Management Misinformation Systems

Author(s): Russell L. Ackoff

Source: Management Science , Dec., 1967, Vol. 14, No. 4, Application Series (Dec., 1967),
pp. B147-B156
Published by: INFORMS

Stable URL: https://www.jstor.org/stable/2628680

JSTOR is a not-for-profit service that helps scholars, researchers, and students discover, use, and build upon a wide
range of content in a trusted digital archive. We use information technology and tools to increase productivity and
facilitate new forms of scholarship. For more information about JSTOR, please contact [email protected].

Your use of the JSTOR archive indicates your acceptance of the Terms & Conditions of Use, available at
https://about.jstor.org/terms

INFORMS is collaborating with JSTOR to digitize, preserve and extend access to Management
Science

This content downloaded from

103.158.42.8 on Tue, 17 Oct 2023 14:16:10 +00:00
All use subject to https://about.jstor.org/terms
 MANAGEMENT SCIENCE
Vol. 14, No. 4, December, 1967
Printed in U.S.A.

MANAGEMENT MISINFORMATION SYSTEMS *

RUSSELL L. ACKOFF

University of Pennsylvania

Five assumptions commonly made by designers of management information

systems are identified. It is argued that these are not justified in many (if not
most) cases and hence lead to major deficiencies in the resulting systems.
These assumptions are: (1) the critical deficiency under which most managers
operate is the lack of relevant information, (2) the manager needs the infor-
mation he wants, (3) if a manager has the information he needs his decision
making will improve, (4) better communication between managers improves
organizational performance, and (5) a manager does not have to understand
how his information system works, only how to use it. To overcome these
assumptions and the deficiencies which result from them, a management
information system should be imbedded in a management control system.
A procedure for designing such a system is proposed and an example is given
of the type of control system which it produces.

The growing preoccupation of operations researchers and management scien-

tists with Management Information Systems (MIS's) is apparent. In fact, for
some the design of such systems has almost become synonymous with operations
research or management science. Enthusiasm for such systems is understand-
able: it involves the researcher in a romantic relationship with the most glamorous
instrument of our time, the computer. Such enthusiasm is understandable but,
nevertheless, some of the excesses to which it has led are not excusable.
Contrary to the impression produced by the growing literature, few com-
puterized management information systems have been put into operation. Of
those I've seen that have been implemented, most have not matched expectations
and some have been outright failures. I believe that these near- and far-misses
could have been avoided if certain false (and usually implicit) assumptions on
which many such systems have been erected had not been made.
There seem to be five common and erroneous assumptions underlying the
design of most MIS's, each of which I will consider. After doing so I will outline
an MIS design procedure which avoids these assumptions.

Give Them More

Most MIS's are designed on the assumption that the critical deficiency under
which most managers operate is the lack of relevant information. I do not deny
that most managers lack a good deal of information that they should have, but
I do deny that this is the most important informational deficiency from which
they suffer. It seems to me that they suffer more from an over abundance of
irrelevant information.

* Received June 1967.

B-147

This content downloaded from

103.158.42.8 on Tue, 17 Oct 2023 14:16:10 +00:00
All use subject to https://about.jstor.org/terms
B-148 RUSSELL L. ACKOFF

This is not a play on words. The consequences of changing the emphasis of an

MIS from supplying relevant information to eliminating irrelevant information
is considerable. If one is preoccupied with supplying relevant information,
attention is almost exclusively given to the generation, storage, and retrieval of
information: hence emphasis is placed on constructing data banks, coding,
indexing, updating files, access languages, and so on. The ideal which has emerged
from this orientation is an infinite pool of data into which a manager can reach to
pull out any information he wants. If, on the other hand, one sees the manager's
information problem primarily, but not exclusively, as one that arises out of an
overabundance of irrelevant information, most of which was not asked for, then
the two most important functions of an information system become filtration
(or evaluation) and condensation. The literature on MIS's seldom refers to these
functions let alone considers how to carry them out.
My experience indicates that most managers receive much more data (if not
information) than they can possibly absorb even if they spend all of their time
trying to do so. Hence they already suffer from an information overload. They
must spend a great deal of time separating the relevant form the irrelevant and
searching for the kernels in the relevant documents. For example, I have found
that I receive an average of forty-three hours of unsolicited reading material
each week. The solicited material is usually half again this amount.
I have seen a daily stock status report that consists of approximately six
hundred pages of computer print-out. The report is circulated daily across man-
agers' desks. I've also seen requests for major capital expenditures that come in
book size, several of which are distributed to managers each week. It is not
uncommon for many managers to receive an average of one journal a day or
more. One could go on and on.
Unless the information overload to which managers are subjected is reduced,
any additional information made available by an MIS cannot be expected to be
used effectively.
Even relevant documents have too much redundancy. Most documents can
be considerably condensed without loss of content. My point here is best made,
perhaps, by describing briefly an experiment that a few of my colleagues and I
conducted on the OR literature several years ago. By using a panel of well-known
experts we identified four OR articles that all members of the panel considered
to be "above average," and four articles that were considered to be "below
average." The authors of the eight articles were asked to prepare "objective"
examinations (duration thirty minutes) plus answers for graduate students who
were to be assigned the articles for reading. (The authors were not informed about
the experiment.) Then several experienced writers were asked to reduce each
article to 2 and 3 of its original length only by eliminating words. They also
prepared a brief abstract of each article. Those who did the condensing did not
see the examinations to be given to the students.
A group of graduate students who had not previously read the articles were
then selected. Each one was given four articles randomly selected, each of which

This content downloaded from

103.158.42.8 on Tue, 17 Oct 2023 14:16:10 +00:00
All use subject to https://about.jstor.org/terms
 MANAGEMENT MISINFORMATION SYSTEMS B-149

was in one of its four versions: 100%, 67%, 33 %, or abstract. Each version of
each article was read by two students. All were given the same examinations.
The average scores on the examinations were then compared.
For the above-average articles there was no significant difference between
average test scores for the 100%, 67%, and 33 % versions, but there was a
significant decrease in average test scores for those who had read only the
abstract. For the below-average articles there was no difference in average test
scores among those who had read the 100 %, 67 %, and 33 % versions, but there
was a significant increase in average test scores of those who had read only the
abstract.
The sample used was obviously too small for general conclusions but the
results strongly indicate the extent to which even good writing can be condensed
without loss of information. I refrain from drawing the obvious conclusion
about bad writing.
It seems clear that condensation as well as filtration, performed mechanically
or otherwise, should be an essential part of an MIS, and that such a system
should be capable of handling much, if not all, of the unsolicited as well as
solicited information that a manager receives.

The Manager Needs the Information That He Wants

Most MIS designers "determine" what information is needed by asking

managers what information they would like to have. This is based on the as-
sumption that managers know what information they need and want it.
For a manager to know what information he needs he must be aware of each
type of decision he should make (as well as does) and he must have an adequate
model of each. These conditions are seldom satisfied. Most managers have some
conception of at least some of the types of decisions they must make. Their
conceptions, however, are likely to be deficient in a very critical way, a way that
follows from an important principle of scientific economy: the less we under-
stand a phenomenon, the more variables we require to explain it. Hence, the
manager who does not understand the phenomenon he controls plays it "safe"
and, with respect to information, wants "everything." The MIS designer, who
has even less understanding of the relevant phenomenon than the manager, tries
to provide even more than everything. He thereby increases what is already an
overload of irrelevant information.
For example, market researchers in a major oil company once asked their
marketing managers what variables they thought were relevant in estimating
the sales volume of future service stations. Almost seventy variables were
identified. The market researchers then added about half again this many
variables and performed a large multiple linear regression analysis of sales of
existing stations against these variables and found about thirty-five to be statis-
tically significant. A forecasting equation was based on this analysis. An OR
team subsequently constructed a model based on only one of these variables,
traffic flow, which predicted sales better than the thirty-five variable regression
equation. The team went on to explain sales at service stations in terms of the

This content downloaded from

103.158.42.8 on Tue, 17 Oct 2023 14:16:10 +00:00
All use subject to https://about.jstor.org/terms
 B-150 RUSSELL L. ACKOFF

customers' perception of the amount of time lost by stoppong for service. The
relevance of all but a few of the variables used by the market researchers could
be explained by their effect on such perception.
The moral is simple: one cannot specify what information is required for
decision making until an explanatory model of the decision process and the
system involved has been constructed and tested. Information systems are
subsystems of control systems. They cannot be designed adequately without
taking control in account. Furthermore, whatever else regression analyses can
yield, they cannot yield understanding and explanation of phenomena. They
describe and, at best, predict.

Give a Manager the Information He Needs and His Decision

Making Will Improve

It is frequently assumed that if a manager is provided with the information

he needs, he will then have no problem in using it effectively. The history of
OR stands to the contrary. For example, give most managers an initial tableau of
a typical "real" mathematical programming, sequencing, or network problem
and see how close they come to an optimal solution. If their experience and
judgment have any value they may not do badly, but they will seldom do very
well. In most management problems there are too many possibilities to expect
experience, judgement, or intuition to provide good guesses, even with perfect
information.
Furthermore, when several probabilities are involved in a problem the un-
guided mind of even a manager has difficulty in aggregating them in a valid way.
We all know many simple problems in probability in which untutored intuition
usually does very badly (e.g., What are the correct odds that 2 of 25 people
selected at random will have their birthdays on the same day of the year?).
For example, very few of the results obtained by queuing theory, when arrivals
and service are probabilistic, are obvious to managers; nor are the results of
risk analysis where the managers' own subjective estimates of probabilities
are used.
The moral: it is necessary to determine how well managers can use needed
information. When, because of the complexity of the decision process, they
can't use it well, they should be provided with either decision rules or perform-
ance feed-back so that they can identify and learn from their mistakes. More on
this point later.

More Communication Means Better Performance

One characteristic of most MIS's which I have seen is that they provide
managers with better current information about what other managers and their
departments and divisions are doing. Underlying this provision is the belief
that better interdepartmental communication enables managers to coordinate
their decisions more effectively and hence improves the organization's overall
performance. Not only is this not necessarily so, but it seldom is so. One would
hardly expect two competing companies to become more cooperative because

This content downloaded from

103.158.42.8 on Tue, 17 Oct 2023 14:16:10 +00:00
All use subject to https://about.jstor.org/terms
 MANAGEMENT MISINFORMATION SYSTEMS B-151

the information each acquires about the other is improved. This analogy is not
as far fetched as one might first suppose. For example, consider the following
very much simplified version of a situation I once ran into. The simplification of
the case does not affect any of its essential characteristics.
A department store has two "line" operations: buying and selling. Each
function is performed by a separate department. The Purchasing Department
primarily controls one variable: how much of each item is bought. The Merchan-
dising Department controls the price at which it is sold. Typically, the measure
of performance applied to the Purchasing Department was the turnover rate
of inventory. The measure applied to the Merchandising Department was gross
sales; this department sought to maximize the number of items sold times
their price.
Now by examining a single item let us consider what happens in this system.
The merchandising manager, using his knowledge of competition and con-
sumption, set a price which he judged would maximize gross sales. In doing so
he utilized price-demand curves for each type of item. For each price the curves
show the expected sales and values on an upper and lower confidence band as
well. (See Figure 1.) When instructing the Purchasing Department how many
items to make available, the merchandising manager quite naturally used the
value on the upper confidence curve. This minimized the chances of his running
short which, if it occurred, would hurt his performance. It also maximized the
chances of being over-stocked but this was not his concern, only the purchasing
manager's. Say, therefore, that the merchandising manager initially selected
price P1 and requested that amount Q, be made available by the Purchasing
Department.
In this company the purchasing manager also had access to the price-demand
curves. He knew the merchandising manager always ordered optimistically.

o \\I
E Q2 --u-\

03 0 1

I E~SS4IM IC

Pt P2 P3
PRICE

FIGURE, 1. Price-demand curve

This content downloaded from

103.158.42.8 on Tue, 17 Oct 2023 14:16:10 +00:00
All use subject to https://about.jstor.org/terms
B-152 RUSSELL L. ACKOFF

Therefore, using the same curve he read over from Qi to the upper limit and
down to the expected value from which he obtained Q2, the quantity he actually
intended to make available. He did not intend to pay for the merchandising
manager's optimism. If merchandising ran out of stock, it was not his worry.
Now the merchandising manager was informed about what the purchasing
manager had done so he adjusted his price to P2. The purchasing manager in
turn was told that the merchandising manager had made this readjustment so
he planned to make only Q3 available. If this process-made possible only by
perfect communication between departments-had been allowed to continue,
nothing would have been bought and nothing would have been sold. This out-
come was avoided by prohibiting communication between the two departments
and forcing each to guess what the other was doing.
I have obviously caricatured the situation in order to make the point clear:
when organizational units have inappropriate measures of performance which
put them in conflict with each other, as is often the case, communication be-
tween them may hurt organizational performance, not help it. Organizational
structure and performance measurement must be taken into account before
opening the flood gates and permitting the free flow of information between
parts of the organization. (A more rigorous discussion of organizational structure
and the relationship of communication to it can be found in [1].)

A Manager Does Not Have to Understand How an Information System

Works, Only How to Use It

Most MIS designers seek to make their systems as innocuous and unobtrusive
as possible to managers lest they become frightened. The designers try to provide
managers with very easy access to the system and assure them that they need
to know nothing more about it. The designers usually succeed in keeping man-
agers ignorant in this regard. This leaves managers unable to evaluate the MIS as
a whole. It often makes them afraid to even try to do so lest they display their
ignorance publicly. In failing to evaluate their MIS, managers delegate much of
the control of the organization to the system's designers and operators who
may have many virtues, but managerial competence is seldom among them.
Let me cite a case in point. A Chairman of a Board of a medium-size company
asked for help on the following problem. One of his larger (decentralized) divi-
sions had installed a computerized production-inventory control and manu-
facturing-manager information system about a year earlier. It had acquired
about $2,000,000 worth of equipment to do so. The Board Chairman had just
received a request from the Division for permission to replace the original
equipment with newly announced equipment which would cost several times
the original amount. An extensive "justification" for so doing was provided
with the request. The Chairman wanted to know whether the request was
really justified. He admitted to complete incompetence in this connection.
A meeting was arranged at the Division at which I was subjected to an ex-
tended and detailed briefing. The system was large but relatively simple. At
the heart of it was a reorder point for each item and a maximum allowable

This content downloaded from

103.158.42.8 on Tue, 17 Oct 2023 14:16:10 +00:00
All use subject to https://about.jstor.org/terms
 MANAGEMENT MISINFORMATION SYSTEMS B-153

stock level. Reorder quantities took lead-time as well as the allowable maximum
into account. The computer kept track of stock, ordered items when required
and generated numerous reports on both the state of the system it controlled
and its own "actions."
When the briefing was over I was asked if I had any questions. I did. First
I asked if, when the system had been installed, there had been many parts
whose stock level exceeded the maximum amount possible under the new system.
I was told there were many. I asked for a list of about thirty and for some graph
paper. Both were provided. With the help of the system designer and volumes
of old daily reports I began to plot the stock level of the first listed item over
time. When this item reached the maximum "allowable" stock level it had been
reordered. The system designer was surprised and said that by sheer "luck" I
had found one of the few errors made by the system. Continued plotting showed
that because of repeated premature reordering the item had never gone much
below the maximum stock level. Clearly the program was confusing the maximum
allowable stock level and the reorder point. This turned out to be the case in more
than half of the items on the list.
Next I asked if they had many paired parts, ones that were only used with
each other; for example, matched nuts and bolts. They had many. A list was pro-
duced and we began checking the previous day's withdrawals. For more than
half of the pairs the differences in the numbers recorded as withdrawn were very
large. No explanation was provided.
Before the day was out it was possible to show by some quick and dirty
calculations that the new computerized system was costing the company almost
$150,000 per month more than the hand system which it had replaced, most of
this in excess inventories.
The recommendation was that the system be redesigned as quickly as pos-
sible and that the new equipment not be authorized for the time being.
The questions asked of the system had been obvious and simple ones. Man-
agers should have been able to ask them but-and this is the point-they felt
themselves incompetent to do so. They would not have allowed a handoperated
system to get so far out of their control.
No MIS should ever be installed unless the managers for whom it is intended
are trained to evaluate and hence control it rather than be controlled by it.

A Suggested Procedure for Designing an MIS

The erroneous assumptions I have tried to reveal in the preceding discussion

can, I believe, be avoided by an appropriate design procedure. One is briefly
outlined here.

1. Analysis Of The Decision System

Each (or at least each important) type of managerial decision required by the
organization under study should be identified and the relationships between
them should be determined and flow-charted. Note that this is not necessarily the
same thing as determining what decisions are made. For example, in one com-

This content downloaded from

103.158.42.8 on Tue, 17 Oct 2023 14:16:10 +00:00
All use subject to https://about.jstor.org/terms
B-154 RUSSELL L. ACKOFF

pany I found that make-or-buy decisions concerning parts were made only at
the time when a part was introduced into stock and was never subsequently
reviewed. For some items this decision had gone unreviewed for as many as
twenty years. Obviously, such decisions should be made more often; in some
cases, every time an order is placed in order to take account of current shop
loading, underused shifts, delivery times from suppliers, and so on.
Decision-flow analyses are usually self-justifying. They often reveal important
decisions that are being made by default (e.g., the make-buy decision referred to
above), and they disclose interdependent decisions that are being made in-
dependently. Decision-flow charts frequently suggest changes in managerial
responsibility, organizational structure, and measure of performance which can
correct the types of deficiencies cited.
Decision analyses can be conducted with varying degrees of detail, that is,
they may be anywhere from coarse to fine grained. How much detail one should
become involved with depends on the amount of time and resources that are
available for the analysis. Although practical considerations frequently restrict
initial analyses to a particular organizational function, it is preferable to perform
a coarse analysis of all of an organization's managerial functions rather than a
fine analysis of one or a subset of functions. It is easier to introduce finer in-
formation into an integrated information system than it is to combine fine sub-
systems into one integrated system.

2. An Analysis Of Information Requirements

Managerial decisions can be classified into three types:
(a) Decisions for which adequate models are available or can be constructed
and from which optimal (or near optimal) solutions can be derived. In such
cases the decision process itself should be incorporated into the information
system thereby converting it (at least partially) to a control system. A decision
model identifies what information is required and hence what information
is relevant.
(b) Decisions for which adequate models can be constructed but from which
optimal solutions cannot be extracted. Here some kind of heuristic or search
procedure should be provided even if it consists of no more than computerized
trial and error. A simulation of the model will, as a minimum, permit comparison
of proposed alternative solutions. Here too the model specifies what information
is required.
(c) Decisions for which adequate models cannot be constructed. Research is
required here to determine what information is relevant. If decision making
cannot be delayed for the completion of such research or the decision's effect is
not large enough to justify the cost of research, then judgment must be used to
"guess" what information is relevant. It may be possible to make explicit the
implicit model used by the decision maker and treat it as a model of type (b).
In each of these three types of situation it is necessary to provide feedback
by comparing actual decision outcomes with those predicted by the model or
decision maker. Each decision that is made, along with its predicted outcome,

This content downloaded from

103.158.42.8 on Tue, 17 Oct 2023 14:16:10 +00:00
All use subject to https://about.jstor.org/terms
 MANAGEMENT MISINFORMATION SYSTEMS B-155

should be an essential input to a management control system. I shall return

to this point below.

S. Aggregation Of Decisions

Decisions with the same or largely overlapping informational requirements

should be grouped together as a single manager's task. This will reduce the
information a manager requires to do his job and is likely to increase his under-
standing of it. This may require a reorganization of the system. Even if such a
reorganization cannot be implemented completely what can be done is likely to
improve performance significantly and reduce the information loaded on man-
agers.

4. Design Of Information Processing

Now the procedure for collecting, storing, retrieving, and treating information
can be designed. Since there is a voluminous literature on this subject I shall
leave it at this except for one point. Such a system must not only be able to
answer questions addressed to it; it should also be able to answer questions that
have not been asked by reporting any deviations from expectations. An extensive
exception-reporting system is required.

5. Design Of Control Of The Control System

It must be assumed that the system that is being designed will be deficient
in many and significant ways. Therefore it is necessary to identify the ways in
which it may be deficient, to design procedures for detecting its deficiencies, and
for correcting the system so as to remove or reduce them. Hence the system
should be designed to be flexible and adaptive. This is little more than a platitude,
but it has a not-so-obvious implication. No completely computerized system
can be as flexible and adaptive as can a man-machine system. This is illustrated
by a concluding example of a system that is being developed and is partially in
operation. (See Figure 2.)
The company involved has its market divided into approximately two hundred
marketing areas. A model for each has been constructed as is "in" the computer.
On the basis of competivive intelligence supplied to the service marketing
manager by marketing researchers and information specialists he and his staff
make policy decisions for each area each month. Their tentative decisions are
fed into the computer which yields a forecast of expected performance. Changes
are made until the expectations match what is desired. In this way they arrive
at "final" decisions. At the end of the month the computer compares the actual
performance of each area with what was predicted. If a deviation exceeds what
could be expected by chance, the company's OR Group then seeks the reason
for the deviation, performing as much research as is required to find it. If the
cause is found to be permanent the computerized model is adjusted appropriately.
The result is an adaptive man-machine system whose precision and generality
is continuously increasing with use.
Finally it should be noted that in carrying out the design steps enumerated

This content downloaded from

103.158.42.8 on Tue, 17 Oct 2023 14:16:10 +00:00
All use subject to https://about.jstor.org/terms
B-156 RUSSELL L. ACKOFF

PROPOSED POLICIES

MARKETING EVALUATED PROPOSALS MARKET AREA MEMORY a

MODELS COM PARATOR
MANAGEMENT SELECTED POLICY (COMPUTER) PREDICTED (COMPUTER)
OUTCOME

2 Ui INFORMATION F INQUIRIES Z
-j2
<~~~~~~~~~~~
oz DATA D
FIELD _MOTHERA

SOURCES

MARKET DATA OPERATIONS

o~~~~~~~~~~~~~~~~~~~ UJ DEVIANT AREAS
INFORMATION INFORMATION RESEARCH
SYSTEM GROUP

M ~~~~~~DATA
oz 1

FIELD DATA ACTUAL PERFORMANCE

MARKET
SELLINGARS
FORCEARS

FIGURE 2. Simplified diagram of a market-area control system

above, three groups should collaborate: information systems specialists, oper-

ations researchers, and managers. The participation of managers in the design
a system that is to serve them, assures their ability to evaluate its performance
by comparing its output with what was predicted. Managers who are not willing
to invest some of their time in this process are not likely to use a management
control system well, and their system, in turn, is likely to abuse them.

Reference

1. SENGUPTA, S. S., AND ACKOFF, R. L., "Systems Theory from an Operations Research
Point of View," IEEE Transactions on Systems Science and Cybernetics, Vol. 1 (Nov.
1965), pp. 9-13.

This content downloaded from

103.158.42.8 on Tue, 17 Oct 2023 14:16:10 +00:00
All use subject to https://about.jstor.org/terms