APR 02 1990
SUPPORTING SENIOR EXECUTIVES'
MODELS FOR
PLANNING AND CONTROL
Michael
E.
Treacy
90s:
85-008
Management
in
the 1990s
SUPPORTING SENIOR EXECUTIVES'
MODELS FOR
PLANNING AND CONTROL
Michael
E.
Treacy
90s: 85-008
June, 1984
CISRWP#125
Sloan
®
This paper appears in
WP# 1676-85
M.
E.
Treacy
The Rise of Managerial Computing: The Best of the Center for
F
Rockart and C. V. Bullen, eds., Dow Jones-Irwin, Homewood,
J
Information Systems Research,
Illinois,
1986.
Management in the 1990s
Sloan Scnool of Management
Massachusetts Institute of Technology
Management
in the 1990s is an industry and governmental agency supported
research program. Its aim is to develop a better understanding of the managerial
issues of the 1 990s and how to deal most effectively with them, particularly as these
issues revolve around anticipated advances in Information Technology.
Assisting the
work of the Sloan School
working partners
in
scholars with financial support
and
as
research are:
American Express Travel Related Services Company
Arthur Young and Company
British Petroleum Company, p. I.e.
BellSouth Corporation
Equipment Corporation
Eastman Kodak Company
General Motors Corporation
International Computers, Ltd.
MCI Communications Corporation
United States Internal Revenue Service
Digital
The conclusions or opinions expressed in this paper are those of the author(s) and
do not necessarily reflect the opinion of Massachussetts Institute of Technology,
Management
in
the 1990s Research Program, or
its
sponsoring organizations.
SUPPORTING SENIOR EXECUTIVES' MODELS
FOR PLANNING AND CONTROL
by
Michael E. Treacy
1.
Introduction
By now the successful
use of models
documented and easily observable.
in corporate
Mclnnes and
decision making is well
Carleton
[1982],
Wooller [1977], Naylor and Schauland [1976], and Traenkle, et
all
observed
accelerating use of decision support system
functional
corporate planning.
needs
in
Very few explicit
or used by senior executives
marketing,
finance,
-
or formal
operations
-
[Rockart and Treacy,
[1975] have
(DSS)
Most of these models are built and used by mid-level
packages.
support
the
al_
Grinyer and
modeling
analysts to
and
management,
models appear to be built
1982].
If
these are the
people responsible for the formulation of broad strategies and the design of
organizational
control mechanisms,
then it follows that explicit modeling has
not had as much impact upon these areas as some proponents would desire.
Yet,
models
reality,
simplify
these
are
senior executives use models for planning and control.
intuitive
abstractions
the
decision
and
of
implicit.
complex
process,
They
decision
to
are
mental
contexts,
identify
that
important
Their
representations
of
use
to
and
to
executives
variables,
They are not usually put down on paper,
generate and evaluate alternatives.
built
into
a
computer
analyzed
program,
quantitatively,
even
or
viewed
as
They are often simple and inelegant and may bear
models, but they are models.
little relationship to "good theory," but they are arguably the most important
models for planning and control, because they are powerful
and they get used.
As one executive put it:
I
Just
bring a lot of knowledge to the party.
scanning the current status of our operations
enables me to see some things that those with less
time in the company would not see as important.
[Rockart and Treacy, 1982]
This
clearly
executive
operation.
These
models
well-developed
many
has
allow
him
to
quickly
implicit
someone less experienced would have to formally analyze.
in
Figure
How well
trouble because
is this company doing?
sales
are
coimionly held implicit model
that
illustrates
1
declining.
One reaction is that the firm is
This
conclusion
is
that sales are always rising in
a
derived
sales
pattern may
model
is required to interpret the data.
these
numbers may
actually
model
of expected
sales was much
only becomes
models
simply
indicate
see
a
seasonality,
very
lower.
and
a
more
from
a
healthy company.
But what if the time frame represented is only twelve months?
These
his
Looking at the sales performance reflected in the trend line, one
this point.
might ask:
of
infomation
new
assess
models
Then
complex
the
informal
An experienced executive looking at
healthy
situation because
This example
his
illustrates
or
her
that data
information when it is interpreted through some implicit model.
are
nothing
experience and knowledge.
other
than
the
accumulation
of
an
individual's
3-
Sales
Time
Figure
1
Analyzing Sales Data With implicit Models
-4-
to understand
Does the executive really need an explicit or formal model
his
analyzed using
modeling
of
range of techniques.
a
understood
who
Explicit models
operations?
her
or
operations
firm's
the
easily
^ery
argue
can
a
manager
benefit
by
formal
that what
the
executive
but support for his or her implicit modeling,
needs is not formal models,
be
that even
could
well
they
and
shared
Some would argue
Others might
operations.
those
are
so
that the weaknesses of this abstract mental modeling can be reduced.
This paper contributes
toward
the
debate
over whether
senior executives
should be supported with explicit models or by other measures.
The answer depends upon the type of decision contexts faced
is a simple one.
We analyze the nature of explicit and implicit models and
by the executive.
alternatives for supporting these different modeling processes.
this,
the
section
rest
Our position
of
discusses
senior executives'
paper
the
the
nature
divided
is
into
of executive
The
next
needs and concludes
that
several
analytic
most important problems are complex
that complete and formal
To accomplish
sections.
unstructured
and
and
analyses of these problems are not always possible.
The section concludes with
a
simple categorization of analyses that is later
used to discuss alternate types of support systems.
Section 3 develops a set of criteria for evaluating the appropriateness of
alternate
forms
executives.
of
analysis
of
support
for
the
different
analytic
This is done by examining the potential
may
incur,
for
a
alleviate or avoid such problems.
of decision support systems.
major
objective
The final
of
needs
of
senior
problems that each form
support
systems
is
to
section discusses alternate forms
These different forms derive from differences in
-5-
tools,
DSS
time
some
so
devoted
is
to
a
discussion
of
these
One
tools.
conclusion drawn in this section is that new, emerging forms of modeling, such
as
fuz^ modeling, may provide support
for many
of
most
senior executives'
important analytic problems.
2.
The Nature of Senior Executive Analysis
There
is
a
organizations,
explanations
of
lot
little
but
suggest
of
use
formal
them
of
use
themselves.
models
may
It
senior
by
control
in
Several
executives.
senior
that
be
and
planning
for
executives
do
not
building and analysis or for using
possess the necessary skills for model
a
But Rockart and Treacy have found that even among senior executives
computer.
who use computers for analytic support there is little use made of explicit
It may be that executives don't need
models [1981, 1982].
fomal models; or
that their information requirements are simple enough that ad hoc analysis is
sufficient.
are
too
Or, it may be quite the opposite,
complex
for
explicit modeling
to
that the problems they confront
be
of
much
benefit.
This
third
of
senior
possiblility we believe is often closest to the truth.
Much
has
executives.
with
been
written
about
the
roles
information
and
needs
Mintzberg [1973] observed that senior managers are often dealing
overwhelming
computer-generated
complexity
under
information
was
great
of
time
little
pressure.
use
to
these
He
found
people.
that
They
preferred current and timely information that could be obtained only through
informal
channels.
Gorry and Scott Morton
[1971] build upon Anthony's
[1965]
-6-
and
framework
characterized
largely unstructured.
role
the
of
senior executives
strategic
as
and
generating
They require support for defining problems,
alternative solutions, and evaluating them against strategic criteria.
issues that senior executives
Most of the important planning and control
must face, such as the creation of plans, the control of subordinates, or the
analysis of competitive advantage, are highly dependent upon the exact context
of
the
problem.
operating
standard
They
deal
with
Sufficient
procedures.
problems is very difficult to obtain.
work
under great uncertainty.
problems
unique
There
and
complete
Therefore,
is
no
for
which
there
knowledge
are
no
these
of
senior managers must often
"normative"
approach
planning
to
and control; there is no clear set of decisions which arise in the course of
this process.
as
future
Rather, data concerning both past and current results as well
plans,
are
scanned
and
analyzed,
variable decision opportunities occur.
and
As Charles
a
set
of
ad
hoc
,
highly
Hitch [1957] observed some
years ago:
The sort of simple explicit model which operations
researchers are so proficient in using can certainly
reflect most of the significant factors influencing
traffic control on the George Washington Bridge, but
the proportion of the relevant reality which we can
represent by any such model or models in studying, say,
a major foreign policy decision, appears to be almost
trivial, [p. 718]
Adding to the difficulties of
what Bellman
[1957] calls
aspire to plan and control
a
lack of structure and great uncertainty is
"the curse
of multidimensional ity."
Managers may
their firms for the single objective of long
term
-7-
objectives,
profitability, but in practice they require many operational
as present period profit plans and gaining X% of market share,
along
the way.
Thus,
there
as milestones
independent and
their
objective of long run profits is often not well
relationship to the overall
understood.
not
objectives are
operational
These
such
further
is
introduced
uncertainty
by
multiple
and
fuz?y evaluation criteria.
We may conclude that the most important problems of planning
and control
These two
faced by senior executives are complex and relatively unstructured.
characteristics, complexity and structure, define a categorization of analytic
situations that is helpful
executives.
In
Figure
2
in
discussing alternate ways of supporting senior
have
we
illustrated
categorization.
this
For
expository purposes, quadrants have been used to define four types of analytic
The qualifier
Situations.
"relatively"
indicate that there
has been used to
are not clear demarcations between these categories.
important problems
The
are
analytic
of
planning
illustrated
situations
and
the
in
control
upper
faced by
right
senior managers
they
quadrant:
are
complex, relatively unstructured problems.
In
the
lower left quadrant are
situations
analytic
where
extent of
the
knowledge about the problem is relatively complete and the problem is not very
In
operating
procedures
quadrant
complex.
are
such
problem
situations,
complex.
analytic
can
often
situations
For these problems,
be
solving
routinely
that
are
is
fairly
applied.
relatively
easy
In
well
the
and
standard
upper
structured,
left
but
standard operating procedures or rules of thumb
-8-
«•«
PROGRAM COMPLEXITY
relatively
complex
•••• •••ai
relatively
simple
•••I
relatively
relatively
structured
unstructured
PROBLEM STRUCTURE
Figure 2
Four Types of Analytic Situations
•9-
Formal
are often inadequate.
typically in the form of an explicit
analysis,
model, is required to integrate many pieces of information in a structured way.
difference
The
between
these
problems can be illustrated with
can
problem
be
viewed
as
simple
a
economic order quantity can
problem of
the
one,
applied,
be
formal models are needed to pull
to
or
inventory
which
as
well -structured
relatively
of
types
two
rules
thumb
of
such
Material
relevant information together.
all
It can
requirements planning is an example of the latter view of the problem.
be
implemented
with
only
the
of
aid
large,
Which
models.
formal
as
for which
complex problem,
a
This
reordering.
view
is
appropriate for structured problems depends primarily upon the level of effort
one wishes
to
expend
to
generate
a
satisfactory
solution.
For unstructured
problems, it also depends upon what is and is not known about the problem.
It is evident from the preceding examples that some problems can be viewed
An inventory problem can
as fitting in any of the four quadrants of Figure 2.
and the greater one's
be simple or complex, depending upon how it is viewed,
experience and understanding of such problems, the more structure they appear
to
The more
have.
provide
a
unstructured the
complete represention;
important it is to provide
a
problem
area,
the
given
executives,
a
choice.
by
the
valid formal analysis.
nature
able
are
the more complex the problem area,
of
the
problems
we
to
the more
Thus, given a choice, one
would opt for the most complete analysis and perhaps
senior
less
a
they
formal
face,
analysis,
are
not
but
often
10-
For each
the
of
illustrated
situations
analytic
analytic approaches can be applied.
In
particular,
Figure
in
different
2,
one can choose
explicit or implicit models to any of these types of problems.
to
apply
Which form of
modeling is most appropriate for each type of analytic situation is examined
in the next section by developing a list of strengths and weaknesses for each
lists are
These
analytic approach.
also
used
in
subsequent
the
section
to
determine which form of support system is most appropriate for each type of
analytic situation.
Potential Problems with Explicit and Implicit Models
3.
Little
models
are
has
developed
simple,
a
list
robust,
easy
criteria
of
control,
to
with, and as complete as possible.
executives
have
derived
been
cases, and critical assumptions can all
to examine alternate
scenarios,
current
observation negates part of
a
a
Subtle
to
Good
communicate
implicit mental models
from
The mental models
observing
relationships,
complex
exceptional
These
And one can use them in many ways
as part of
information.
account for what was observed.
as
[1970].
be part of a good mental model.
implicit models are also easy to control.
with
easy
They can be robust.
relationships and events over many years.
conjunction
adaptive,
By these criteria,
be carried around in a person's head.
experienced
models
good
They are simple, or at least simple enough that they can
can be good models.
of
for
a
larger analytic
Mental
models
are
process,
adaptive.
-
or in
If
manager's model, it will probably be modified to
They are as complete as possible, as complete
manager has been able to comprehend.
-11-
criteria
Little's
can
however,
good models,
for
They can be viewed as a list of the potential
way.
models
Formal
criteria.
Little's
meet
easily
so
problems introduced when
This would account for why mental
implicit mental models are made explicit.
models
another
interpreted
be
tend
be
to
complicated; they can sometimes give absurd results when certain complexities
are
They
missing.
only
can
controlled
be
They usually are difficult to
facilities are built into the modeling system.
adapt
situations
other
to
and
explicit
And
information.
new
to
whatever
through
exercised
and
models
certainly are not easily communicated with by a typical manager unaccustomed
to their use.
Not all
pick out the good ones.
than
ills
That
research
Elam,
for
et
al_
But,
why
is
those
[1980],
and
at
Or,
least
formal
to
Will
[1975]
their
suffer more of these
shortcomings
are
important
and
new
models.
Konsynski
and
Dolk
is
each
present
conceptual
more
of
area
an
management
model
conmiitted
formal models
in general,
models.
informal
do
obvious.
Little provided a list that helps to
formal models are this bad.
[1982],
for
schemes
modeling systems that would help automatically to reduce some of the problems
with explicit models.
If these
ideas could be implemented, then we might see
more managers approaching problems using formal modeling methods because they
would be perceived as more valuable relative to implicit mental models.
Perhaps the most important limitation of mental models is that they do not
handle
complexity
relationships
remarkably
well.
between
broad
range
It
is
variables.
of
difficult
Our
knowledge,
integrators of that knowledge.
but
to
mental
people
mentally
integrate
capacities
are
not
can
complex
handle
particularly
a
good
-12-
A
manipulate.
with
problem
second
implicit models
these
adequately describes
Once a model
a
is
that
they
to
phenomenon, one might like to
use it to predict or to optimize some aspect of the phenomenon.
This requires
trying alternative scenarios, assessing the impact
manipulation of the model,
the other techniques that have become familiar through
of uncertainty, and all
modeling
line-oriented
hard
are
This
systems.
cannot
done
be
as
easily
with
as
a
formal mathematical model.
Another problem with mental
because
planning
it
This
experience.
that
they
are
difficult to
share
And it is even harder to convey the supporting evidence for the
with others.
model
models is
is
intangible
and
gathered
important problem because
an
control
and
often
is
is
a
large
and
the
widespread
years
through
of
process of management
undertaking.
requires
It
specialists' in particular areas, such as a strategic planning, accounting, and
finance.
can
only
In
addition,
be
provided
it requires the insight,
by
senior
line
analysis,
management.
Thus,
and guidance which
it
is
a
shared
activity, involving many different actors with only partly shared perspectives.
A fourth problem with
implicit models
Mental models are difficult to validate.
stems
from problems
one
and
two.
If they cannot easily be manipulated
and if they cannot be easily shared with others, then they are also difficult
to
test
against
impartial
information processors.
mental models.
evidence.
This may
result
We
in
know
that
imperfect,
people
perhaps
are
imperfect
badly
Without proper validation, these flaws may not surface.
flawed
-13-
Finally,
implicit models
are
vastly
underutilized.
Managers often know
much more about situations than they bring to the decision-making processes of
planning and control.
There is a major, uncatalogued resource out there that
is not being adequately utilized.
In
Figure 3 we have summarized the preceding discussion as a list of the
primary potential
of
a
support
problems with formal
system
and mental
executive modeling
for
is
models.
to
A major objective
alleviate
some
of
these
problems.
Formal Models;
*
*
*
*
*
complicated and hard to understand
difficult to make robust
difficult to control
difficult to adapt
difficult to communicate with
Mental Models
*
*
*
*
do not handle complexity well
difficult to manipulate
difficult to share
difficult to validate
uncatalogued and underutilized
*
Figure 3
Primary Potenti al Probl ems wi th
Formal and Informal Models
4.
Alternate Forms of Analytic Support
Among all
the
buzzwords, jargon,
and acronyms of the information systems
field, decision support systems (DSS) has proven to be a term of practical and
philosophical
substance.
As
a
practical
matter,
DSS focuses on the analysis
of data and explicit modeling of the business and its environment.
A manager
-14-
not
viewed
is
as
passive
a
consumer
information,
of
but
as
active
an
participant in the development and use of his or her personalized, supporting
process
development
is
necessary
Adaptive design and user control
with
perspective,
user's
the
manager
wants
desires.
to
maintain,
require
system
a
that
the
the
in
manager's
a
of
process
parts
those
evolve
to
the
of
the DSS
Adaptive
approach.
the
of
fit
to
direction
support
job
that
and
job.
begin
that
the
or
she
he
The user is assumed to be knowledgeable about his or her role and,
if provided with the appropriate data,
better
toward
evolve
will
tenet
supporting
and
not just involvement in,
over,
fundamental
a
are
design
evolutionary
user control
Thus,
computer system.
equivalent
Lindblom's
to
software tools, training and guidance,
management.
It
"muddling
[1959]
is
information
the
through."
This
systems
non-invasive,
adaptive philosophy has been key to DSS's extraordinary practical success.
During
past
the
ten
years,
a
host
decision
of
support
systems
(DSS)
generators and software packages have been developed that allow analysts and
managers
to
differences
forms
of
develop
among
analytic
the
and
major
support
directly
categories
presently
decision
use
packages
of
available.
support
define
Figure
systems.
the
4,
The
alternative
adapted
from
Montgomery and Urban [1969], shows the capabilities that have been provided by
different DSS generators.
A
decision
support
system
provides
a
manager
with
another
source
information on his or her internal and external business environment.
an
interaction and display facility that may
of
Through
include a command language and
query, report writing, and color graphics facilities, the manager can access a
>
-15-
MANAGER
DECISION
MODELS
i
—
Y
I
INTERACTION
AND DISPLAY
STATISTICS
AND
DATA BASE
MANIPULATION
1
ENVIRONMENT
Figure 4
Comprehensive DSS Capabilities
•16-
base
of
data,
and
functions,
statistical,
perform
explicit
create
arithmetic,
models
and
his
of
data
manipulation
firm,
competitors,
other
her
or
industry, and the economy.
Figure 4 represents
an
ideal
generator capabilities
DSS
set of
With
.
these capabilities it is now possible to build many different types of support
systems.
The question we
the different analytic
face is which form of support is best for each of
situations discussed in Section 2?
In
Figure
5,
four
different forms of support are indicated for the four analytic situations.
the rest of this section we will
explain
how
they
relieve
the
In
describe each of these forms of support and
potential
problems
explicit
of
and
implicit
models discussed in the last section.
Model -Oriented DSS for Complex, Well -Structured Problems
Complex problems that are
with
formal
explicit
models.
relatively well
In
a
structured are
well -structured
problem
better analyzed
area,
there
is
enough knowledge to construct a fairly complete representation of the problem
context.
is
the
complex,
Hence,
analytic
fomal modeling
approach
well -structured
modeling environment
reduced.
in
of
is
possible and for significant problems it
choice.
problems
should
which modeling
system
A
be
is
to
aimed
support
toward
the
analysis
providing
easier and potential
a
of
formal
problems are
17-
PROBLEM COMPLEXITY
relatively
complex
relatively
simple
MODEL-ORIENTED
DSS
18-
focused
support
decision
of
the
on
just
Many
such
form
a
generators
system
capabilities
The
support.
of
line-oriented modeling packages are diagrammed in Figure
modeling
line-oriented
A
spreadsheet,
gives
interrelated
several
time
manipulate
manager
a
several
model,
to
model
calculate
alternate
These are
the
results
line-oriented
reports and graphs.
and generate
They do not,
of
of
model
over
model
the
user
easily
to
scenarios,
systems
defined line by line, with each line
electronic
an
as
the
and
assumptions
these
of
explicit
an
allow
packages
the
known
define
to
are
6.
in
and
the
to
sense
representing another
The packages usually allow a manager to define,
variable.
a
is
and
test
commonly
more
ability
Typically,
perform sensitivity analyses.
that the model
the
variables
periods.
the
system,
available
presently
and analyze
solve,
in general,
manage
supporters,
but they
a
data base or offer ad hoc analytic capabilities.
These
line-oriented modeling
systems
have won
many
are still not user-friendly enough to have gained a significant base of senior
executive users, except in financial
the creation of formal
to
create
also
a
address many
designed
providing
to
a
make
of
the
modeling
powerful
however,
ideal
for
models if the analyst has enough knowledge to be able
complete
fairly
They are,
functions.
but
representation
potential
easier,
problems
less
uncomplicated
building and model analysis commands.
of
the
with
formal
complicated,
syntax
that
These
problem.
and
models.
more
incorporates
packages
They
friendly
both
are
by
model
-19-
-20-
The conmand languages used in this type of software also make it easier to
control
data
Changing
models.
values,
different
performing
analyses,
and
changing the fomiat of printouts are all easier with these packages than with
a
standard programming
Because the models
language.
are
syntactically
less
complicated and easier to understand, they are also easier to adapt.
The one potential problem that these packages do not help to alleviate is
difficulty
the
in
making
might be used to create
a
formal
models
For example,
robust.
this
software
sales forecasting model, but because the package is
only a language and has no built-in intelligence about sales forecasting,
it
can in no way assure that the model is correct.
modeling
line-oriented
These
accounting-based
models
in
fairly
systems
used
are
well -understood
areas,
for
extensively
such
evaluation of projects, budgeting, and consolidation analysis.
as
financial
Most managers
understand enough accounting to be able to evolve useful and usable financial
models through an interactive process of model
one
way
that
the
semantic
potential
problems
building and analysis.
of
formal
models
have
Thus,
been
overcome is by basing the models upon well -understood accounting principles.
Data-Oriented DSS for Simple, Well -Structured Problems
Simple
and
well -structured
monitoring and tracking sales,
data
that
they
can
use
to
problems
are
for example,
trigger
rules
fairly
easy
to
support.
In
what managers need is access to
of
thumb
or
standard
operating
-21-
This is the type of support offered by data-oriented DSS.
procedures.
These
systems are built to provide managers with ready access to a pre-defined base
They are often used as supplements for paper-based reporting systems
of data.
because they offer more flexible reporting formats as well as a limited set of
These systems are used to support monitoring in areas where
analytic tools.
an executive
has
fairly well -formed mental
a
model
of expected performance.
But because the data covers only a limited area and analytic tools are also
data-oriented
limited,
DSS
for
appropriate
only
are
recurring,
reasonably-structured problem areas such as monitoring performance vs. budget.
The software that data-oriented decision support systems are based upon is
quite
different
usually
called
management
in
RAMIS,
software
the
"friendly"
systems
illustrated
FOCUS,
from
(DBMS).
Figure
NOMAD,
(that
The
Some
7.
is,
and
well-known
and DBASEII.
performing
very
easy
to
capabilities
facility for managing and accessing
graphs,
for model -oriented
used
These
of"
learn
this
products
systems.
and
use)
class
in
packages provide
a
data
base
software
are
marketplace
are
of
the
is
It
manager with
a
large base of data, creating reports and
a
limited
analysis
upon
the
data.
They
give
managers the ability to choose the data that they wish to see and to format it
in reports or graphs as they wish to see
the
line-oriented modeling
software,
it.
Unfortunately, as with much of
the user interfaces for these data base
packages are not yet friendly enough to encourage widespread use by executives.
•22-
23-
Information Support for Simple, Unstructured Problems
the more unstructured they are,
Even for relatively simple problems,
difficult
more
appropriate
becomes
it
relative
Concurrently,
for support.
used to address problems
to
its
pre-define
to
implicit mental
the
succession planning or how
such as
competition
become
subject
models
are
that are
firm is doing
a
questionable
to
must be based upon at least three design elements:
a
that
validity,
Support for simple but unstructured problems
manipul ability, and sharability.
(2)
reports
and
data
the
the
large base of data,
a
(1)
set of ad hoc analytic tools, and (3) sharing of the system by managers
with common information needs.
support
system,
which
the
is
These design elements
most
widely
define
form
implemented
information
an
of
support
for
senior executives.
In
an
increasing
number
of
companies,
level
a
of
and
data
software
commonality have been established; their information support systems resembles
Figure 8.
naturally
planners
Bases of data have been established for groups of individuals who
draw
and
upon
the
same
controllers
type
created
of
one
data.
In
base
of
managers and market researchers created another.
Chicago,
corporate
staff
[Rockart and Treacy,
1982]
and
In
executives
all
both firms,
data,
while
from
one
the
strategic
the
marketing
Industries
At Northwest
draw
it was
company,
one
base
the managers
and
of
in
data.
analysts
who established what data would be in the data bases and what data definitions
would be used.
And,
despite the common data bases,
also maintain their own personal data files.
in every case
the
users
24-
USER INTERFACE
Specific
-25-
With present information requirements analysis
These data bases evolve.
techniques and with the existing state of data management in most firms, only
about half the needed data can be put into a data base on the first pass.
In
addition, the nature of the executive's job is one of constant change and new
data are often needed to support new problems and opportunities.
Each of these three design elements helps to reduce problems of validity
A large base of data combined with analytic tools
and sharing mental models.
allows
the
manager
against the data.
models.
are
build,
to
explore
test
his
or
her
implicit models
this way, data analysis serves to help validate mental
In
problems with these implicit models is that they
One of the potential
difficult to share,
difficult to
information support system provides
If managers
implicit models.
and
a
transfer to another person.
A shared
language and a medium for communicating
support their implicit models through a system,
then the models become more transferable with the sharing of data and analysis.
The
data
software
management
that underlies
and
ad
hoc
information
analysis
support
capabilities.
systems
The
comprehensive and contain all the elements shown in Figure
must
have
packages
both
must
be
4.
Fuzzy Modeling and Expert Support for Complex and Unstructured Problems
Most implicit mental models are fuzzy, but computer-based modeling systems
are precise.
computer-based
Until
we can develop fuzzy modeling systems, most of the useful,
support
for
implicit
models
will
be
provided
through
data
-26-
and
analysis,
retrieval,
the
Fuzzy modeling systems will
internal
inconsistent
and
transforming
allow
a
different
into
formats.
manager to externalize crude, incomplete,
without
models
data
of
having
complete and internally-consistent models like those
to
translate
found
them
into
in model -oriented
DSS.
Fuzzy models are used by senior managers all
pricing decision,
(2)
(3)
of
manager might bring into play the following rules:
Our price should be about two times direct
costs.
Our price should be just below our dominant
competitor's price.
If our competitor's prices go too high, we
should price for increased market share.
(1)
None
a
For example, in a
the time.
these
statements
is
standard formal modeling system.
in
form
a
precise
enough
to
be
used
Each statement begs for refinement.
in
a
With a
fuzzy modeling system, however, the above statements form a model of a pricing
decision that could be solved.
be
satisfactory,
The results of that solution clearly would not
but it is a useful
starting point that prompts the manager
for a more refined representation of his or her mental model.
Fuzzy
models
problems,
with.
the
These
complexity.
potential
are
type
of
models
But,
flaws
of
ideally
suited
problems
have
the
formal
precision and consistency.
models
senior
that
power
conceptually,
complex
to
of
fuzzy
such
as
the
relatively
executives
formal
models
and
must
modeling
compensate
requirements
unstructured
commonly
systems
for
to
many
deal
handle
of
the
for completeness,
27-
Fuzzy modeling systems can create and analyze models that are incompletely
for
compensate
systems
inconsistent,
internally
specified,
these
ambiguous
in
their
the models
by
intelligently
or
flaws
in
assumptions that will allow the model to be analyzed.
these
assumptions
automatically
are
added
The
choosing
Along with the results,
model
the
to
causality.
and
shown
the
to
manager who can modify them for further analysis.
The development of fuzzy modeling systems might sound as if is many years
away.
It is not.
artificial
product
The technology for fuzzy modeling has been developed in the
community
intelligence
already
has
an
and
approximate
widely
becoming
is
reasoning
capability
available.
provides
that
One
some
fuzzy modeling features.
Another important development from the artificial
relevant
the
to
support
relatively
of
situations, is human expert systems.
complex
assistants
that
can
and
unstructured
decision
To date, such systems have been oriented
toward replicating expert capabilities on
has been to reorient the
intelligence community,
a
computer.
A
recent development
applications toward providing computer-based expert
support
rather
than
replace
a
decision
example, one company has developed "smart" statistical tools.
For
maker.
When a computer
IS given sales data,
for instance, the system will analyze the characteristics
of
into
the
data,
taking
account such
things as
seasonality,
and
will
then
choose the most appropriate sales forecasting tool.
The conceptual underpinnings of expert support systems have been developed
under
the
rubric
of
"model
management
systems."
According
to
El
am,
et
al
-28-
[1980], a model management system,
can be viewed as a system that dynamically constructs a
This is
decision aid in response to a particular problem.
accomplished by drawing on a knowledge base of models that
captures the technical expertise of a management scientist
plus an understanding of the basic activities involved in
a given decision-making environment...
These fuzzy modeling and expert support tools are not yet available.
is
little
wonder,
modeling
needs
problems.
Of
are
the
meaningful
of
the
therefore,
with
executive
support
complex
models.
support fails as a support mechanism for complex problems.
"real"
problems,
offer
they
But,
information
That is why many
that executive support cannot help them with their
complex
their
the
unstructured
and
because
systems
support for a manager's implicit mental
senior managers might feel
supporting
infomation support systems
other forms of support,
for
popular
most
faced
difficulty
had
have
we
executives
senior
three
that
It
problems.
For
them,
must wait
we
little
a
longer for the emergence of some "real" executive support tools.
While
we
are
waiting,
models or modeling systems.
for
planning
explicit,
and
control,
it
In
one
benefits
us
might
assume
of
challenge us
to
decision making.
uses of them.
rethink
our concept
that
think
all
problems
This paper has explored another type of model
best support managers'
to
too
narrowly
about
reading some of the literature on models used
representations
mathematical
not
-
in
important
a
models
spreadsheet
implicit models
-
a
model
and
its
role
in
format.
and how to
As support technology evolves,
of
are
it will
managerial
-29-
REFERENCES
Anthony, R.N., Planning and Control Systems: A Framework
Division of Research, Harvard
for Analysis
Graduate School of Business Administration, Boston,
Mass., 1965.
,
Bellman, R.E., Dynamic Programming Princeton University
Press, Princeton, NJ, 1957.
,
Elam, J., Henderson, J., and Miller, J., "Model Management
Systems:
An Approach to Decision Support in
Complex Organizations," Proceedings Conference on
Information Systems 1980, pp. 98-110.
,
Gorry, G.A., and Scott Morton, M.S., "A Framework for
Management Information Systems," Sloan Management
Review V. 13, N. 1, pp. 55-70, Fall 1971.
,
Grinyer, P.H. and Wooller, J., Corporate Models Today: A
New Tool for Financial Management (Second
Edition), The Institute of Chartered Accountants in
England and Wales, London, 1978.
Hitch, C.J.,
"Operations Research and National Planning - A
Dissent," Operations Research V. 5, October 1957.
,
Konsynski
B.
and Dolk, D.
"Knowledge Abstractions in Model
Management," Working Paper MIS Department,
University of Arizona, 1982.
,
,
,
Lindblom, C.E,, "The Science of 'Muddling Through'," Public
Administration Review V. 19, 1959, pp. 79-88.
,
Little, J.D.C., "Models and Managers: The Concept of a
Decision Calculus," Management Science V. 16,
April 1970, pp. B466-485.
,
Mclnnes, J.M., and Carleton, W.J., "Theory, Models and
Implementation in Financial Management,"
Management Science V. 28, N. 9, September 1982,
pp. 957-978.
,
Mintzberg, H., The Nature of Managerial Work
Row, New York, NY, 1973.
,
Harper and
Montgomery, D.B., and Urban, G.L., Management Science in
Marketing Prentice-Hall, Englewood Cliffs, NJ,
,
1969.
Naylor, T.H., and Schauland, H., "A Survey of Users of
Corporate Planning Models," Management Science,
-30V.
22,
N.
9,
May 1976, pp. 927-937.
Rockart, J.F., and Treacy, M.E., "The CEO Goes On-Line,"
Harvard Business Review V. 60, Jan-Feb 1982,
pp. 82-8 8.
,
Rockart, J. P., and Treacy, M.E., "Executive Information
Support Systems," Working Paper Center For
Information Systems Research, Sloan School of
Management, M.I.T., 1981.
,
Traekle, J.W., Cox, E.B., and Bullard, J. A., The Use of
Financial Models in Business Financial
Executives' Research Foundation, New York, 1975.
,
Will, H.,
"Model Management Systems," in Grochla and
Szyperski, Information Systems and Organizational
Structure de Gruyter, Berlin, 1975.
,
63i,Q^,;53
MIT IIBRARJES
3
T060 0056302^
1
mmi
^ -
-
/^
Lih-26-*>7
Mil LIBRARIES DIIPL
3
TOflD
1
DD5fl3D2T
1