Evolutionary Urban Transportation, Vol. II, Was hington, D.C., 1968(a) .
- - -, Tomorrow's Transportation : New Systems for th e Urban Future, Wash-
ington , D.C. , 1968(b).
Goggin, W il li am C. , " How the Mu lti dime nsional Stru ctu re Works at Dow
Corning," Harvard Business Review, 5 2 (Jan uary-February 1974), 54-65 .
Ha ll , John R., Jr., and R. L. Ackoff, " A Systems Approach to the Prob lem s of
Solid Waste and Litter," Journal of Environmental Systems, 2 (Decembe r
1972), 351-364 .
Henry, Jules, Culture against Man , Ran dom House, New York, 1963.
Laing, R. D ., Th e Politics of Experience, Bal lantine Books, New York , 1967.
208 References
Landsberg, Hans, L. Flei shman, and J. Fisher, Resources in America's Future,
The Johns Hopkins Press, Baltimore, 1963.
Lansi ng, J. 8., and G. Hendricks, Automobile Owners hip and Res idential Den-
sity, Survey Research Center, University of Mi chi ga n, Ann Arbo r, 1967.
Lyle, C. Fitch and Associates, Urban Transp ortation and Public Policy, Chandler
Publishi ng Co., San Francisco, 1964.
Management and Behavioral Science Center, Planning and Desig n for Juvenile
Justice, U.S. Department of Ju stice, Law Enfo rc ement Assistance Adminis-
tration, Washington, D.C. , August 1972.
McClenehan, J. W., and H. J. Simkowitz, "Th e Effect of Short Cars on Flow and
Speed in Downtown Traffic: A Simulati on Model and Some Results,"
Transportation Science, 3 (1969), 126---139.
M inicar Transit System, Final Report of Phase I, Feasib il ity Study, pre pared by
the University of Pennsy lvan ia for the U.S. Department of Trans portation,
1968.
Nation al Acade my of Sciences, U.S. Tra nsportation Resources, Perform ance
and Problems, Proceedings of the Transpo rtati on Research Conference,
Woods Hole, M ass., 1960.
Penn-Jersey Transpor tation Study, Vol. I : "The State o f the Region ,'' Philadel-
phia, 1964.
Rapoport, Anatol, Figh ts, Games, and Debates, Th e Universi ty of Michigan
Press, Ann Arbor, 1960.
Roles et Fo nctions Futurs de Paris : Ville Internationale, Documentation du
Gouvernement Fran cais, Paris, Pu blication No. 39, 1973.
Sagas ti, Franci sco, and R. L. Ackoff, "Possible and likely Fu tu res of Urban
Tra nsportati on," Socio- Economic Plan ning Science, 5 (1971), 413-428.
Schon, Donald, Beyond the Stable State, Random House, New York, 1971 .
Singer, E. A., Jr., In Search of a Way o f life, Colu mbia Un iversi ty Press, New
York, 1948.
Toffler, Alvin, Future Shock, Bantam Books, New Yo rk, 1971.
Waid, Clark, D. F. Cl ark, and R. L. Ackoff, " Alloca tion o f Sales Effort in the
La mp Divisio n of the General Electric Compa ny," O perations Research 4
(1956) , 629-647.
Wi l bur Smith and Associates, Transp ortation and Parking for Tomorrow's
Ci ties, New Haven, 1966.
Index
ckoff, Ru ssell. L., 26, 121. 126n, 162n. 164, Barin storming, 54
174n, 202, 203 Baruch, Jordan, 122
Activity units , 1471 Bay Bridge, 59
Adams , James, 205 Beaut y, 16
Adaptability, 142 Beer, 162-173, 174-188
Advertising, 101, 102, 162-173, 174-166 Behaviora l science, 203
mess ages, 175, 176 Bell System, 59-60
Advocates, 150 Bever age con tainers, 87-88
Aesthetics , 14, 15, 16, 22 Bierce, Ambrose, 45, 79
Afr ica, 169 Bill boards, 172-173
Airl ine, 61 -64 Blood suga r, 106
Alcoholic beverages, 174, 176. See a/so Bonus, 94
Beer Botti ny, Wah r, 129
Alco ho lism, 1'15 Brea kdowns, 116
Algorithm , 200 BUDWEISER, 162ff
Allpo rt, G.W., 23 Bure au of ?ublic Road s, 130, 132, 134
America, see Un ited States Bureaucracy , 155
American , 66 Burro, 73
Anhe user-Busch, In c., 162-173, 174-168 Busch, August A., Jr., 162, 168, 170, 171
Applia nces. 62, 64-65 Busc h Center, 122
Arbitration, 40 Busses, 93-95
Aristotle, 17
Art, 13 Campesino, 73
Association of variab les, 101 , 110, 117. See C ancer , 102
also Relations Candy, 105
Assumptions, 9, 78, 202 Capi tal punishment, 4Sf
Authority, 142 Carbohydrates, 106
Automation, 199 Ca r pooling, 139
Automobiles, 29, 66, 128-140 Cars for Ci1ies. 139
Causality, 100-120, 164
Banking, 67-68 Central business district, 137
Bargaining, 40 C himney sweeps, 113-114
210 Index
'~ witty, literate and, most of all, convincing reflection ... [Ackoff]
shines an of ten bright light into corners where problems hide, showing
the manager how to understand the consequences of his own behavior;
identify real, rather than supposed, elements of problems; perceive
another's aims; determine what is controllable; and deal with other
nettlesome factors."
-Inc.
Russ Ackoff- author, on ultant, and tea h r xtraordinaire. During his long career,
he has shown thousands of managers, architects, engin ers, attorneys, advertising
people, software developers, and scientists the way to more creative, artful prob-
lem solving. This new paper edition of The Ari of Problem Solving is perhaps the
best example of Ackoff in action . St p by step, this pra tical guide shows you how
to develop an understanding of the art of creative thinking and the design of
creative <;Olut ions.
Using ':i\ckoff's Fa bles"- humorous yet eminently practical parables, ba~ed on real
problems by real managers- you'll see why solving a probl m seldom solves the
prob! m, but why approaching it from a n w, unorthodox angle of ten does. The
result is vintage Ackoff- controversial, funny, and always on target.
If you like to dig beyond simple solutions- to imaginative solutions that work- this
book is for you.
II
John Wiley & Sons, Inc. I SBN 0 - 471-85808-0
Professional, Reference and Trade Group 52295
605 Third Avenue, New York, NY. 10158-0012
New York • Chi he ter • Brisban • Toronto
• Singapore
9 80471 8 8089