i)ewey
HPZ8
wo.|7S'6-?fc
May 1986
GSR WP No. 140
Sloan WP No. 1786-86
90s WP No. 86-022
Center for Information Systems Research
Massachusetts
Institute of
Sloan School of
Technology
Management
77 Massachusetts Avenue
Cambridge, Massachusetts, 02139
^
The Management of Data:
Preliminary Research Results
Dale
L Goodhue
Judith A. Quillard
John
F.
Rockart
May 1986
CISRWPNo. 140
Sloan WP No. 1786-86
90s WP No. 86-022
«1986 Massachusetts
Institute of
Technology
Center For Information Systems Research
Sloan School of Management
Massachusetts Institute of Technology
THE MANAGEMENT OF DATA
PRELIMINARY RESEARCH RESULTS
INTRODUCTION
1.0
Today, corporations are placing increasing emphasis on the management of
A number
data.
of
factors,
which include a significantly more competitive
environment, more ubiquitous personal computers,
and
computer-literate
more
better
personnel,
Acknowledging
data.
this,
follow the exhortations of authors
have
more
led
to
more
and
such as
rapidly improving software,
companies
Diebold [1979],
and Appleton [1986] to "manage the data resource".
argue
information
that
achieving
and
organizational
data
play
a
performance
demand
a
and
trying
are
Edelman
to
[1981],
These, and other authors
significant
and
therefore,
and,
more
for
growing
must
be
role
in
managed
proactively.
The
available
literature,
however,
contains
little
suggest
to
that
organizations have successful track records in managing data from a business
perspective,
about
the
resource",
companies.
rather
than
characteristics
we
studied
Almost all
17
a
technical
of
viewpoint.
effective
data
of these
approaches
management
efforts
order
In
to
in
efforts were viewed as
information systems staff and user management.
to
learn
more
the
data
"managing
a
set
of
successful
diverse
by
both
Our research focused on the
managerial motivations, planning processes, and types of outputs achieved.
The authors would like to thank Judith Green Carmody, William A. Gillette,
Lisa A. Johnson, Ruth E. Kocher, John S. Lively, and Judy E. Strauss, who did
their Master's thesis research in conjunction with this study.
2-
They are:
Three major themes emerged from our study.
management of data is becoming increasingly important.
Major business and technology trends suggest that, if anything,
this importance will increase.
The
the
improving
to
approach
dominant
single,
is
no
There
multiple
adopted
firms
have
Rather,
of
data.
management
approaches, contingent upon business needs.
A set of fundamental managerial issues must be considered
selecting the appropriate approach to data management.
Section
of
rest
The
describes
1
research
the
method,
in
growing
the
importance of data management, and some current approaches to managing data
Section
in the literature.
we
Section
saw.
several
presents a framework and describes the choices
summarizes
3
managerial
key
2
issues
they consider the options
Section
framework.
the
managers
that
4
explicitly
should
address
Section
suggested by our framework.
addresses
then
5
as
concludes
the paper by emphasizing the several major themes in our results.
1
Research Approach
.1
These
being
first phase of
resulted from the
themes
conducted
the
by
Center
Information
for
data management
a
Systems
Research
study
(CISR)
at
Case studies, each involving 4 to 6 days
MIT's Sloan School of Management.
of interviews, were conducted in six large corportions during the spring of
1985.
In
addition,
CISR reseachers spent 1/2 to
3
days in each of 5 other
organizations to learn about their data management efforts.
Figure
summarized
in
industries,
including
In
1
with
disguised
electronics,
consumer
The companies,
represent
names,
goods,
a
insurance,
range
and
of
energy.
each firm, we focused on one to four data management programs or projects.
These
firms
discussed
with
us
the
data-related
policies,
processes,
controls, standards, and tools that were in place (or had been tried).
Also
discussed were the factors which motivated each organization to take action,
and the results which were achieved.
In
addition, our interviews collected
staff
professionals,
regarding the most important problems and issues concerning
the management
the
opinions
of
and use of data.
I/S
managers,
line
managers,
and
Figure
Disguised
1.
Name
Firms
in
the CISR Data
Management Study
Description
Data Management efforts
paper
cited in this
Blame Corporation
Personal care products; Fortune 500
Data access services for end users
Crockett
Canadian subsidiary of Fortune 500
U.S. computer company
Integrated database for finance
and administrative applications
and managerial reporting
Diverse Conglomerate
Fortune 500
Information database for use by
senior
Among Top
Dobbs Insurance
Foothill
Computer
10 in assets
Fortune 250
management
new
Strategic data plannmg and
Systems for one division
1.
Information centers provide
data consulting
2.
Set of standard data definitions
established by corporate task
force
3
Strategic data planning used for
order flow function
Global Products,
Inc.
Matrac Corporation
Sierra
Energy
Spectrum Electronics
Fortune 500 manufacturing
Subject area databases being
gradually implemented
Fortune 100 manufacturing
Data access services
Among
Information database
for financial reporting
top 10 energy companies
Fortune 100
Fortune 500 diversified chemical
end
users
Integrated manufacturing
dataoase
Waverly Chemicals
for
1.
company
in
one
Common
division
manufacturing
Systems largest division
2.
Common
3.
Strategic data
accounting Systems
used by multiple divisions
modeling
done
for l/S planning in
largest division
4.
Set of standard data
definitions established by
corporate group
Windsor Products
Fortune 100
1.
Several information
databases built
2.
'Data Charting" effort
-4-
1
.2
Managing Data Is Becoming Increasingly Important
A rationale
Huber'
for
characterized
that
prediction
[1984]
s
by
importance of data management
growing
the
Huber suggests,
changes,
will
require
organizations,
Thus,
These
all
at many
sources
survive, will
to
be
management of
the
in
increasing
and
organizations to have up-to-date information from multiple
levels of aggregation.
will
(author's emphasis).
complexity , and more and increasing turbulance "
environmental
more
knowledge ,
increasing
and
more
society
industrial
"Post
found
is
need improved
designs for information acquisition and distribution.
As the business need
hardware costs
in
so
has
gains
and
the benefits to be gained from timely,
the costs
have decreased.
effective
for
functionality.
Thus,
as
accurate information have increased,
illustrated in Figure
As
capability to
have been dramatic
there
software
in
technical
the
Over the last twenty years,
deliver information.
declines
increased,
has
it
2,
become cost
has
organizations to manage new types and greater quantities of
information with computer technology.
Because
electronic
rapidly
a
managing
form,
importance,
but
also
manage the data
problem
with
integrating
data
applications,
autonomous,
resource
data
were
development
accuracy differences,
of
access
accessible,
to
activities,
initiatives.
start from a clean
designed
often
to
developed
today
led
to
corporate
the
is
in
widely
needs
quality
even
data
hinders
constrains
today
often
system
the
in
attempts
to
difficulty
This
data
problems.
development
to
of
isolated
of
almost
problem of
definitions,
The
limits managerial
ability
underlying
dispersed,
in
in
only
An
slate.
the
discrepancies
not
growing
meet
and timing and coordination
information,
and
has
available
now
is
is
Realistically,
organizations
in
originally
which
effectively
data
not
do
data
of
sub-organizations within a large corporation.
decentralized
lack
that
complexity.
in
managing
amount
increasing
and
undertake
resulting
and
staff
maintenance
strategic
5-
Figure
2.
HOW MUCH INFORMATION IS IT COSTBENEFICIAL TO
"MANAGE"
cost = benefit
20 years ago
cost = benefit
today
Due to decreasing costs and increasing benefits, the sets of data that have been
computerized in most corporations have expanded from accounting data to include a
complete range of operational activities, staff functions, and line managerial activities. In
addition to numerical data, increasingly, text is being stored. Efforts are underway to
store, through "expert systems', some of the knowledge previously available only
eQnlp'<; hpaH<
That much mnrp ri;»t;i \a/iII hp '.tnrcri Anri arro^^ori piprf rnnir aIIu in fhp
m
-6-
1
Current Approaches to Data Management
,3
What means
manner which
which
have
should firms use
best
been
contributes
discussed
existing
the
in
objectives?
business
to
of managing data
achieve the goal
to
a
approaches
Various
resource
data
in
management
(Dfy^)
literature can be categorized as follows:
Approaches with a technical focus.
These include tools and
management
such
database
systems,
data
techniques
as
dictionaries [Ross, 1981], data entity relationship modeling
[Chen, 1976].
Approaches with a focus on organizational
responsibilities
These include the establishment of organizational units such as
database
administration
and
data
administration
[Tillman,
1984], [Voell, 1979], and the formulation of administrative
policies and procedures covering areas such as data ownership,
access, and security [Appleton, 1984].
.
Approaches with a focus on business-related planning.
These
include planning processes and methods such as Strategic Data
Planning [Martin, 1982] and BSP [IBM. 1981] that link the
acquisition and use of data with business objectives.
It
increasingly evident that no one category provides
is
adequate
approach.
Coulson
[1982]
acknowledges
efforts
many
that
completely
a
to
straighten out data management problems through the implementation of a data
dictionary have failed.
Kahn [1983] presents empirical
evidence suggesting
most data administration groups have had little or no success in correcting
key data management problems.
Because
the
organizational
ultimate
units,
goal
is
not
put
to
in
place
tools
but rather to link data needs with
or
to
needs
the
create
of
business, much attention has focused on the third type of approach.
planning approaches,
however,
done
well,
investments
discuss
BSP
of
some
requires
time
reasons
and
many
and
For example, Zmud [1983] notes that "to be
organizational
effort"
why
These
require significant resource commitments,
are often not easy to undertake.
the
(Chapter
large-scale,
efforts are difficult to accomplish.
10,
members
p. 281).
top-down
to
make
Section
strategic
considerable
2.2.1
data
will
planning
-7-
2.0
A CONTINGENCY APPROACH TO DATA MANAGEMENT
The
picture
management efforts
from
emerges
which
fit
single
no
our
on
diverse in business
motivation,
type of result obtained.
that
effective
are
several
Which path is selected depends
Successful
considerations.
organizational
is
Rather there
clear pattern.
"paths" to improving the management of data.
heavily
studies
case
organizational
scope,
Certainly one must be careful
efforts
are
planning method,
studies
with
a
In
seems
also
to
be
borne out
in
and
in generalizing from
less than twenty examples, but this variety (or contingency approach)
case
very
discussions which we
in our
have
had
number of other companies.
analyzing
the
cases,
we
see
four
key
areas
which
represent
interlinked set of choices that organizations make as they undertake
management effort.
a
These areas are:
In the successful
The identification of a business objective
companies in our sample, data management actions were almost
always justified not by conceptual or technical arguments, but
by one of
four compelling business
needs:
coordination,
flexibility, improved managerial information, or I/S efficiency.
.
The scope of the data management project.
The firms we studied
had explicitly defined and limited the scope of their efforts.
Some focused on a functional area (such as finance), others on
a division, while some were corporate-wide.
The data planning method.
Top-down, in-depth strategic data
In
modeling was not the only data planning process we saw.
fact, there appear to be a number of obstacles in accomplishing
a
large-scale strategic data planning effort. The planning
processes utilized In our cases varied widely in terms of their
formality, their detail, and their emphasis on data models.
The range of options varied from strategic data modeling to
more limited planning approaches to no planning whatsoever.
The "product" of the data management effort.
Much of the
implementation of
existing DRM literature centers on the
subject area databases [Martin, 1982].
In our case studies,
however, we saw five distinct "products", which were the end
results of the data management project team's work.
These
products are: subject area databases for operational systems.
an
data
-8-
common systems, information databases, data access
and architectural foundations for future systems.
Figure
from
generalized
simple
This
cases.
our
components
four
these
presents
3
shows
and
framework
services,
the
choices
a
starting
represents
point for visualizing what organizations are actually doing with respect to
The next four subsections of the paper (2.1
data management.
2.4) discuss
-
We begin with the most tangible, the
each of the elements of the framework.
"products", and then in turn discuss planning processes, scope, and business
objectives.
Five Data Management "Products"
2.1
"product"
The
area
subject
identified
databases
systems
as
most common
databases
used
products
common
These
.
development.
identified
two
by
multiple
systems ,
are
addition
In
other
the
products
three
products:
foundations for the future.
existing DRM literature
in
to
data
This
and
operational
that
in
these
"new
system"
services
access
and
,
of
also
We
information
call
each
set
a
systems.
what we will
similar
is
involves
new
products,
we
architectural
section discusses at some length each of
the products and presents case examples.
2.1.1
Subject Area Databases for Operational Systems
Subject area
databases
important
business
products,
rather
processing
or
entities
manufacturing
applications may share
SADBs.
described
In
four
as
of
SADBs
subject
individual
scheduling.
companies
we
operational
areas,
is
such
Many
In
the
such
and
order
as
operational
single
a
results
first,
around
customers
different
data from
the
organized
as
applications,
studied,
data.
which
data
(both access and update)
the
for
or
around
than
contain
(SADBs)
can
set
of
best
be
strategic
planning was the basis for the effort.
largest
Dobbs
Insurance Companies 1s one of the nation's
insurance companies.
Dobbs has five major divisions which
In the late 1970s, the Group
operate relatively autonomously.
Insurance Division
(GID)
began to address issues of
data
data
9-
m
u
o
OJ
u
c
E
(U
c
Q
O
3
E
m
01
-10-
Due to a changing business environment and
resource management.
recent reorganization, the division wanted to reevaluate its
After a significant amount of preliminary
operational systems.
work, GID undertook a strategic data planning effort based on
From the resulting strategic data
James Martin's approach.
(including customer,
model, a dozen subject area databases
asset, and product) have been implemented as the foundation for
the division's applications systems.
a
Subject
area
databases
for
operational
systems
can
also
be
designed and implemented more gradually.
Global Products, Inc., had a history of using selected reference
files for static data elements such as customers, products,
Integration of application systems was
prices, vendors, etc.
achieved by passing transaction data elements from program to
In the mid-1970s, Global
program following a predefined flow.
sequential,
applicationdecided
replace
this
Products
to
oriented approach with a subject area database approach for its
The goal was
finance, distribution and marketing functions.
achieved not by forcing rewrites of all existing systems, but by
offering new or rewritten applications access to the SADBs.
The
approach resulted in a growing set of integrated subject area
databases.
This
improved
the
accuracy
and
timeliness
of
operational data, and reduced system development and maintenance
effort.
difficulty with such an incremental approach is that each
system's data requirements may not be consistent with the
current design of the SADBs.
Where necessary, the set of
databases must be revised to incorporate the requirements of the
new system.
The advantage, however, is that the changeover can
be accomplished with minimum disruption to existing systems and
a preplanned movement toward a consistent set of databases can
occur, taking into account the resource constraints of the
Today, approximately 40-50 databases are shared
organization.
30
applications at Global Products.
by
The
new
Data management efforts began in the Consumer Products Division
of Spectrum Electronics more than ten years ago with the
creation of a centralized manufacturing database for its twelve
Pressed by Asian competition, the Consumer Products
plants.
Division, with annual sales of almost $1 billion, started by
developing and implementing a logical data model for five key
manufacturing applications.
The division then proceeded to
manufacturing
into
the
incorporate
all
remaining
systems
database, and now has subject area data for vendors, parts,
assemblies, etc.
-n-
Common Systems
2.1.2
A second type of data management "product"
or
which
databases
developed
are
applications developed by
used
multiple
by
multiple copies
discussed
the
system.
which
are
of
Common
systems.
data files
are
systems
single, most often a central, organization to be
organizational
above,
applications,
a
common
for
the operational
is
Physically,
units.
opposed
As
developed
to
shared
be
can
subject
the
to
there
area
by
one
be
or
databases
of
range
a
common systems approach tends to focus on replacing existing
a
systems in one specific application area.
Many firms already have common systems in place, often developed not for
data management purposes, but rather to ensure common procedures or to lower
example
can
although each division has its
own
I/S
A
costs.
typical
at
Blaine
and
sales
seen
be
products
Corporation
force,
where,
common order
entry and distribution systems are used by its three major divisions.
As
at
data management may
Blaine,
common systems effort.
However,
fact
in
definitions)
systems
is
will
not new,
a
be
only
a
by-product
of
a
common systems cannot be developed without
surfacing and resolving data definitional
old
be
discontinued.
disputes,
Thus,
old
since
while
the
systems
(and
concept of common
number of companies appear to be gaining major data
management benefits from implementing common systems and/or from using the
standard, sharable data in them.
The largest division of Waverly Chemicals,
Fortune
500
a
diversified chemical company, operates about a dozen plants.
Since about 1980, the division has emphasized the development of
common systems for manufacturing applications such as production
scheduling and spare parts inventory.
The objectives have been
to increase coordination among the plants and to reduce costs in
A
areas
such
as
inventory.
data
management
was
group
established as a key to the process.
Several
firms
business area
have
which
have
found
in
the
Instituted
era
of
common
fourth
systems
generation
they have derived significant value from the common data.
in
a
critical
languages,
that
12-
Based on data from its common order entry system, which spans
manufacturing and distribution divisions.
numerous
internal
Grand Distributors, Inc., has been able to provide salesmen with
complete sales information on a customer by customer basis.
total
This,
combined with an estimate of each customer's
not
vendors
tells
salesmen
only what
from all
purchases
customers are buying from Grand, but what they are buying from
This
others and, hence, where the marketing possibilities are.
would not be possible if each division had its own order entry
In fact one
systems, since the data would not be compatible.
small division does use its own order entry system, and its data
can not be included in the information provided to salesmen.
Information Databases
2.1.3
A
third
new
"product"
system
information
is
Whereas
databases.
the
first two products provide data to be used by transaction processing systems
and
monitoring
for
real-time
aggregations of data which are
information
operations,
primarily
used
for
staff
databases
analysis
and
are
line
management information.
Most
"information databases"
for managerial
information.
In
can
be
general,
defined
as
subject
area
for data management purposes,
can be distinguished from subject area databases for operational
two
primary
existing
Information
ways.
systems.
Instead,
databases
"bridges"
systems to provide the appropriate
Depending
bridging
on
the
process
information
level
may
databases
external data.
of
be
data
are
to
compatibility
straightforward
usually
contain
databases
not
do
often
the
of
or
from
existing
quite
aggregated
rewriting
data
existing
the
new subject area
the
systems in
necessitate
built
database.
systems,
difficult.
and,
they
the
Also,
sometimes,
[See Note 1.]
At
Windsor
Products,
corporate
management's
demands
for
information led to the development of several new databases.
However, existing applications have not been rewritten and the
operational databases on which the transaction systems depend
remain in place.
Automated bridges from the existing systems
populate the new "information-only" databases for customer,
product and shipment.
Sierra Energy Corporation has an information database for the
key financial information required by the corporate controller's
function.
This database is incorporated in a system called DMS
-13-
DMS was developed in the late 1970s
System).
financial
data
from divisions
of
to
flow
well
to
as
as
then,
corporate,
to
geographic sectors and,
A
generic
bridging
and
security.
increase data
integrity
mechanism to capitalize on the data collection processes in
existing systems was developed, along with download capabilities
Standardization of data
for
forms
and
report structures.
definitions controlled by master tables in a data dictionary
helped create the data integrity feature of DMS.
(Data Management
to
improve the
next
The
originally
case
illustrates
conceived
as
an
effort
an
information
where
an
database,
is
integrated
now
database,
also
used
for
transaction processing.
computer
Crockett,
Canadian subsidiary of a major U.S.
the
integrated
database
to
support
corporation, has developed a single
and
order
including
accounting
all
administrative applications,
It
within
I/S.
was
prototype
The effort began as a
entry.
drain
of
running
originally aimed at removing the operational
report programs on the transaction processing systems, by creating
a user friendly managerial information database.
The system which encompasses the database has evolved to include
data capture and transaction processing as well as to provide
A cornerstone of the new
extensive end user access to data.
system is an active data dictionary, which is used to determine
how to calculate derived data elements and, by end users, as a
directory for data definitions.
Crockett has devoted considerable
hardware resources to the system, but feels it is benefitting from
the investment.
2.1.4
Data Access Services
The first three products discussed emphasize developing new databases or
files
study,
with
pertinent,
however,
accurate,
and
focused mainly on
consistent
improving
data.
managerial
Some
firms
access
to
in
our
existing
data, without attempting to upgrade the quality or structure of the data.
Though the particulars of the efforts differ, each is centered around a
small
cadre of
personnel
whose
goal
is
to
better understand what
data
is
available in current systems and to put in place mechanisms to deliver this
data.
These efforts seem to be widely applauded by managers who are able to
"get their hands on" existing data.
questions
concerning
definitional
make use of the data provided.
They also have the promise of surfacing
inconsistencies
as
managers
attempt
to
-14-
A multi-binion dollar manufacturing firm, Matrac Corporation,
has put in place a "data warehouse" for its corporate end users.
Briefly, the data warehouse organization is a small group within
organization
that
data
access
provides
I/S
the
corporate
for
The group locates data,
arranges
services to users.
in
the
data
the
user's choice of
and delivers
extracts,
The group maintains copies of most of the data it
formats.
The next step for
provides and arranges for periodic updates.
to
data
develop a
data warehouse management team is
the
directory designed specifically to assist end users by listing
Included for each
and cross referencing the data available.
data item will be key information such as update frequencies and
times, contact people for more detailed interpretation, accuracy
characteristics, etc.
Obviously,
provision
of
key
a
both
element
human
of
delivering
assistance
and
data
to
improved
users
end
tools
for
is
the
access
and
analysis.
At Blaine Corporation, corporate I/S provides informal data
"consulting" services to help end users locate, access, and
interpret the data they need.
There are several infonmation centers at Foothill
Computer.
Although
the
centers
are
charged with
the
provision
of
analytical tools and support of end user computing, they also
devote significant time to the provision of data consulting
services.
Not surprisingly,
where
data
is
of
data access services appear more helpful
reasonable quality
than
where
Delivering data in its current form to managers,
data
is,
however,
in companies
frankly,
can
a
mess.
spur action
towards increasing data definition and control mechanisms.
2.1.5
In
Architectural Foundations for the Future
most of the firms we studied, managers focused on a limited set of
data serving a portion of the corporation.
danger
that
by
approaching
data
business unit by business unit,
management
However,
in
a
or subject area by
there clearly
function
by
is
the
function,
subject area manner,
a
company leaves itself open to real problems if, in the future, it is desired
to integrate data across these boundaries.
-15-
attempt
To
organizations
lead
will
incompatibility
problems,
architectural
foundations.
developing
some
By
foundations we mean policies and standards which, when adhered
architectural
to,
on
focused
have
future
these
avoid
to
to
a
well -structured
consistent
and
environment.
data
Managers allocating resources for architectural foundations are investing in
Our cases
the future without necessarily having immediate benefits in mind.
foundations:
emphasized two different types of architectural
wide scope
(1)
strategic data models, and (2) common data definitions.
One
data
view
model
firm's
a
serve
to
development.
a model
of
as
data
an
architecture
underlying
is
a
blueprint
strategic
corporate-wide
for
all
future
systems
James Martin's Strategic Data Modeling approach produces
as one of
its products
[Martin,
1982].
such
IBM's BSP methodology [IBM,
1981] and Holland's methodologies [Holland, 1983] are others that produce a
version of a data architecture.
strategic data model
Proponents of these approaches claim that a
provides an architectural
consistency of information,
more easily
foundation that will
integratable
systems,
lead to
improved
and
productivity in system development and maintenance.
Only one of the ten firms we
strategic
studied developed a
data model
primarily for architectural foundation purposes.
At Waverly's General Polymer Division a strategic data modeling
effort was begun for the entire division after the successful
implementation of common manufacturing systems.
The model has
been completed and will
be an input to the division's
I/S
strategic plan.
It is still
too soon to tell whether that
effort will eventually lead to benefits for Waverly.
A
second approach
definitions.
to
data
architecture
is
standardization of data
the
The choice of which data elements should have corporate-wide,
standard data definitions is an important architectural
there
are
some
data
elements which
are
so
basic
to
issue.
the
In
operation
any firm
of
the
business and which are the basis of so much shared communication, that it is
critical
same way.
for all parts of the organization to refer to these elements in the
Presumably these data elements should be given global, mandatory
definitions.
Below the corporate level,
it may make
sense for a particular
-16-
division
standardize
to
additional
data
within
just
elements,
Thus, the standardization of data definitions can be seen as
that division.
a
certain
on
hierarchical process.
Most of the new system "products" (Sections 2.1.1 to 2.1.3) developed by
firms
the
our
in
personnel
agree
elements,
as
several
a
on
was
the
required
however,
intended
that
building
to
set
a
as
line
an
management
definitions
precise
prerequisite
cases,
definitions
study
of
of
a
standardized
architectural
I/S
specific
product.
the
and
[See
technical
set
of
Note
2.]
corporate-wide
foundation
for
data
In
data
unspecified
future systems development work.
Foothill
Computer formed a task force, chaired by
In 1984,
corporate I/S, to identify and define key data elements being
The members of the task
used in multiple areas of the business.
force came from the I/S groups in the various functions and
divisions.
There were no immediate plans to implement the
agreed upon definitions.
Rather, it was assumed that future
systems development work would conform to these definitions.
In
addition, it was established that any group supplying data to
another group within the corporation would be required to
if
deliver that data in conformance with the definitions,
asked.
After coming to agreement on definitions for over 200
data elements, the task force has refocused its efforts to
concentrate only on those elements for which a specific business
impact can be identified and pursued.
The corporate
I/S
group
at
Waverly
developed
rigorous
a
As of March, 1985,
methodology for data element definition.
approximately 150 data elements had been formally defined by the
Another 500 elements
data management staff using this method.
identified as candidates for the process.
The
have
been
definitions have been documented on a data dictionary but have
not yet been incorporated into applications development work.
Thus,
model
or a
as
these examples
set of
illustrate,
standard definitions can be a product in its own right.
Their usefulness can be questioned,
to guide future
either a wide scope strategic data
however,
unless the data model
is
used
systems development, or the definitions become incorporated
in either operational
or managerial databases.
-17-
A Variety of Data Planning Processes
2.2
planning processes used to select the "products" in
We now focus on the
DRM is synonymous with a large-scale,
To many people,
the previous section.
There are, however, other less
strategic data planning and modeling effort.
This
effective.
studies
categorizes
section
four
into
approaches
planning
all -encompassing
can
equally,
be
modeling;
data
from
our
a
case
"80-20"
"targeting";
These approaches represent
more,
not
if
processes
planning
the
strategic
types:
approaches; and no explicit planning.
of
which
continuum
planning processes which assist the organization to identify the target
for
management
data
continuum
The
pursue.
relatively
and
action,
choose
action
the
well-defined
ranges
from
scope
models
broad
detailed,
to
of
methods
enterprise
the
"product")
(or
to
(producing
and
data),
its
through less formal and rigorous approaches, to no data planning at all.
Strategic Data Modeling
2.2.1
Section
discussed
2,1.5
strategic
foundation.
developing a data architectural
focus on using the approach
modeling
data
this
In
context
the
in
we
section,
of
primarily
situations where the near-term objective is
in
to build new systems or databases.
The underlying assumption
-- that it is
business
contest.
is,
of
the
strategic
planning methodologies
data
impossible to plan effectively if one does not know what the
what
it
However,
of
and
does,
what
data
companies
eleven
it
uses
studied,
we
--
is
difficult
three
only
used
to
a
strategic data modeling approach, and as yet only one of those has succeeded
in
implementing
many
that model
other firms
have
actual
in
surfaced many
Informal
systems.
disappointments with
discussions
such
with
approaches,
and few successes.
The
diagram
Figure
in
4,
adapted
from
James
Martin's
Planning Methodologies , is representative of several
planning
approaches.
planning approach,
of the diagram.
The
left
side
of
the
Strategic
Data -
data-oriented strategic
diagram
shows
the
top-down
leading to the bottom-up design shown on the right side
-18-
Figure
4.
Strategic Data Planning
.
-19-
model,
(Box
in
1
with
begins
process
The
Figure
development
the
4).
of
enterprise
The
enterprise
an
depicts
model
business
or
functional
the
areas of the firm, and the processes that are necessary to run the business.
The
next step
processes
or
is
identify corporate
to
activities,
enterprise model,
the
(Box
2).
leading
to
Data
the
entities
data
requirements
identification
and
are
of
them
link
to
thus mapped
subject
to
onto
areas
for
which databases need to be implemented, (Box 3).
only
general
In
selected portions of
enterprise model
the
Building the logical data
area databases are selected for bottom-up design.
model
the first step.
is
The data model,
(Box
5),
of detailed end user and management data views,
the
previous
subsequent
logical
As
model.
design
analysis,
of
application
in
Section
(Box
programs,
results from
synthesis
a
(Box 4), with the results of
2).
(Boxes
Database
6),
design
and
from
the
proceed
data model
mentioned
developed
future
entity
top-down
subject
and
a
strategic data model
systems,
but has
2.1.5,
Waverly
Chemicals
for use as an architectural
not as yet
implemented any
successfully
has
systems
foundation for
using
Dobbs Insurance is so far the only reasonably successful
the
data
example in
our study of this approach leading to actual systems development.
Dobbs did successfully use the top-down bottom-up approach in
its group insurance division, modifying the approach somewhat to
reduce the detail required.
Within a few years, the division
rewrote virtually all its systems to conform to the data model
it had developed.
Some departures from the model have been
necessitated on occasion to meet the requirements of short-term
deliverables.
Efforts to resolve these "nonconforming" systems
with the data model
are made at a later point in time.
While
the
strategic
data modeling
approach
resulted
in
new
systems
Dobbs, it did not at Foothill Computer.
In the late 1970s, Foothill Computer recognized data management
as an important issue.
A group, composed primarily of senior
I/S professionals, began to study and then to develop top-down,
data-oriented methods for strategic systems planning.
After
much exploration, the group developed a strategic data planning
method that integrated both data-oriented and process-oriented
views of design.
at
-20-
In 1982, the method was used to analyze one major functional
The project team
area, the flow of orders through the company.
worked for six months to define and get consensus from the
They also
worldwide business units on a logical data model.
Logical
reached agreement on data elements and definitions.
subject area databases were designed, and much effort was spent
The
systems,
validating those databases with user views.
A host of problems led to
however, were never implemented.
These included the difficulty of
discontinuing the project.
consolidating disparate user views, doubts about whether all the
key user views had been obtained, and not least of all, a
multi-million dollar price tag for implementation.
The course
of action finally chosen was to make extensive modifications to
an existing system to address its shortcomings.
There
studied.
were
In
14
none
other
data
these
of
management
did
the
efforts
planning
the
in
and
eleven
firms
implementation
process
proceed as suggested by the strategic data modeling methodology.
we saw a variety of other approaches used successfully.
we
Instead,
Figure 4, discussed
Figure
previously, depicts a generalized strategic data planning process.
5
illustrates the actual planning process in four of the case study sites.
shown in Figure 5a and 5b, both Spectrum Electronics and
Waverly Chemical chose to develop manufacturing systems without
having used any top-down, data-oriented planning methodology to
arrive at that decision.
They started with a particular set of
applications (1), then developed a logical data model (2), and
finally designed the physical database and the applications (3).
As
Sierra Energy, (Figure 5c), started with a particular user's
view (l)--that of the corporate controller—and designed a data
model (2) and physical data bases (3) from that perspective.
Methods for "bridging" data from existing applications were then
developed (4).
At Windsor Products, (Figure 5d), the data management group in
corporate I/S consulted extensively with the distributed I/S
groups and with senior management to reach a consensus on what
new corporate-wide databases would have the greatest impact on
the business (1).
The data management group then consulted with
all users of that data as it built the logical data models for
those databases (2).
They essentially replaced the entire left
side of the diagram by wide-ranging deliberation within the
distributed I/S organization, then proceeded to build databases
and "bridges" (3).
In all
side"
or
of these cases, the firms either skipped or abbreviated the "left
top
down
portion
of
the
top-down
planning
bottom-up
design
21-
Figure
5.
The Planning Reality
Figure 5a.
Figure 5b.
SPECTRUM
WAVERLY
(General Ploymer Div.)
PHYSICAL
APPLICATIONS
DATA MODEL
USER VIEWS
USER VIEWS
SIERRA
PHYSICAL
APPLICATIONS
DATA MODEL
Figure 5c.
DATABASE
PHYSICAL
DATABASE
DATABASE
-22-
process.
In
fact,
they
followed
discussed in Sections 2.2.2
-
alternate
the
planning
processes
to
be
2,2.4.
Some companies, like Foothill, that do follow a top-down, strategic data
planning
process
reflection
more
there
firms
are
have
significant
appear to
not
be
difficulty
number of
a
successfully
using
implementing
reasons which
strategic
data
the
help
plan.
Upon
explain why
planning
methods.
Among them are:
The strategic data planning process, done in detail and with
wide scope, can be very time consuming and expensive.
Also, for the process to be successful, key managers must
This commitment is often
commit significant time and effort.
difficult to obtain (and keep) from these busy individuals.
a
It is not always clear to the planners or top management
whether strategic data modeling is being done to create an
It
architecture or to immediately build new systems.
is
difficult to manage the expectations of those
involved
regarding the results and benefits of the process.
Total implementation of a wide scope data modeling effort can
be extremely expensive.
There is a tendency to avoid these
new costs, especially if many of the existing systems which
represent a huge investment are still effective.
When implementing only a subset of the plan, it can be
difficult to bridge the gap from the top-down plan to
bottom-up design.
If proposed and existing systems interface
along different boundaries, it may be hard to isolate and
replace a subset of existing systems with a subset of
The use of application packages also
proposed systems.
creates interface and boundary problems.
Because of the upfront effort needed, organizations face a
longer and more expensive development process for the initial
Line managers
systems developed with this planning method.
do not like to see project schedules lengthened.
Similarly,
I/S managers, who have incentives to deliver quickly and
contain costs, may resist the additional effort involved.
Often the business will
developed and implemented.
change
while
the
plan
is
being
The methodology requires new I/S skills.
Thus,
process
for all
and
these reasons,
manage
expectations
it is
difficult to gain commitment to the
regarding
the
outcome
of
strategic
data
-23-
planning.
We
will
discusses several
return
these
of
some
to
items
Section
in
which
4.0
issues affecting the implementation
fundamental managerial
of data management efforts in general.
Targeting High Impact Areas
2.2.2
Most
corporations
that
abbreviate
or
skip
top-down
the
planning
There are a variety of alternative
methodology do not act without a plan.
planning processes which can be used.
The most common process
other
business
area.
the
is
a
particular function or
problem
important
companies
some
In
"targeting" of
opportunity
or
areas are quite clear without an extensive analysis.
Sierra Energy, the controller knew from his experience that
needed, (and did not have), consistent, accurate data for
At Windsor Products, it was quite
corporate reporting purposes.
clear that consistent and accurate customer data and product
data needed to be shared to support changing business demands
and, after some discussion, it became clear that shipment data
At Spectrum Electronics, the head
was also critically needed.
of the I/S department was certain, and was able to convince the
manufacturing vice president, that an integrated manufacturing
database and common systems were the answer to coordination
problems and high systems costs.
In
^
he
none
In
But,
improved
data
quality.
In
rigorous
a
key managers
there were
all,
in
these companies was
of
each
data
who could
case,
visualize
data
a
method
planning
the
management
used.
benefits
program
of
was
undertaken which was limited in scope, but was effective and implementable.
"80/20" Planning Methods
2.2.3
In
necessary
in
global
having
these
cases
is
(bottom-up),
planning
invest
to
full-scale
carry out a
to
implemented
there is a desire to get the major benefits of global
without
planning
data
aim
firms,
some
to
zero
while
(top-down)
in
strategic
quickly
reducing
phase.
amounts
the
This
the
type
on
data
the
amount
of
of
time
planning
key
of
and
money
The
process.
"products"
effort
spent
to
in
be
a
planning can be termed an
-24-
"80/20" approach, after the adage that for many undertakings,
80 percent of
the benefits can be achieved with 20 percent of the total work.
Windsor Products, having carried out its first round of data
management in a quick targeted manner (as shown in Figure 5d),
found itself uncertain as to the next subject databases to
Management decided, however, that a full strategic
implement.
What it needed
data model was not necessary for its purposes.
from a planning process was a means to fairly rapidly identify
It
the next round of candidates for subject area databases.
which
planning
approach
abbreviated
developed
its
own
therefore
This involves identification of the
it calls "Data Charting".
major data aggregates in the corporation and the groups which
The use of this method took less than six man months
used them.
of effort, and involved reviews by line managers in all parts of
the business. This "Data Chart" is serving as the basis of
planning for the next round of information databases.
A
different
type
"80/20"
of
approach
was
used
to
identify
key
data
elements that should have common corporate-wide data definitions at Foothill
Computer.
In
Foothill
1984,
put priority on developing common
data
definitions to promote data consistency, as described in Section
A Key Data Task Force was formed to identify and resolve
2.1.5.
impact
The method used was to
high
data definitional problems.
I/S
ask distributed
groups to each nominate 50 important, cross
functional data elements.
There was a 30 to 40 percent overlap
among the twelve lists of data elements nominated.
Work then
proceeded on the "product" of developing common definitions for
the identified critical data elements.
The problem with strategic data planning approaches is the investment of
time and dollars needed to obtain results.
probable
the
inconsistencies
that
will
Drawbacks of "targeting" include
arise
from
multiple
targeted
projects, and the fact that, in some companies, the most appropriate targets
may not be evident.
An
"80/20" approach, while not providing the detail
of
strategic data planning or the quick hit of targeted approaches, does offer
the major benefits of both previous approaches.
2.2.4
No Planning Process
There are also data management actions that can be taken without doing
any data-oriented planning.
For example,
if a
decision is made to provide
25-
better access to existing data,
that data,
then
without addressing changes
data planning methodology
no
is
needed.
form of
the
in
the case
This was
at Matrac with its data warehouse approach, described in Section 2.1,4.
Bounded Scope
2.3
our
In
corporation;
all
attempted
manage
data
used
by
the
limited the focus of the effort in one or more ways.
An
study,
firm
no
to
the
all
important factor in the success of data management efforts is that the scope
of
effort be carefully
the
which
defines
part
selected.
organization
the
of
As
described
to
is
in
this
section,
included
be
the
in
scope
data
management effort.
focused
Most firms
which
on
functional
a
targeted manufacturing
finance
corporate
targeted
management efforts
was
Section
(see
Section
(see
divisional
An
.
such
as
Spectrum Electronics
2.1.1),
or
Sierra
The
scope
of
some
case
of
the
area,
2.1.3).
example
the
is
Energy
which
data
Group
Insurance Division of Dobbs, described in Section 2.1.1.
In
some of our cases,
However,
small
in
set
all
of
but
two
standard
of
selection of corporate -wide
data.
staff.
small
In
within
these
instances,
definitions.
data
planning effort was corporate
scope of the
the
subject areas:
the
product was
Windsor
Products
customer,
relatively
focused
product,
and
on
a
shipment
Matrac focused on making data available to the corporate headquarters
Thus,
these
while corporate
efforts,
in
were
scope,
limited
to
a
subset of the total data used by the corporation.
addition
a
to
functions
corporation.
and
Among
districts, and product lines.
divisions,
these
are:
other
suborganizations
groups,
show,
the
scope
of
the
sectors,
exist
geographic
Some organizations may choose one or another
of these units as a locus for data management efforts.
will
a
.
data
determined by the business objective.
management
As the next section
effort
is
substantially
26-
Business Objectives, Not Conceptual Justifications
2,4
The proponents of data management far too often base their arguments on
either
data-centered
underlying
rationale
of
soundness
conceptual
the
viewing
data
systems
design.
as
a
or
resource
assert
They
the
that
processes change while data is relatively stable, and therefore data should
the
be
element
key
in
management is essential
the individual
pieces.
not fit together,
will
I/S
because one
in
Or,
needs
a
they
argue
global
plan
that
developing
before
While these arguments are appealing,
pragmatic,
a
time
be entrenched inconsistencies
and the result will
the
data
global
Without such a plan, the pieces built one at
myopically designed systems.
not engender action
planning.
they
in
do
cost-conscious world of the business
manager.
The
for
successful
most
the
exploiting
problems
an
and
part,
data management processes we
at
solving
clear
a
opportunity.
This
opportunities
which
section
and
observed
motivate
been
the
executives
kinds
to
aimed,
problem
business
specific
discusses
have
of
or
business
consider
more
intensive management of the data resource.
2.4.1
Coordination within or across Business Units
A major motivation for data management action is the perceived need for
better
coordination
of
activities,
business units, or across them.
either
within
specific
terms this means more than the rapid sharing of data.
shared usefully,
it must
be
or
Improved coordination requires an enhanced
ability to communicate within or between organizational units.
be
functions
in
a
In
In
practical
order for data to
language understood by
all
involved.
This implies common data definitions and common codes for key data elements.
There are a number of clear examples of coordination as a motivation in
our case studies.
Both Spectrum Electronics and Waverly Chemical felt the need to
standardize their manufacturing systems so that many plants
could be coordinated more effectively.
In both these cases.
-27-
Better
led
to
significant
benefits.
have
efforts
their
production
of
separate
plants
schedules
the
coordination between
has led to reduced in-process and inter-plant inventories; the
coordination of spare parts inventories has led to smaller
inventories and reduced downtimes; and coordinated purchasing
of
and
brought
economies
scale
special
operations
has
arrangements with vendors.
Sierra Energy is a different type of example.
In order to meet
needed
external
reporting
requirements,
Sierra
to
better
coordinate the financial information being fed to corporate from
Again the need was for a
diverse decentralized divisions.
cotmon "language" so that critical information could be rapidly
and accurately available to the headquarters units.
2.4.2
Organizational Flexibility
A second category of motivation
data management
is
desire
the
flexibility to allow either an internal
greater organizational
of the organization,
for
for
restructuring
or a refocusing of the organization due to changes in
the environment.
Waverly Chemical merged two large manufacturing divisions in the
mid 1970s, and faced major problems with accounting systems
which had been designed and implemented separately in the
The company faced similar problems in the
original divisions.
when
it
again
reorganized, this time combining five
late 1970s
It was quite clear to senior
old divisions into two new ones.
management that there would be other reorganizations in future
As long as each division had its own accounting systems
years.
with much incompatible data, the problems would persist.
In
faced
regards
with
to
refocusing,
important
changes
a
in
number of companies
the
marketplace
we
which
studied
have
have
put
been
intense
competitive pressures on them to change from a product focus to a market or
customer focus.
At Dobbs, each product line in the Group Insurance Division sold
its
directly
to
customers
and
maintained
own
customer
information.
Each product line had its own set of customer
As the competitive environment changed, the inability to
codes.
present a single face to a customer interested in many types of
services put Dobbs at a disadvantage.
-28-
flexibility
Organizational
have
designed to
been
to
be
able
to
structures
data
by
support particular applications
not flexible enough
but which are
hindered
often
is
which
suborganizations,
or
strategically
new
provide
important "views" of the business.
2.4.3
Improved Information for Managers
A third motivation for more effective data management is
and
improved information for senior managers, middle management,
These information consumers want two things:
personnel.
for
need
the
key
staff
improved access to
data and improved data quality.
Top management at Windsor asked for a list of their fifty
They were told that because the data was
largest customers.
organized around product lines with different customer codes
within each product line, it was wery difficult to respond to
their request.
At one bank, the quality of computerized management information
such items as loan position and bankers acceptances was so
poor that managers
relied on manually prepared,
redundant
reports.
Since it took almost three months before reported
errors were corrected, some managers stopped notifying systems
personnel about errors and only used the manual reports.
on
When
problems
such
these
as
become
important
enough
to
management,
strong motivations arise to improve data quality and access.
2.4.4
I/S Efficiency
As backlogs
adopt
methods
productivity
increase, and as
and
or
tools
reducing
that
I/S costs climb,
show
maintenance
considered to be a way of addressing
standard
avoid
and
having
definitions.
stabilized
to
data
constantly
Reduced data
promise
environment,
"reinvent
redundancy
the
increasing
of
Improved
costs.
both
there is a clear need to
of
these
should also
of
management
problems.
programmers
wheel"
data
should
data
trim costs.
development
In
be
a
more
able
structure
is
to
and
Presumably,
reduced inconsistencies will make it easier to access information for either
new applications or new reports.
-29-
few of the cases we
In
for
taking action.
the
To
efficiency
primary motivation
studied was
I/S
contrary,
number of firms felt that the
a
a
management actions they were undertaking would involve
greater,
not
data
lower,
I/S costs in the short run.
Some
did
firms
improvements
for
look
in
efficiency
I/S
as
secondary
a
benefit, and some have achieved it.
Spectrum eliminated separate programs and programming staffs in
data processing and using a
its plants by centralizing all
centralized database and common software for all manufacturing
The company also claims to have reduced maintenance
systems.
costs by eliminating not just redundant data, but the programs
that updated redundant data, and the programmers who maintained
those programs.
Crockett claims a 40% reduction in development costs through use
of its active data dictionary, the refocusing of the development
process on eliciting business rules rather than programming
procedures, and user generation of their own reports.
2.4.5
Summary of Motivations
What
linked
these
all
the
to
None
factors.
motivations
them
in
reflect
They
business.
of
have
draws
on
the
conrion
competitive,
conceptual
should be considered a corporate resource.
our research are taking steps
to
that
is
The
they
managerial,
justification
companies
better manage data
tightly
are
and
cost
that
data
participating in
for concrete
business
reasons.
In
a
few cases,
these motivations to better manage data emerged
line manager's need to
departments,
successful
however,
from a
respond to changes in his or her environment.
have
also
efforts we have seen.
had
a
key
role
in
initiating many
of
I/S
the
Many line-sponsored efforts have occurred
after months, or even years, of education by I/S managers.
-30-
FOUR CRITICAL COMPONENTS FOR DATA MANAGEMENT: SUMMARIZING THE CASES
3,0
The
four
previously
process,
components
critical
Figure
in
These
"product".
and
business
are:
3,
discussed
choices
the
framework
for
provide
each
in
area
a
rough
corporation.
In
choices
actual
data
planning
framework
Section
In
for
we
2,
this
section
made
by
the
case
the
A few evident patterns emerge.
study firms.
Figure 6 shows the filled in framework.
under the
appears
a
presented
action,
scope,
separately.
together with the
presented
is
objectives,
components
thinking about data management actions
management
data
for
alternative
used
it
in
each column the firm's name
In
major data management effort.
a
Where a firm had more than one major effort, its name appears several
times,
Refer back to Figure
with a subscript to distinguish between the efforts.
1
for capsule descriptions of each effort.
observations are evident from the figure.
Several
from
a
coordination
flexibility
or
planning process, to a functional
reflects
been
existence
the
able
to
see
of
business
objective,
to
targeted
a
The path
scope, to a variety of products .
managers
positive
a
there is one
This path, with 4 companies,
especially prominent path through the diagram.
is
First,
in
specific
benefit/cost
functional
ratio
who
areas
their
in
have
area
of
responsibility.
Second, another path that stands out (with 3 companies involved) is from
an objective of improved management information , with various scopes,
planning
process ,
to
a
"product"
improved
of
data
This
access .
to
is
no
an
important option, with real benefits at minimal cost.
In
addition to these two prominent paths. Figure 6 shows a rich array of
guided predominantly by the particular business
choices,
firm,
several
and
its
unique
appropriate
situation.
planning
Strategic
processes.
data
objective
modeling
Likewise,
subject
is
only
area
files
provide
useful,
usable
data;
information
one
the
of
databases
are not the only "product" through which benefits can be delivered.
application
of
Common
databases
-31-
1/1
E
<u
o
a
—
V/1
4>
i/l
a;
_
f7
O
O
t^ w«
*
c
—
> t
— c
q;
•
c
Q
ru
Q.
E
o
u
c
E
03
lo
$ $
CO
Q
<
w
a»
a.
O
«/^
e
i
i
2 o g o
UO
u.
O a
t/i
c
o
E
E
o
u
> >
5$
32-
"bridged"
application
existing
from
contain
systems
data
useful
for
managers; and even the improvment of data access services can have important
benefits.
There
is
path
contingent
are
improving
to
options
Multiple
corporation.
organization
single
no
correct
The
exist.
upon
management
the
readiness
the
data
of
choices
a
any
for
of
needs
and
in
that
organization at a particular point in time.
FUNDAMENTAL MANAGERIAL ISSUES AFFECTING DATA MANAGEMENT IMPLEMENTATION
4.0
Our interviews with I/S and line managers, and our own reflection on the
concerns
seem
to
five
fundamental
specific
problems
raised,
have
suggested
underlie
many
of
the
effectively implement data management efforts.
managerial
affecting
which
issues
ability
the
These are (1)
to
the trade-off
between short-term deliverables versus long-term plans or architectures;
the tendency to centralization inherent in data management;
new
organizational
roles
responsibilities;
and
the need for
(3)
impact
the
(4)
(2)
of
data
management on I/S culture; and (5) the process of effective introduction of
innovations
into
thoughts
these
on
organization.
an
issues
as
In
means
a
this
section
laying
of
the
we
present
groundwork
early
our
for
future
research.
4.1
The Trade-off between Short-term Deliverables versus Long-term
Architectural Foundations
Managers
trade-off
considering
between
foundations.
The
data
short-term
development
management
actions
deliverables
of
a
data
and
today
architecture
and
often
diffuse.
A
fact
short-term focus of line managers.
architecture
are
in
demonstrable results,
direct
of
life
in
is
with
a
architectural
long-term
infrastructure decision, like the building of roads.
term
faced
are
in
essence
an
The benefits are long
American
business
is
the
Decisions to allocate resources to data
conflict
with
managers'
focus
on
near-term earnings per share, and this year's
quickly
return
-33-
efforts reported in
Not surprisingly, most of the successful
on investment.
this paper have tended toward the short term end of the spectrum.
The danger of limiting the scope of data management action to areas with
a
near-term payoff is that this may result in problems later on if the local
ultimately
solutions
end
"architecture"
Efforts
to
strategic
prevent
usually
are
also
corporate-wide
developing
by
many
have
approach
to
and
the
to
problems.
its
has
expensive,
wery
Moving
other.
each
however,
inconsistencies
data
models
data
continuum,
the
of
with
incompatible
prove
not
succeeded.
What
is
needed
is
a
extensive,
less
expensive
less
architecture than strategic data modeling.
In
order to develop such
data
new
a
approach, we need to understand better what aspects of current data design
conflict
might
efforts
in
future
with
constructive,
what
and
needs,
minimally
principles
restrictive
ways.
can
The
planning effort at Windsor Products seems a very useful
move
current
guide
"80/20"
in
data
right
the
direction.
4.2
The Centralizing Tendency of Data Management
Underlying
organization
greater
any
effort
reality
the
is
toward
centralization
of
more
that
decision
effective
improved
data
in
an
lead
to
standardization
of
management
data
Increased
making.
management
can
data does facilitate increased central control.
For
example,
standard
systems are developed.
executives,
If
data
the
definitions
may
be
on
this
instituted
as
common
resulting data is made accessible to senior
they may have a vastly enhanced ability to compare operational
details of business units under their jurisdiction.
act
established
data.
to
In
fact,
facilitate
several
"central"
of
the
There is a tendency to
systems
coordination
and
described
control.
here
Even
increased centralization is not a design objective, it may be a result.
were
if
34-
For
Data management thus presents corporations with a familiar dilemma.
a
particular business unit
unique
its
in
environment,
there
is
make explicit choices concerning the balance of centralization
flexibility that will
greater,
best serve the business as
importance is the recognition of the
to the implementation of actions which
if not
resistance
almost inevitable
shift real
to
versus local
Of equal,
whole.
a
need
a
or perceived power in the
organization [Markus, 1983].
4.3
Responsibilities
New Organizational
Data
management
organizations.
established
It
is
is
changing
old
apparent
that
creating
responsibilities
new
units
organizational
new
data-related
various
handle
to
and
activities.
in
being
are
addition,
In
policies are being set that describe the responsibilities of personnel with
respect
areas
to
such
as
collection,
data
access,
and
security.
Perhaps
most significantly, the new responsibilities have an impact on not only the
I/S department, but also the roles of line managers.
In
addition
which
groups,
companies
some
access
groups".
sometimes,
as
an
administration
database
highlighted
are
1985],
responsible
to
A
for
few
the
firms
part
had
development
architectural
important
study
our
in
the
in
the
and
and
planning
DRM
data
end
"data
groups,
guidelines,
plans,
supporting
[Gillenson,
groups"
The emergence of
organization
administration
1981]
DRM
corporate
corporate
data
[Ross,
"corporate
created
foundations.
of
literature
had
of
groups
access
and,
groups
computing
user
deserves to be emphasized.
For the end user,
assistance with hardware and
software
Increasingly,
user
is
consulting.
not
enough.
end
organizational
responsibilities
redesign
must
also
effort.
be
of data owners,
data suppliers,
To
assigned
administrative policies and procedures
clarified.
includes
data
(See Section 2.1.4.)
Establishing the appropriate organizational
the
support
units
achieve
for
just one
effective
establishing
related to data
data custodians,
is
part of
implementation,
and
quality.
executing
The
roles
and data users need to be
-35-
administrative
the
clear assignment
of
how
example
of
from
Diverse
data
quality
comes
responsibility
helped
Conglomerate, where an extensive information database used by the
chief executive was fed by many operational systems from many
The quality and consistency of the CEO's database
divisions.
depended upon coordinating the update data from the divisions so
that the database reflected a view of the corporation at a single
In order to manage what was becoming a difficult
point in time.
problem, the group in charge of the CEO's database developed a
"master data input calendar" for critical data files coming from
The MDIC specified the schedule of updates from
the divisions.
each division, and the person responsible for data arriving on
Thus it formalized the operational coordination problem.
time.
One
The Key Data Task Force at Foothill Computer, mentioned in Section
definitions
for
some
200
data
standard
2.1.5,
has
issued
elements.
As part of the process, the task force identified a
The "controller" of a data
"controller" for each data element.
element is responsible for its definition, but not necessarily for
The
the acquisition, updating, or accuracy of the physical data.
Not all of the
typical "controller" is a senior business manager.
readily
took
on
the
data
controllers
managers
selected as
responsibility.
At this point in time,
the organizational
to data management are still
the
cases,
these
fledgling
Significant thought must be given to
evolving.
establishment of appropriate
groups
roles
must
responsibilities with respect
be
and
organizational
given
strong
units.
sponsorship,
In
many
and
be
protected from the vicissitudes of corporate politics.
4.4
Impact on the I/S Culture
A fourth managerial
culture.
DRM
implies
issue is
the
impact of data management on
the
I/S
significant changes in the ways systems planning and
design are carried out.
As Durrell
[1985] points out,
"data administration
challenges the basic process-oriented approach that has been employed during
the
last
20 years.
This
can
be
disconcerting
and
sometimes
insulting
many long-time DP professionals" (p. ID/29).
The new data-centered system used at Crockett (Section 2.1.3),
has had a major impact on the work of programmers and analysts.
The major activity now of these people is working with business
managers to define the business rules governing the meaning of
data
elements,
and
entering
data
those
rules
into
the
dictionary.
There is really no such thing as an applications
to
.
-36-
Not surprisingly, this has had
programmer at Crockett anymore.
significant impact on the attitudes and turnover of the
A great many programmers found
programmer/analyst staff there.
their skills of no value in the new environment, and many left
While the level of excitement of
during the transition period.
those who remained is high, Crockett personnel admit that they
have more luck hiring recent graduates than they do bringing in
experienced programmers to fill the new applications development
a
role.
problem
The
professionals
new
support
move
the
incentive
not
is
skills.
schemes
one
of
There
must
also
data-oriented
toward
programmers
reward
teaching
only
for
information
changes
organizational
be
example,
For
design.
implementing
systems
systems
and not for conforming to, or petitioning for changes to,
to
current
schedule
on
the data model
or
Without changes to these incentive structures, when system
data standards.
deadline
pressures
develop
their
become
own
programmers
high,
local
data
have
structures
a
tendency
strong
into
enter
that
rather
time-consuming negotiations with the data administrator for changes
to
to
the
corporate data model
The
cultural
changes
organizational
and
data-oriented approach to
systems
implied
development are
adopting
by
important to
a
anticipate.
The incentive structure appropriate for programmers and the skills needed by
them are
two
obvious
areas.
Other important changes will
probably become
apparent as organizations gain more experience in this area.
of
new
methods
and
tools
reflecting
approach to systems development will
a
major
change
not be successful
to
The adoption
traditional
the
if the organizational
change is not carefully managed.
4.5
The Process of Effectively Introducing Innovations into the Organization
The
viewed
implementation of
as
the
process
of
initial
data management
effectively
methods or tools of unproven value,
that
in
general
the
diffusion
of
introducing
innovations
is
can
usefully
be
innovations,
into the organization.
things, on five characteristics of the innovation.
(1)
efforts
new
i.e.,
Rogers suggests
dependent,
among
[Rogers, 1962]
other
These are
the relative advantage of the innovation over its alternatives;
(2)
the
•37-
results;
observabil ity of the
the compatibility of
(3)
existing values, past experience, and perceived needs;
the
and
innovation;
trialability,
its
(5)
innovation can be experimented with on
the complexity of
(4)
extent
advantage
most
of
practice is not known.
results
observe
to
First of all
compared
actions
management
data
are
available,
it
the
current
to
there have been few
in most organizations,
To date,
Where results
.
the
scale, low risk basis.
small
to implement data management actions in many corporations.
relative
which
to
argument can help explain why it has been difficult
theoretical
Rogers'
a
the
or
innovation with
the
extremely
is
hard
to
separate, and quantify, the impact of data management from other related (or
unrelated) actions.
As
noted
Section
in
4.4,
data-focused
a
approach
I/S
to
not
is
compatible with the existing appli cation/procedure/project focus, where the
is to build
goal
create
a
architecture
data
data management
applications
functions,
involves
to
have
meet current
great deal
down
and organizational
efforts.
financial
a
torn
are
management actions
risk
systems to specification on time, rather than to
individual
DRM
complexity
units are examined.
has
future
presented as
been
sold
as
the
Finally,
between
systems,
yery often data
low
scale,
small
basis
the
walls
between
trial able ,
on
Certainly,
needs.
interrelationships
and
not been
Instead,
of
and
that
major
a
investment and top to bottom commitment in the organization will
be needed to achieve results.
If one does adopt Roger's perspective on the diffusion of innovation,
is
in
easy to understand the difficulties that organizations
implementing
large-scale
data
management
have
This
efforts.
it
encountered
perspective
argues for more limited approaches where the relative advantage is clearer,
the impacts more observable, and where complexity is lessened.
limited approaches can be viewed as trials or experiments.
trial
is
ambitious
successful,
the
second trial,
infrastructure".
organization
will
probably
be
and ultimately for significant
In
addition,
When the first
ready
for
investment in
a
more
"data
-38-
CONCLUSION
5.0
Three major themes emerge from our study.
that managing
research
The
ability
data
firms
to
reorganize,
to
significant
a
access
of
operations,
is
or
managerial ly
change
to
First,
issue
it
corporations
in
relevant
their
clear from our
is
data,
to
strategic
today.
coordinate
can
focus
be
severely limited by poorly managed data.
Second,
there
one,
no
is
clear-cut approach to managing data.
range of options exist that can
depend
heavily
scope.
As
particular
the
on
needs
the
of
data
in
management
and,
and
objective
business
demonstrated by our case examples,
considered
be
to
a
particular
The appropriate planning process to use and "product" to deliver
business.
to
tailored
be
A wide
organizational
there are many contingencies
multiple
thus,
paths
to
take.
Another view of these contingencies is that data management actions can have
an impact on three aspects of a corporation's data:
its infrastructure,
its
content, or its delivery.
By
infrastructure we mean design or definitional
current
flexibility
local
in
favor
Enforced definitional
future.
of
greater
coding
and
standards which limit
flexibility
global
standards
are
one
the
in
example.
The
development of common systems, with their implicit enforcement of a single
set
of
which
definitions,
data
can
contribute
infrastructure
tend
to
to
is
a
be
another
many
the
infrastructure.
data
both
of
difficult
and
data
management
Actions
expensive,
to
build
and
the
actions
a
data
benefits
from such actions are most often realized only in the longer term.
Content
refers
to
the
choice
of
what
data
policies that address the accuracy of that data.
to
maintain,
which
affect
data
content.
These
also
to
Systems to capture new or
more detailed data, and decisions to purchase external
actions
and
efforts
data are examples of
tend
to
be
moderately
expensive, with benefits in the middle term.
Del
it.
i
very refers to making existing data available to managers who need
Data consulting services, extract policies, and the provision of fourth
39-
of mechanisms
reporting tools are examples
generation
improve
to
delivery.
Actions in this area tend to be less expensive, and have short term benefits.
The choice of where a firm should best allocate its resources
infrastructure, content, and delivery
between
-
depends very much on the willingness
-
Current
and ability of senior management to invest now for future benefits.
systems,
resource constraints,
management style,
strategic
vision,
and
the
evolution of the industry are all important factors.
The
third
theme
is
number of managerial
order
to
affinity
and
successfully
for
management,
that
no
matter which
organizational
implement
short-term
data
results,
path
is
issues which must
management
the need for new organizational
efforts.
Treating
introducing
practical
tendency
a
in
managerial
of
data
and
present issues that, if
severely inhibit the effectiveness of data management
initial
innovation
The
roles and responsibilities,
the impact of data management on the I/S culture all
not managed well, will
are
addressed
be
efforts.
centralizing
the
there
chosen,
can
forays
help
in their organizational
into
data
managers
environment.
management
understand
as
which
a
process
efforts
of
are
-40-
NOTES
NOTE
1 .
The two diagrams in Figure 7 show alternate visions of the relationship
In
the first
between operational databases and information databases.
diagram, transaction processing (TP) systems are segregated from end user
information databases to allow better tuning of the high volume TP
applications and databases.
Only selected , and usually aggregated , data is
information
downloaded to the
database, making the latter smaller and more
easily tuned to the very different load and demand pattern of end user
Thus, the TP database has complete data, but only for a limited
queries.
The information database has only selected and aggregated
time horizon.
data, but for a longer time horizon.
This two block model requires either that we know
detailed historical data managers will be interested in, or
to the information database all detailed data purged from
or that we accept that some future questions will have to go
advance what
that we download
the TP database,
unanswered.
in
The three block model, pictured in the second diagram in Figure 7,
the
recognizes
that the
questions of
interest to managers
querying
information database will change over time, in ways that cannot be predicted
in advance.
Therefore, it adds the historical database which is intended to
carry al 1 data about company transactions at the lowest level of detail
This is obviously a large database but one with very lenient
available.
efficiency demands.
Extract programs to feed the information database can
change as needs change, and if it is necessary to view data in a new
configuration, both current and historical data can be put into that form.
An extreme example of problems with historical data comes from
Southern
Cross
Products,
which
consumer
and
gift
sells
products.
There is currently a major push at Southern Cross to
develop
line
better product
sales
forecasts
by
including
Thus,
marketing
economic
indicators
in
the
anlysis.
the
analysts want data on past sales to as far back as 1950.
Unfortunately, the data processing group maintains data for one
year on tape, then discards it.
The solution has been to locate
"old timers" who have been with the company for years and might
have old printouts of sales data hanging around on a shelf
The data which the marketing analysts have found in
somewhere.
this manner has been rekeyed to create a patchwork database for
the forecasting analysis.
A
more typical example comes from Boston Mutual.
In
this
insurance company, a front-end policy preparation system records
all agents who share in the credit (and compensation) for any
force,
When the policy actually takes
particular policy.
however, the preparation information passes to an information
-41
Figure
7.
Conceptual Views of Operational and
Informational Data
Applications Applications
i
Applications
s
42-
Recent
database which carries only the first agent listed.
forced
to
dock
Boston
begin
Mutual to
business pressures have
agents for policies which are not maintained beyond the first
year, but the data to do this accurately is not available.
The two models in Figure 7 illustrate an important architectural choice
Depending on the business demand for data and on
that firms must make.
firms may make different choices for different
technical
constraints,
For instance, detailed sales-related data may be
categories of data.
captured using the three block model with an historical database because of
the likelihood of having to reaggregate in different forms as marketing
Detailed manufacturing
data
requiring
time
change.
real
structures
historical
significance,
might
purged
from
accuracy, but with no presumed
be
it
into
without
loading
the
historical
database
the transaction processing
database.
NOTE 2
.
Establishing data definitions was important in many of our
definitions and codes are key standards for a consistent data
The process of resolving specific data definitions, however,
While as many as 50 percent of the data elements in a given
defined without any disagreement, those that remain may be yery
resolve.
cases.
Data
environment.
is not easy.
area may be
difficult to
As mentioned in the section on common systems, data management
began at Waverly' s General Polymer Division as part of an effort
develop
to
common
manufacturing
systems.
Today,
the
standardization of data definitions is viewed as central to GPD'
data management program.
The cornerstone of the program is an
internally-developed data dictionary, with over 5,000 total and
500 standardized data elements. A committee composed of the data
administrator,
systems
analysts,
and
user
representatives
facilitates the definitions of standard data elements.
These
definitions are then used by all new applications.
Many of the companies we studied, including Waverly mentioned above,
used a task force of both I/S and user managers to resolve definitional
problems.
I/S personnel are familiar with the technical aspects of the
definition, including how implementing the definition might affect other
systems.
User managers can see the impact on their business practices from
different definitions.
A technique we saw at Foothill Computer for resolving definitional
disputes is to refocus the dispute from "what should the definition be" to
"who has final authority for deciding".
This person or position is then the
"owner" of the data element, and has both the right to decide on the
definition, and the responsibility of responding to others who are unhappy
with the current definition.
-43-
Metadata standards add structure to the definition process by explicitly
Several of
setting out what a data definition must include to be complete.
the companies in our study had formal metadata standards to improve the
quality of data definitions and the productivity of the definition process
These standards ranged from being data-type
[Symons and Tijsma, 1982].
classification schemes to being format "languages" with syntax rules for the
structure and completeness of definitions and a controlled vocabulary of
permitted terms to be used in definitions.
In spite of the importance of developing and managing standard data
definitions, tools in this area are not yet technically adequate.
Most data
dictionaries on the market were designed to support a particular software
environment, and thus are extremely cumbersome or unworkable in the typical
corporation with data on multiple hardware and software systems.
This has
forced several of the corporations we studied to invest in designing their
own data dictionaries to be able to meet their needs.
Satisfaction of the
corporations with their proprietary data dictionaries varies, but it is
clear that there is an unfulfilled need here.
44-
REFERENCES
Appleton, Daniel S.
October 1984.
"Business Rules: The Missing Link", Datamation ,
"Infonnation Asset Management", Datamation, February 1, 1986.
Chen, P.P.S. "The Entity-Relationship Model - Toward a Unified View of
Data", ACM Transactions on Database Systems , Vol. 1, No. 1, March 1976.
Coulson, Christopher J. "People Just Aren't Using Data Dictionaries",
Computerworld , August 1982.
Diebold, John "Information Resource Management—The New Challenge",
Infosystems , June 1979.
Durrell, William "The Politics of Data", Computerworld , September 9, 1985.
Edelman, Franz "The Management of Information Resources: A Challenge for
American Business", MIS Quarterly , Vol. 5, No. 1, March 1981.
Gillenson, Mark L. "Trends in Data Administration", MIS Quarterly , Vol. 9,
No. 4, December 1985.
Holland, R.H. "Tools for Information Resource Management" Paper presented
at The Guide Conference, New Orleans, Louisiana, November 9, 1983.
P. "The Nature and Design of Post-Industrial Organizations",
Management Science , Vol. 30, No. 8, August 1984.
Huber, George
IBM, Business Systems Planning
,
IBM Manual
#GE20-0527-3, July 1981.
Kahn, Beverly, K. "Some Realities of Data Administration", Communications of
the ACM , Vol. 26, No. 10, October 1983.
Markus, M. Lynne "Power, Politics, and MIS Implementation", Communications
of the ACM . Vol. 26, No. 6, June 1983.
Martin, James Strategic Data-Planning Methodologies , Prentice-Hall, Inc.,
Englewood Cliffs, New Jersey, 1982.
Rogers, Everett M. Diffusion of Innovations , Free Press, New York, 1962.
Ross, Ronald 6. Data Dictionaries and Standard Data Definitions: Concepts
and Practices for Data Resource Management , New York, Amacon, 1981
Symons, C.R. and Tijsma P. "A Systematic and Practical Approach to the
Definition of Data", The Computer Journal (British Computer Society),
Vol. 25, No. 4, November 1982.
Tillman, George D. "Data Administration vs. Data Base Administration
Is a Difference", Infosystems , February 1984.
-
There
-45-
Ronald F. "The Evolution of Data Administration", Text of an address
delivered at the Information Management Conference, New York, October 17,
Voell
,
1979.
Zmud, Robert W, Information Systems in Organizations
Company, Glenview, Illinois, 1983.
,
?.
Scott, Foresman and
13^7
Qkb^
Date Due
Lib-26-67
flSif 'T'/'Tf-i/
3
loao 00^
„,,in
Jst'
si'!,