Accounting for Non-technical Issues
in Computer System Design
by
DARRELL LEE GRAY
S. B., Massachusetts Institute of Technology
(1974)
SUBMITTED IN PARTIAL FULFILLMENT
OF THE REQUIREMENTS FOR THE
DEGREE OF
MASTER OF SCIENCE
at the
MASSACHUSETTS INSTITUTE OF TECHNOLOGY
September 1980
Q
Darrell Lee Gray, 1980
The author.hereby grants to M.I.T. permission to reproduce
and to distribute copies of this thesis document in whole or
in part.
Signature of Author
Sloan Scho 1
f Management
ugust 1, 1980
Certified by
Stuart F. Madnick
Thesis Supervisor
Accepted by
Michael S. Scott Morton
Chairman, Department CorLI:ttee
Accounting for Uion-technical
Issues
in Computer System Design
by
DARRELL LEE GRAY
Submitted to the Sloan School of Manacement
on August 8, 1980 in partial fulfillment of the
requirements for the degree of Master of Science
in Management
ABSTRACT
This thesis develops a basic methodology for evaluating
the non-technical issues faced by a computer system designer
or implementor. It models the political and social environment that the analyst and the system must perform in, to
look for potential trouble spots. Some specific recommendations are then made to assist the analyst in avoiding
problems once they are revealed.
It
A case study provides a basis for the analysis.
offers a systematic account by an insider of a major system
development effort. The case deals with several examples of
resistance encountered, from users of the system and from
other groups within data processing who were involved in the
project. The organizational structure of data processing
was found to be a major factor in
the intra-departmental
problems.
Thesis Supervisor: Stuart E. Madnick
Title: Associate Professor of Management Science
-2-
ACKNOULEDGEIENT S
Stuart
to thank my thesis advisor,
I would like fist
I owe a great deal to 1. Lynne
Madnick, for his guidance.
Markus, my thesis reader, whose ideas and suggestions made
She helped give my vague
it possible for me to get here.
ideas this form, and freely gave assistance and encouragement along the way.
Without the great people at Worldwide Insurance, of
course, this document sinply wouldn't exist. They put up
not only with my mistakes, but with my writing about theirs
as well. I think they will recognize themselves and accept
my gratitude.
And lastly, I owe special thanks to a couple
friends for keeping me cjoincj through the last few r.onths.
M-3-
TABLE OF CO:ITEUTS
Page
Chapter
Abstract
2
Acknowledgements
3
List of Figures
5
1
Introduction
6
2
Background of the Case
14
3
History of CAS
32
4
The New CAS
56
5
Problems Encountered With CAS
75
6
Conclusions
90
Appendix
110
Bibliography
117
-4-
LIST OF FIGURES
Figure
Page
1
Corporate Organization Chart (Partial)
16
2
MIS Organization Chart (Partial)
16
3
Organization Chart --
18
Corporate
Accounting, Financial Analysis and
Reporting Section
4
Organization Chart --
MIS -
APD
18
Financial Reporting Section
5
CAS System Flow Chart
27
6
CAS History in
31
7
CAS Reporting Backlogs
8
Su~mary List of Acronyms and
Terminology
Brief
43
116
l
CMAPTER
INTRODUCT I1
Many valuable insights
into
resolving
encountered in implementing computerized
the
problems
information
sys-
have come from the application of work in
tems (CIS)
fields to the problems.
Early
work
with
computers
necessarily very much technically oriented.
were very
The
expensive and difficult to program.
they were only used when the
value
of
the
obvious (usually because of the volume of
be done).
other
was
machines
Therefore,
computer
was
calculations
to
At first, the projects were also very limited in
scope, because of
the
scale
of
what
the
computer
was
capable of doing.
As
the
programmers
changed.
machines
and
got
support
bigger
and
software
got
faster,
and
better,
The capability grew for developing
no longer served a single simple function.
things
systems
The
Programmers began to develop programs
took on more of the task
they
were
helping
with.
benefits were not always quite so clear as before,
effects on task and the people involved grew.
that
easy
program tasks with obvious benefits to automation had
done already.
the
and
to
been
which
The
the
The programs
began to be used to produce the reports instead of just the
-6-
contents of the reports,
not just the original user,
those reports.
the corputers
the
but all of
end
were
now
being
files
data
used
to
In
of
First through printeC reports,
that
cases,
rrany
and
data
store
were
processed
another user's program, the corputer became an
forwarder.
users
Instead of just Leing used to process data,
transfer it between beople.
then through
affecting
and now the procram; was
this
again
information
that
reant
by
the
human
interactions were changing or being cut out entirely.
into
Here is where programmers began to get
trouble.
run
Programmers were used to finding ways to make machines
better.
They had
no
experience
in
problems that their programs caused.
to separate people,
processes,
dealing
It
with
became
and programs.
human
difficult
Before,
the
computer had been a tool used by a single person as a
part
of a system for getting things done.
were
being rearranged to take advantage of
capabilities.
Now, the people
the
computer's
The system for getting things done
was
new
now
being changed to take advantage of the tool's abilitie-s for
improving on information
flow in
the organization.
And the
programmers suddenly needed more 'people' skills.
The data processing organizations
realized
changes
were happening, and Systems Analysts joined programmers
those departments.
They were to develop
the
new
in
systems
for getting the firm's work done, and the programmers would
-m7-
analysts would develop the package
forms,
replacing
the old syster with the new.
But still, no
profitability of the firm.
about
the
the
on
effect
really
was
one
the
in
people
a
the
to improve
and as a processor,
means of cormunication,
mechanism,
storage
took advantage of the cor-puter as a
concerned
that
Their job was to develop systers
change in the firr.
of
centers
became
were fitted into the orcranization they
analysts
the
Ps
in
aid
and
system
new
the
that constitute
etc.
progrars,
computer
of
The
-roblems.
again only have to worr" a1out the technical
organization.
It was the late 1960's when a real look was
taken
why computer systers failed, that the people issues
began to cone out.
Work from
other
fields
began
at
really
to
be
adopted and brought into use in the field, in an attempt to
technically
make system's theory 'people' sound as well as
sound.
The earlier work
psychologists,
of
and organizational specialists
problems with CIS.
reward,
was
Motivation,
Theory Y and similar words
applied
sociologists,
to
perception,
and
phrases
recurring
resistance,
began
to
appear in the literature.
Through
the
1970's
research
continued to explain problems;
much richer.
the
in
these
literature
areas
has
has
grown
But, to date most of the work has either been
explanatory or simply descriptive.
-8-
A great deal of
effort
has been put into identifying problem areas to be
studicd,
into trying to measure the extent of the problems,
and into
understanding the problems and why they occur.
to be a shortage of work
that
makes
There seers
analyst about what to do to avoid such problems.
then,
is
use in
to
suggestions
My
the
goal,
to try to develop a methodology for the analyst to
accounting for these
non-technical
issues
in
design phase, rather than in a post mortem of the
To do this, I have drawn
from
recent
work
the
project.
in
political
science on evaluating investments in light of political and
social risk.
Approach to the Problem
In my ongoing work with Worldwide Insurance* ,
begun to run into many instances of resistance to
that I was working on,
I
had
the
CIS
the Cost Analysis System (CAS).
I
was fortunate to be working for a person who was well
read
in the current literature,
and
was
current ideas in CIS implementation.
carefully
follow
introducing
System.
the
current
best
advice
technology
into
Many possible forms of user
willing
We had
we
to
attempted
could
the
try
find
Cost
resistance
out
to
for
Analysis
had
been
thought about in advance, and we felt that we
had
ourselves pretty well.
encountering
Yet
we
were
problems, very few of them technical.
-9-
still
covered
basis for a look at how one
1orldw1ide
work at
I hoped therefore to use
hopefully
and
predict
might
I was
prevent "people" nroblems in CIS implenentation.
a position unusual for researchers
that I
the firm.
I was
that
a real part of the development project
in
in
in this field,
was already established as an "insider" in
the
as
wanted
to
insider
in
I
study.
an
There are several advantages to being
this
I
situation.
didn't need to justify my presence.
to people and records I needed.
had
access
subjects
need
never
The
analyst.
I
already collected most of the facts that I would need --
files
documenting CAS.
most
of
the
old
schedules,
memos,
remained, given to me as the person
for CAS.
and
reports
participants.
And
my
note-taking
already legitimized as well,
the
in
allowing me to
had
that
informally
with
the
meetings
was
having
was
my
responsible
currently
Much of my interviewing could be done
as a part of the regular meetings I
I
but
good,
Records were not very
had
paperwork
existing
the
of
most
held
already
I
I already
systems
know that I was anything but a
so
researcher,
a
becoming
before
established myself
had
and
legitimately,
there
was
record
points
for my thesis as well as for the project.
The data collection went pretty much as
actually informed people
in
December
-10-
that
planned.
I
was
the
CAS
the
subject of my thesis, but not what the exact nature of
It
thesis was.
I
and I did not bother to clarify.
history informally with
of
each
of
written
questions, the written evidence and
I
fortunate
was
that
find
With the exception of some timing
the line of duty.
well.
in
didn't
I
that
already have a copy of, but I was again able to
out in
as
informally
them
available
documents
and
system
the
There was very little
well, as a normal part of my job.
the way
interviewed
formally
several people, but I was able to discuss
its
technical,
was
was assumed that the topic
to
enough
memories
corroborated
fairly
several
find
complete system status reports among the rest, which helped
never any question in my
that
mind
I
what
was
there
Thus,
put things into a consistent time frame.
had
as
was
accurate as people remembered, and reasonably complete.
Much of the early
of
history
documents, and other assorted memos,
their
way
to
my
files.
whole
The
came
CAS
etc.
that
rounded
was
those
from
had
made
out
especially-
by
who
interviews with Rob West and Don
Massey,
had both been on the task force.
It is interesting to note
that no one from the original six members of the task force
was still
at Worldwide
four years later,
and only
Don were left of those who had served on it
Rob
and
at any time.
been
associated
with the system since the task force dissolved.
I was able
It was easier to find people who had
-11-
to talk with several prcgrarmers,
analysts,
and accountants
who had worked with or on CAS at some point.
I
was
also
able to reach a couple of the oricinal prograr2.ers from the
task force period for their comments
possible,
as
well.
I took the stance of simply being
Whenever
interested
the past of the system that I was now involved
asking technical questions about the early
in
with.
system,
By
I
was
even able to legitimately take notes at these times without
making anyone uncomfortable.
My thesis
thus
what
provides
is
probably
a
more
accurate account than an outsider would be able to produce.
I have had the advantage of
first-hand
knowledge
thinking behind many of our decisions.
Thus,
of
this
the
thesis
serves the purpose of providing a systematic account of the
resistance issues involved in a
resistance from
users
processing group.
framework for
In
and
from
addition,
previewing
the
finally,
I
have
implementation,
the
rest
of
I have tried to
non-technical
might become a problem in CIS
And
CIS
design
attempted
the
both
data
provide
issues
a
that
and
implementation.
to
make
concrete
recommendations for avoidance of some of those problems.
Plan of this Thesis
In Chapter 1, I have presented
this
project
and
some
of
the
-12-
the
reasoning
conclusions
behind
developed.
Chapter
will
2
providc
understand the case,
in
project,
a
of
the
fairly
nece ssary
to
and
the
terms of the major actors
early history of the first
out of some
background
the
that
problems
complete
interactions is important.
during
up
came
history
Chapter 3 takes up the
through
of CAS from the time I began work for the company,
The
with the development of the new system in
story
continues
Chapter
4.
complete
the
I
with
take a more detailed look at the problems encountered
the project in Chapter 5, and
the
earlier
the
of
outline
the decision to rewrite the system.
sense
To make sore
CAS systen.
analysis
and
of
the
conclusions in Chapter 6.
acronyms
appears
as a reference in
case the
A summary of
Appendix,
overwhelming.
-13-
at
the
end
abbreviations
become
CHAPTER 2
BACKGROUND OF THE CASE
This chapter will provide the necessary background
to
understand some of the problems encountered in the development of CAS.
It begins
with an overview of the firm and a
The early history of
look at the major actors.
the
CAS is then outlined to establish the setting for
first
Chapters
Three and Four.
The Firm
Worldwide
London based firm,
Boston.
It
is the American affiliate
Insurance
with its U.S. headquarters
collected
nearly
one
billion
of
a
located
in
dollars
in
premiums in the U.S. in 1979, primarily in commercial policies.
The firm has grown rapidly in the
resulting in a lot of pressure
on
last
computer
other support areas, to cope with the rising
to expand the
ment has
made
services available.
many
major
the rapid growth.
with far
and
workload, and
including
reorganizations, in the past few years,
years,
services
manage-
corporate
The
changes,
few
trying
corporate
to
manage
Unfortunately, a number of these changes
reaching consequences (especially for
cessing) have been made with very little
-14-
notice
data
pro-
to
those
who might be affected,
problems, such as
and more
late,
This hcs resulted in
a
number
of
monthly reports being published 6 months
and
in
a
lot
of
bad
feelings
toward
management.
The MIS Division
The Management Information Services Division (MIS) has
overall responsibility for all data
vice
MIS has its own senior
at Worldwide.
activities
processing
president
who
Within
MIS
reports directly to the executive comrittee.
there are four departments, plus an administrative group.
The Applications and Programming department
responsibility for all application
maintenance
for
non-MIS
department the current
lines,
by the
name
departments.
organization
parallelling the
organization.
system
of
structure
development
and
the
APD
Within
is
along
the
responsible for.
within the APD
the
major
system
financial
reporting
the
accounting
subsystem
section,
it
is
function
of
accounts
to
is
of,
usually
This corresponds
50 people headed by a junior vice president,
-15-
the
member
to
Worldwide.
Within MIS the financial reporting group consists of
sibilities for general ledger,
of
referred
Thus, the group that I am a
called Cost Analysis or simply C-A-S.
the same area of
or
functional
rest
Each programming team is often
of
has
(APD)
with
payable,
about
responvarious
WORLDWIDE INSURANCE
Corporate Organization Chart
(Partial)
Figure 1
WORLDWIDE INSURANCE
MIS Organization Chart
(Partial)
Reporting
Statistical
Figure 2
-16-
federal and state
systems,
insurance
reports,
risk
outstanding
and
claims,
premiums,
profitability,
finc'ncial
req>uired
Cost
the
and
some workload measurement systems,
Analysis System.
day to day computer operations --
preparation,
data
all
provides for
The Data Processing department (DPD)
scheduling,
entry,
pro-
and quality control for running existing
grams.
The
Systems
Department
Technology
(Technology)
system software.
They
providing
with
charged
also
are
other
and
responsible for all hardware, operating systems
is
technical expertise to APD when required.
The Systems Management Department
a technical quality control monitor.
has a role
(SMD)
as
inter-
They assist in
facing with user departments, develop the functional specifications for most major MIS projects,
In
-benefit analyses and post-completion audits.
they
monitor
performance.
ongoing
application
cost-
provide
and
addition
and
usage
program
Sue, our representative from SMD, started
Worldwide at about the same time that I did.
(and new) CAS project requests.
then has been spent with CAS,
Much
despite
of
her
immedi-
She
assigned
ately became involved with CAS as one of her first
systems, trying to manage and interpre t the
stack
her
of
other
old
since
time
respon-
sibilities, because of the scope of the modifications
-17-
at
that
WOPLDWIDE INSURANCE
Corporate Accounting
Financial Analysis and Peporting Section
. O'Donnel
lanual Rpts.
pecial
rojects
J. Sullivan
Budget and
Planning
Reports
Figure 3
WORLDWIDE INSURANCE
MIS -
APD
Financial Reporting Section
Al Garber
Staff: 53
Figure 4
-18-
we decided to make to the systen.
MIS at present includes over 300 staff members and has
time.
training/edu-
ware products and the availability of a good
cational program,
employee turnover (the
average
programmer
new
personnel
attract needed people.
This tends
not
job
perpetuate
to
makes
and
troubles,
the
in
is that Worldwide has gained a bad reputation
any earlier
will
Part of the problem now
remain a full year at this point).
market as a high turnover firm.
high
very
from
suffers
the MIS group
soft-
and
Despite investments in advanced hardware
any
at
about 50 additional people as contract procramers
hard
it
to
As a result, Worldwide is forced to
maintain a large number of
contract
fill
to
programmers
slots that should be filled by permanent staff.
Most of the MIS staff
are
issued
copies
a
of
volume loose-leaf manual which describes policies and
cedures of the
division.
This
provides
guidelines
six
proof
general MIS policy, as well as attempting to routinize some
of the normal operations.
For example, standard procedures
for submitting tests to be run are layed out in
detail
in
it.
Mr. Bradley, the vice president
who
runs
several of the other corporate officials, appears
regular change in his organization.
Like
MIS,
fond
corporate
like
of
man-
agement, he tends to mandate change and only later discover
-19-
what the effects were.
his decision in
A recent exam,ple is
response of several users was to
tional unit outside
MIS.
of
One
such
group
was
the
who
took
two
accounting area's financial reporting group
SMD people
new
(including Sue)
section
functions.
SMD
the
replace
later
Although no official retraction was made, two months
informed
casually
were
people
some of the remaining SMD
a
and an APD manager and forred
to
themselves
func-
similar
a
create
immediate
The
August 1979 to disband the SMD department.
that the group would not be dissolving in the near future.
A less drastic but typical change is the rearrangement
years that I have been
at
reported to five different
really only changed once,
than
two
Lou,
has
function
has
less
the
In
of senior managers under Bradley.
boss,
Worldwide
my
people.
Lou's
to
that did not correspond
and
any of the 5 changes above his level.
MIS Project Requests
The MIS Division is
a
as
run
in
There is no billing mechanism for any of these
services, so these requests are
really
the
monitoring and controlling the volume and
services.
writing
in
formally
Their services must be requested
most cases.
organization.
service
The suggestion
for
a
project
only
disbursement
may
of
of
come
from
a
user
within MIS, but the request itself must be made by
-20-
way
of the system.
For some services, such as
for
request
a
this may take the form of a
DPD to run an existing program,
For the development or modification of a program
memo.
CIS, the request is
Project
MIS
an
completing
by
made
or
Request Form.
that user.
assigned to
and to the SMD representative
The coordinator also assigns a
number
project
System,
Control
Project
the
in
and records the project
APD
appropriate
who forwards copies to the manager of the
programming team,
MIS,
in
Requests for programmr.ing go to a coordinator
which is used to monitor the status of all projects.
The
APD
manager
project
the
consult to determine whether
then
representative
SMD
the
and
significant
is
rule
enough to require formal specifications (the informal
of thumb in use is
weeks of programmer
project
that any
Project Request Form is
formal
gets
time
with
more
the
If
specs).
not completely clear,
2-4
than
may
then SMD
either choose to return the request to the user for further
with
clarification, or to consult
the
user
Otherwise,
develop formal specifications.
if
to
and
APD
the
request
is clear and complete, APD may choose to accept or deny the
project as defined in the request.
Whenever the
enough,
of
intent
a
requested
becomes
the APD manager has a right, after consulting
to reject the request if
it
seems unreasonable.
-21-
This
clear
SMD,
may
a duplicate
benefits the requestor rather than the ccmpany,
of an existing system, or a simply impossible idea.
for
wise, APD may accept the project
and defer it
accept it
resources.
as far as I
(In practice,
or
other
or
time
determine
could
project has ever been forr.ally rejected.
Other-
work,
immediate
for available staff
that
system
a
such as
inappropriate nature of a project,
the
for
or
be for reason of excessive resources required,
no
Instead, they are
accepted and deferred permanently).
case,
In any
is
user
the
written
a
to
entitled
response to a request in a "reasonable time".
(Again,
the
manual fails to specify what reasonable is;
normally
it
seems to be about one month).
If
accepted,
project
the
is
programmed by the responsible APD group.
are
to
supplied
the
When all
delivers them to the users.
(on the formal
specs
or
the system.
requirements
request)
project have been satisfied, the user must
may be run at
it
off"
any
the
on
submits
and
APD then completes documentation
for production,
of
"sign
the system to DPD to be put into "production".
accepts it
turn
in
who
stated
original
the
tests
Results of
representative,
SHID
and
scheduled
then
Once
DPD
time
the
user makes a written request.
The user has no option to change
requirements
cancel the project once programming commences.
-22-
Until
or
to
that
are
They
request.
any
time they may withdraw
so
usage,
neither for development nor for computer
charged
their
incentive for making reasonable requests depends largely on
simultaneous
major
The project's status is reported throughout the
cycle
the inability of APD
many
handle
to
requests.
to
going
issued on a weekly or monthly basis, some
are
reports
Various
through the Project Control System.
users
and some to MIS, and to all levels of management.
The Cost Analysis Department
Department
The Cost Analysis
within the accounting
function
evaluate the worth of various
business.
(CAD)
was
the
early
in
portions
of
established
1970's
to
company's
the
They were to look into the profitability of each
office doing business for
the
company,
and
ability of each line of insurance business.
the
profit-
They would use
cost chargeback analysis, which invoved finding appropriate
ways to allocate each of the indirect costs of doing
ness to offices and lines
of
insurance.
Then,
ability statements would be developed for each
busiprofit-
office
and
line.
The goal of the
department
has
remained
through several staffing changes, changes
and in actual personnel.
both
consistent
in
number
The staff has been made up mostly
-23-
of
cost
with
along
reorganized
August 1979, this department
was
most of the accounting area.
Its function was expanded
staff
The
corporate financial reporting.
larger
its
of
increased
was
and
Cost
back to about 15 and the name was changed to the
Expense Department to recognize
to
range
wider
a
for
include additional responsibility
In
trainees.
accountant
cost
and
accountants
the
in
role
organization.
essentially
to
automate
system
The
the primary user of the system.
the
expand
and
was
designed
profitability
a man-
Rob West,
reports that CAD was already producing.
been
has
Throughout the life of CAS, this department
ager in the department, has been a primary user and contact
with MIS since his arrival in early 1976.
The Early CAS
The Cost Analysis System had its beginnings in January
1975,
with a directive
company.
That created a
force
task
review existing (manual) methods
president
the
by
issued
with
a
charter
allocating
of
of
the
to
expenses,
and where necessary, to develop new procedures for charging
out those expenses.
was
The goal
allocation of the full costs
of
a
doing
office and to each line of insurance.
that use of
the
computer
would
-24-
aid
fair
and
business
equitable
to
The expectation
in
producing
each
was
such
reports in
a recular and timelv manner.
people
The project team. was formed from the top three
in
IS.
the Cost Analysis Department and three people from
would
developed
procedures
They quickly decided that the
the
indeed have to be automated to be used effectively, so
MIS staff continued with the project.
September 1975, the
gathered
team
local and regional office operating
to
February
From
reviewed
information,
then
and
procedures,
began to develop allocation methods and report layouts.
In
to
October a software house was brought in
They
the bulk of the programming.
the
with
worked
handle
MIS
members of the team to develop final system specifications.
In November 1975,
programming was started using an
initial
design that incorporated all of the specifications that had
been completed at that point.
under
The task force was
project,
complete
to
pressure
staff
and since the software firm had limited
devote to the project,
into two pieces.
was decided to split the
it
The first
to
project
part- that
the
piece would be
the
would extract all necessary information from other
systems
the
system.
and format it
The second
for use by the second
section
would
do
all
of
part
of
analysis
the
and
reporting to accomplish the project's objectives.
The first
section
was
brought
back
in
house,
several contract programmers were hired to do the
-25-
uork
and
on
it..
The
second
force,
task
the
according to the procedures developed by
expenses
the
allocated
which
section,
remained with the software house.
and implemented in
step was completed
The first
two steps.
had
that
features
in June 1976 and offered all the
developed
was
system
the
During the next two years,
been
requested prior to November 1975.
any problems
The second step was to resolve
the step one design,
and in
addition included many enhanceduring
the
At the end of December 1977,
the
complete.
The
the
ments that were developed by
programming of step one.
found in
system was considered to be
more
force
task
or
project team had gradually dissolved
less
the
towards
end
of
1976 as the members began to meet less and less often.
The
(CAD),
had
official
re-
primary users, the
Cost
Analysis
Department
begun to request further changes through the
quest process by that point, rather than by working through
the project team.
June
Between February 1975 and
force had changed several times;
the old in
the team.
of
The MIS members of the task
both MIS and CAD had changed.
members remained by mid-1976.
management
1976,
none
of
the
original
The new CAD manager replaced
According to an August
1976
project
status report, this turnover had not only resulted in a lot
of new ideas and points of view, but in a
nearly
complete
Figure 5
continuous
from
pressure
there
".eanwhile,
redefinition of the original plan.
immediate
for
.top management
was
results.
to
Because of tie pressure, there was no chance
stop
any
changes.
As
decisions were made about what to add next,
the task
force
and review the program design and make
and passed them on to
produced specifications
Each
nrograrintnq.
for
software house
and
the
enhancement
was
but
fitted into the system as well as possible,
was
design
one
the
specifications.
Therefore, it
the
from
developed
was not
all
of
the
task
force
basic
the
1975
November
changed
easily
accomodate some of the early oversights.
1977
IS
the
At
specifications
end
had
to
of
been
implemented and the known problems resolved.
In
preparations were being made
January 1978,
the CAS into 'production',
tion, etc.,
completing
that is,
to
documenta-
so that the responsibility for regular
of the system could be given to DPD.
A
requested by the user dated back to 1976,
put
list
running
of
changes
but the
original
request as proposed to MIS had finally been fulfilled.
So,
the system had been "signed off" by the users,
MIS
project request rules,
to meet
and to make requests for changes
to
CAS legitimate.
Even before it
that the system was
was accepted,
an
the "live"
operational
-28-
tests
nightmare,
showed
requiring
massive amounts of r-acine tire, tapes,
its monthly runs.
The
allocation
and disk space
for
required
processing
final monthly run of 15-20 hours of continuous time in
a
the
computer with no possibility of restarting in the middle in
case of problems.
At the
same
DPD
time,
was
averaqing
success-
about 10 hours between machine failures, making a
ful CAS run a case of
extreme
good
luck.
The
reports
of
produced were completed 2-3 months after the close
And there were still major
month that they were reporting.
reporting problems that made
it
to
impossible
from where and how expenses were allocated to
final numbers.
the
determine
produce
the
According to Rob West:
There was no way to audit the results that came
out of the system. The Policy Transfer reports
didn't match the Allccated Cost reports or the
extracts.
The prograr.mers called it
a rounding
problem and said it couldn't be helped. We were
using a dozen allocation passes at the time,
so
the results were next to impossible
to recreate
by hand, not that we didn't try.
Even our
extract reports had to be manually totalled to
get the base figures for checking percentages.
As a result,
the reports were
not
accepted,
trusted,
or
used by the offices that eventually received them.
At the end of 1977 Don,
CAS transferred to SMD.
the lead
analyst
This was a major step up for
which he credited to the success they
had
Although he kept some responsibility for
tially, he was "quite fed up with CAS and
involved with it".
working
had
the
all
him,
with
CAS.
system
ini-
the
people
During January, responsibility for
-29-
on
CAS
hod Leen shifte.
Lou).
to a new vanager in MIS
(my eventual
At that point, Tom was the only person left
on the system.
year by then.
He
had worked on CAS for a
person with any real
left
little
knowledge
of
But Tonm,
the
over
for another company in March of
he left, Beth was somewhat familiar
worked, but lacked
any
real
knowledge
workings of the programs or of
She was further
with
the
in
the
This
By
tine
the
the
the
system
internal
allocation
hampered
last
the
1978.
how
of
a
Beth,
programs
was just before the February processing run.
being used.
working
Lou had assigned a second programmer,
from his group to work on CAS as well.
system,
loss,
procedures
the
because
processing was done in a single huge PL/l program
major
and
she
was a trainee and only knew COBOL.
When the February processing
was
attempted
March/early April 1978, the PL/l program
to a halt
part
way
through
the
failed,
several
message about an attempted illegal
several attenpts to fix the program
division.
using
it
a
was
brief
Beth- made
available
uals, but she was unable to solve the problem.
late
grinding
hours
supposed to run, and printing in explanation only
in
man-
No one else
in IIS with enough knowledge of PL/l to help was willing to
try to fix the program.
At that point a decision was
to try to hire a PL/l programmer to help,
I joined the CAS group.
-30-
made
and in June 1978,
CAS hISTORY I1" 3RIEF
1975
January
. Directive creates Task Force
. Task Force begins Analysis of needs
November
. Programring of CAS begins (stage 1)
. Task Force continues analysis
1976
June
. First round of programming completed
. Stage 2 prograriing begins
1977
December
. CAS "Signed off" by CAD
1978
March
. Tom leaves, CAS stops working
June
. Darrell hired
. Corporate Reorganization announced
1979
August
. CAS running, Improvements underway
January
. First suggestion of rewrite to CAD
. Regular meetings with CAD begin
May
. CAS rewrite proposed again
July
. Rewrite proposal to CAD management
August
. Feasibility study requested
. CAD reorganized
1980
October
. Feasibility Study results presented
November
. Rewrite requested, terminals ordered
March
. User terminals installed
April
. Interface problems resolved
June
. Scheduled system delivery date
Figure
-31-
6
.
C:HAPTEP. 3
THE HISTORY OF CAS
When I arrived at Worldwide, CAS had
accepted by the primary user, CAD.
officially
been
This does not mean they
were happy with it, only that under the rules
of
project request system they had to "sign off"
the
once the request, as stated, had been fulfilled.
the
MIS
project
And
CAS
met its original specifications, however badly it did so.
But, at that point, the February report was
late and CAS wasn't working.
accepted by DPD as
continuing
tation.
a
months
The system still hadn't been
production
system
because
operational problems and the lack
Some programs
4
of
of
docuren-
had been almost continuously
state of flux since they were written, sometimes with
than one change being tested at a time.
had
never
production.
even
been
requested
to
As a
accept
the
in
a
more
result,
DPD
those
for
Reports had been issued from the system to the
eventual users several times, but there
remained
problems and questiors about the validity of
the
that the reports probably wouldn't have been used
had been timely.
so
many
results,
if
they
There was a general consensus among
both
MIS and CAD that although the basic idea of the system
was
good, the output, such as it was, had never been put to any
-32-
real use.
The extract programs
worked.
system,
the
of
end
front
processing, which made up the
for
available
data
made
that
They consistently ran to their normal completions,
changed by a nur'ber of
been
had
programs
The
and produced reasonable results.
programmers over the two and a half
years of their existence.
Some had already been changed by
more than a dozen people,
most
programmers.
of
being
those
no
There was essentially
contract
for
documentation
the programs and no consistency in either methods or naming
conventions.
Internally, the programs operated in a
style
principles.
Any
accounting
which was far from accepted
data that failed to meet the "rules" of
were
program
the
explanation,
simply rejected, with no warning message,
even totals of rejected data.
Bad
data,
codes, data that failed edit or
selection
or
untranslatable
were
criteria,
all simply ignored.
typically
The reports, as a result, were
cent short of the
system.
totals
to
required
But, since there were no
match
obvious
a
few
feeder
the
biases
per
in
the
results, and they were only used to develop percentages for
cost allocation in CAS, those errors
were
accepted
(that
is, the users considered the problem much less serious than
many others with CAS).
The only saving
point
about
that
part of the system was that the programs were fairly simple
to
and
start
scimehow
acceptable
produced
still
they
(although incorrect) output.
actual
The rest of the system, however, which did the
cost allocations,
program.
consisted of a single huge PL/1
I think the program may have violated every modern programming rule.
Trying to work with that program would convince
struc-
and
design
anyone of the value of modular program
It is simply not possible to
tured programming techniques.
once,
read and understand several hundred pages of code at
of
this
used none of the ideas
that
logic
let alone trying to follow the complicated
At the same time, it
program.
make a programmers life easier.
and the
margins
structure.
were
all
There were no blank lines,
yet
different,
Variable names had almost no meaning,
two letter names were used where ever
no
indicated
or
one
possible,
and
were
On
one
page
used again for different purposes elsewhere.
'XX' might be a string of characters containing the name of
a line of
reappear
insurance,
and
a
few
away
pages
as a counter or a premium dollar amount.
was no documentation, either internal or external,
addition the program used a number of
will
'XX'
-There
and
in
advanced features of
PL/l that made it even more difficult to
understand
(such
as the extensive use of pointer variables).
There was a suspicion that
the
last
programmer
sabotaged the program when he left, but the program was
-34-
had
so
cor.1plicatcd that chanccs are that if
he wouldn't have bothered.
When
it
the
he had had
the
worked,
required 12-15 hours of continuous computer time
up several disk units and several
moment it
wasn't working at all;
and
tied
At
the
current
something in the
data was causinu it to reach a point where it
processing.
program
drives.
tape
urce,
simply
quit
It had been designed under several assumptions
that had changed even before the first code was completed.
When they started the design process, the organization
was simple and compact,
lines of insurance.
build the system
The table
around
those circumstances.
for
as was the reporting structure
was
Under
they
structure
reasonably
the
new
chose
under
efficient
conditions
the
old
the table structure became
design was not practical,
very
inefficient with a larger, more complex organization to
represented.
Also, some
of
the
PL/l
simpler
processes
originally combined in
the
allocation
processing efficiency.
But, as those processes were
to
program
be
were
for
rede-
fined, they often became more complicated, and the logic to
allow them to remain combined became more complex as
well,
until eventually, they should have been separated again for
efficiency.
the
But the software house had neither
the inclination,
nor the incentive to rewrite
the
tion program during the middle of the project.
alloca-
The extract
portion of the system had fared much better.
The
were so simple to becgin with that
simply
-35-
they
had
time,
programs
gron
messy with chances.
During the remaining week of
June
and
through
July
1978, most of my time was devoted to trying to unravel
the
hundreds of pages of PL/l in "the" allocation program.
I
needed to find out why it wasn't working, and
to
try
any
possible solutions that might solve the problem temporarily
until I could figure it out.
"experts" of the firm,
me and my boss that it
them questions.
was not a
appeared,
but it
quickly became
PL/l
apparant
was not worth the effort
of
to
asking
They had no particular interest in helping
to solve the problem.
it
I was introduced to the
technical
In any case, it began to appear that
problem
but a problem in
with
PL/l
as
it
first
the complex processing that
was
going on internally.
Originally, as it
failed,
the
system
had
message about an illegal division computation.
this to the specific numbers however,
and the result should be legal.
But
then
that the variable containing the result
the time of failure from what we
allows you to specify the number of digits to
one
they
like
discovered
redefined
it
a
traced
looked
I
was
thought
for a number, and the result here had
We
and found that
It therefore
PL/1 was failing for some reason.
given
was.
be
at
PL/l
reserved
digit
too
few
attack
on
the
reserved.
I
problem,
began
a
systematic
brute
force
chanring anthine that miTht lead to the
-36-
proble-M,
changes
and then chanjing any new problerms that those first
had caused.
For the most part,
the number of digits reserved for
the number allowed for
the
involved
this
Increasing
variables.
result
the
of
in a computation that in turn becare too large.
all
the
fields,
was
requirements for computer memory and disk space
much worse (one digit would cost roughly 10%
used
I couldn't
large
already
our
or
had
division
caused another problen later on, when the result
simply increase
increasing
get
would
storage
more
space).
repeating
After
So I proceeded one step at a time.
the cycle several times, I was finally rewarded with a test
that worked.
Eventually, the method had paid off,
the end of July,
had been wrong,
operation.
although I didn't really
understand
at
what
the program was working and CAS was back in
The problem had been that the numbers had grown
too large for the program to handle, but
why.
and
it
wasn't
It could have stermed from the data or from the
clear
con-
trol tables set up by CAD, but it hadn't appeared until far
into the processing.
I tracked it
It wasn't
down to an unusually
until two weeks later that
large
(but
legitimate)
data item.
By the first week of August 1978, then, the system was
again working.
Rob West suggested to
his
superiors
that
they skip most of the monthly reports that were overdue, in
-37-
a single run of the system and
already behind schedule.
the
that
amount
They agreed
that
weren't worth producing at this point.
we
were
reports
those
An effort would
Le
and
made to produce a June report more or less on schedule
to keep up on time delivery fror
week in July a
announced,
major
corporate
that point on.
The
last
had
been
reorganization
but so far no one knew what the impact would
be
on MIS.
At this point I was responsible for 3 programmers
5 systems.
Beth was working on the COBOL part of CAS
and
(the
extract programs that provided data to the system).
was also working on the only
controlled,
feeder
system
but that system required little
no requested changes.
and Bob,
CAS
The other
two
She
that
support and had
programmers,
worked on the remaining three
systems
Sharon
under
control.
These were vendor supplied packages that had
problems,
but required many small maintenance jobs,
changes in report headings, etc.
went back to school
employee.
full
time
At the end
and
we
of
became
a
our
few
mainly
August
part
My boss began to look for a replacement
I
time
for
me
as the group leader.
During August, Beth began to try and get CAS back into
shape and try again to put it into production.
couple edit programs to the system to allow
make their own changes to system tables.
-38-
the
I
added
users
We also began
a
to
to
Although the system ladn't been accepted until t:ihe
1976,
there were several projects
changes
requesting
On reviewing then, we discovered that they
ranged
logic
from simple report modifications to major processing
would
correct,
simplify,
and
of
dated
that
outstanding
from the last few months of that year,
to CAS.
end
changes
that
extend
system.
For instance, they wanted to not only report
the
data
rejected by the extracts, but to be able to correct it
and
that
the
include the corrections in later processing,
data used would match what
systems.
we
from
expected
so
the
feeder
numbers
They also wanted to be able to trace how
were produced for the allocation reports.
We discussed these with the users, and I was
make changes to the
system
requests out of the way.
to
get
two
of
able
the
to
simpler
One additional project, involving
a major change to the way one of the tables was used in
extract program, was also chosen.
an
Beth and I began to work
out the best way to handle the change to processing.
When
I went back to school she was beginning to make program and
file changes necessary to the project.
came up and that project was set
most offices to
ization.
a
new
things
information
about
It involved reassignment
regional
and
Agents had been reassigned to
-39-
other
aside for a time.
At that point we began to get more
the corporate reorganization.
But
territorial
new
of
organ-
offices,
and
the whole change was to be efffective back to the
January 1978.
All
were
reports
that
of
realigned
the
show
That meant
structure in the July runs.
and other long term reports
to
first
year-to-date
(such as our own) would have to
have all of their data realigned for all earlier periods to
Also, because
meet this requirement.
of
reassign-
agent
ment, much of the source data for a number of systers would
have to be entirely reprocessed.
and claims are
responsible for
reported
that
to
the
policy
is
This is because
office
that
currently
premiums
the
agent
assigned
to.
Since most reports are only at the office level, data files
are frequently summarized to that level as they
processed, thus losing
reorganization.
the
information
are
needed
first
for
this
And because office numbers in most of
the
numbering systems being used include a region indicator
one of the code
characters,
many
office
as
location
codes
happened.
The
would have to be recoded.
This is not the first time this
had
last reorganization, during the initial CAS development
in
1976, had been the cause of much of the inefficiency of the
system.
Fortunately, CAS didn't have
about that time.
suffered.
It
had
been
the
any
data
other
to
systems
Once the executive staff makes such a
it seems that they are unapproachable
about
Our response was limited to persuading Sue
-40-
worry
that
decision,
changing
to
champion
it.
a
It
proposal for a corporate-wide numbering system.
provide a common set of
reports, to replace the codes
each system.
all
for
codes
systems
and
individually
for
new
developed
would
would
It wouldn't avoid reorganizations, but
as
make the consequences easier to bear,
as
well
making
reports from different systems much more comparable.
to
CAS
system
for
Not only are
the
The several systems which provide information
unfortunately each use a
different
numbering
coding offices, lines of insurance, etc.
different
codes different, but the level of detail is very
from one system to another.
Yet
in
order
CAS
for
utilize the data, it has to be matched across systems.
accomplish this, a number of translation tables
tained.
are
to
To
main-
The system then attempts to translate each code it
receives in the
extract
processing
to
a
Since the
numbering system set up by CAD.
single
common
incoming
codes
would all be changing, the cross reference tables would all
have to wait for the source data
to
before we could run the July report.
changes necessary to the
extract
be
would
We
have to be rebuilt to process the recoded data.
converted
anyway
There were some minor
programs
to
print
the
correct headings, etc., but no changes were required to the
PL/1 program to handle the reorganization.
The project to modify the extract tables
hold while Beth worked on
the
-41-
changes
for
was
the
put
on
reorgan-
ization.
week in SepCember,
But the first
that she was getting married --
two
hadn't
ever
Fortunately,
Beth
I
run
Initially,
asked just to figure out the many necessary
cycle,
order to run the jobs and which
to
was
files
correct.
to
get
from
which
to
which
(There
and others had set up for testing various
our
Sharon was
use,
several cases multiple versions of the programs
she
in
steps
the system through the next processing
version of which program
before
programmers
group was reassigned to work on CAS.
where
that
had almost everything set up for the July
other
to
a
person
and
bothered looking at the COBOL programs.
Sharon, one of the
taking
closer
I was now the only
knew anything abcut the CAS programs,
left.
announced
and that she was
job on the other side of the state to be
they had decided to live.
Beth
were
that
changes.
in
Beth
Beth
had made the changes for the July run, but had forgotten to
tell us which version to use.)
Eventually, Sharon
was
to
take over maintenance of the extract programs as well.
In October, we actually got the July report out.size of the translation tables used
by
the
increased significantly because of the new
structure.
larger
tables
generating
the processing tables that drive the
had
organizational
Because of the increased complexity of the
organization and the
data,
extracts
The
actual
allocation program became much larger as well.
new
extracted
expense
The
July
CAS REPORTIUC PAC:KLOGS
1976
1977
1978
1979
1980
June
July
August
September
October
November
December
January
February
March
April
May
June
July
August
September
October
November
December
January
February
March
April
May
June
July
August
September
October
November
December
January
February
March
April
May
June
July
August
September
October
November
December
January
February
March
April
May
. First complete test (Step 1 reports only)
.
(occasional tests)
. January reports run
. February and March reports run
. April reports run
* May reports run
.
.
.
.
.
.
.
June and July reports run
August reports run
September reports run
CAS "Signed Off" by CAD
October and November reports run
December reports run
officially issued)
January reports (first
. February reports run
. March reports run
. July reports run
. October reports run
. November and December reports run
. March and revised Decerber 1978 reports
. June reports run
September reports run
October reports run
December reports run
January reports run
February reports run
Miarch and Anril reports run
F igure 7
-43-
run did complete succe-ssfully,
but required ncrrly twice as
much disk space and twice as much computer time as
previously.
For several months then,
to handle the many special cases that had been
So,
had
they continued to add
more and more to the processing tables each run,
done incorrectly.
it
in
order
dropped
the resource requirements
or
continued
to grow every time that CAS was run.
A decision was again made to skip
several
processing because July came out so late.
months
of
CAD's management
wanted to get the system closer to on schedule rather
than
have the additional reports.
In October, my boss hired two
new
people.
He
had
finally been successful in finding a new group leader,
but
Andy didn't know PL/l.
So, instead of Andy taking over the
CAS group, a new group was formed for him.
He
took
over
the three systems unrelated to CAS that we supported,
with Bob from my group,
who worked
on
them
(and
along
another
programmer from my boss's other group).
The
second
programmer.
person
hired
was
PL/1
contract
Marie was hired to join CAS until a
permanent
person could be found (and probably
backup in
a
also
case I should decide to leave).
to
She
serve
and
as
Sharon
once again began the job of trying to get the system
to put into production.
And I again began
list of project requests for CAS.
-44-
to
I attempted
ready
review
to
a
the
explain
the complicated table structure necessary for
the
project
Beth and I had started, but neither Sharon nor Marie seemed
to understand.
So,
I left
that project
on
hold
time being and we made plans to begin work on a
for
the
couple
of
the simpler requests.
At the end
of
October,
management
announced
desire for us to run an October report (November
the next one we planned on producing).
their
had
been
One of the projects
that we had activated required a complete run of the system
with a special set of processing control tables.
to produce a test file for use
by
another
decided to make the special run for
(It
was
system).
We
October
as
well,
so
that we wouldn't have to make any extra extract runs and we
would have a
encountered
normal
a
run
number
of
to
compare
problems
totals
with
with.
the
We
extracts
ourselves, and were running late, when OAE notified us that
one of the files they provide us
properly before.
December.
and
the
That
problem
had
not
was
been
realigned
corrected
by
mid-
But because of our growing resource requirements
increasingily
heavy
demands
on
the
general, our final processing run for October
pleted until New Year's Day, 1979.
system
wasn't
The special
run
in
comwas
completed the following weekend.
During January we didn't have
requested by the users underway.
-45-
any
Sharon
projects
was
directly
working
on
several chances that
would
be
required
in
These
were
programs to process January.data.
marily changes resulting from the corporate
--
the
extract
still
pri-
reorganization
often undoinq temtporary changes that were made
mid-year
to handle the conversion.
Marie had been brought in as a PL/1 programmer,
as it
turned out, she had had a course but had
the language before.
but
never
used
She had no interest in learning about
CAS, and since she
was
forcing the issue.
She continued the process
temporary
it
didn't
seem
of
worth
cleaning
up and documenting the system so that we could put it
production.
Ily boss Lou and I agreed that
going to be of much use to us after the
completed,
and she was told
that
her
Marie
into
was
not
documentation
was
contract
would
be
allowed to expire at the end of the month.
We felt that the operation of the system was
so
impractical,
projects,
that
rather
I would spend
my
than
time
on
resources required for a run of CAS.
of weeks analyzing CAS, trying to
could improve the system.
my
working
trying
to
cut
isolate
areas
that
I read through the code
the code was so involved that
change,
on
rearranging
So I looked for places
the
that
but
the
we
we
looking
most
logic
might
without affecting logic, such as disk usage.
-46-
user
I then spent a couple
for obvious flaws that could be streamlined,
dangerous.
becoming
of
was
effect
I had come to some
Toward the end of January,
sions.
would
First, the system
to
have
the users.
The idea of
desijn.
the
control
the
tables,
processing
be
essentially
rewritten to avoid some central problems with
The entire systcm was table driven, under
conclu-
of
however,
had
been firmly followed throughout the design, with no account
taken of the fact that the tables might not be "full".
incoming data and the flexibility of
the
The
controls
users'
allowed the tables to be very sparsely filled.
Imagine
10 by 10 table to be processed,
8
entries being zero.
totals for the
If
table,
we
with
want
to
get
but
row
the system worked on that basis,
of
and
we have to look at each of
numbers even though most are zeroes.
in
all
a
the
column
the
100
All of the processing
except that the
tables
were much larger and the calculations, often quite complex,
were performed even if the number was a zero.
Each
indi-
vidual process in the program would have to be looked at to
determine what rules might be used to recognize
work might be bypassed.
Or the system could
to ignore the irrelevant parts of
each
table
that
be
some
rewritten
or
to
not
treat it as a table at all.
The second conclusion was that because the disk
space
required was managed the same way, there was probably
little data using all of that
space.
When
I
tests, I discovered that only 5-10% of the records
-47-
did
very
some
created
had any non-zero entries (and many of those
non-zero entries).
So,
had
very
without changing the logic
program, I could decrease the disk requirements
90%,
if
read
of
by
and
write
requests
and
the
manage
almost
space
I
way "fool" the program by not
record
storing
there was a non-zero entry.
Then,
requested,
the
if
it
wasn't
in
the
when
file
my
this way,
I could save eight of
that we currently required without
the
nine
affecting
myself
could
the
recognize that a record of all zeroes should
logic.
the
I were to write some routines to intercept all disk
independant of what the program was doing.
In
few
that
unless
record
routine
be
was
would
returned.
disk
drives
the
program
Compared to improving the processing, improving the
disk space requirements would be fairly
simple.
same time, it would isolate all data storage and
into routines that could later be replaced by
At
the
retrieval
DBMS
access
routines.
There were some other thoughts at
this
system was complicated, poorly understood,
use, and impossible to run.
point.
The
difficult
to
Many of the problems seemed to
stem from the original design flaws.
Rewriting the
system
would allow us to take advantage of all the things that had
been learned the first
since, and design for
changes.
time,
all
of
everything
the
we
changes
had
and
learned
proposed
It would also be possible to design the system to
-48-
be run as a series of
massive run.
transactions
The idea of
an
instead
allocation
expense that is identified by a coded
of
is
a
to
expense.
For instance,
be
charged
regional
take
location,
for
an
insurance
line, etc. and split it up among the locations,
we have determined should
single
etc.
that
of
that
part
administrative
expenses
may be split among the branch offices in that region on the
basis of the total nurber or size of
each.
policies
written
The processing was controlled by exhaustively
list-
ing all expense codes in tables, with instructions for
the money is
to be divided,
and which codes should
by
how
receive
the allocation.
Those tables amount to no more than a list
of instructions,
each saying "take the expense from there,
divide it
If
up this way,
and put it
here and here and here."
we could somehow process each one of those
alloca-
tion instructions separately, it would be possible to
stop
the system at any time.
each
Assuming that we had
of them as done when we finished processing
marked
it,
start back up without losing what had already
plished.
As it
was,
we
been
the system was trying to do
"in its head" instead of keeping records.
not only "forgot" which entries it had
If
down".
Ideally,
if
recorded after each instruction
-49-
was
much
stopped,
already
enough
accom-
too
it
information
processed,
it
processed,
but lost track of where it had recorded everything
"written
could
we
had
were
could
proccess each indiviually, or in
any' size group.
Then,
could have transactions that said "do everything
we
necessary
to report office X", or "run the next 5 instructions".
This would make it possible
small enough pieces to make
tionally.
it
to
break
much
more
any
run
viable
into
opera-
That, in turn suggested that we might be able to
allow on-line
inqiries
of the sort that
single transaction to answer a questicn
would
process
a
(such as: What were
the total expenses charged to fire insurance at the Houston
office?
small,
large.
).
The input
and
output
files
were
reasonably
only the intermediate work space was at present
If
all the data were stored in
on-line
could also allow on-line access to original
files,
and
we
processed
data to replace thousands of pages of reports.
capturing the output in
too
Simnply
by
a file, we would have the advantage
of being able to reprint the reports.
As the system was, a
lost report or bad heading could mean a complete 15-20 hour
rerun;
all of the output was printed directly to paper
a product of the ronstrous
final run.
The
as
massive-, terc-
porary files would probably benefit from being put
onto
a
DBMS .
I also realized that development of a new system could
proceed much faster than fixing the old.
The
reason
that where it was impossible to get more than one
test
was
of
the old system in a week, we would be able to test a better
--J
designed system as often as needed, both because
require far less resources to begin with,
and
could run much sraller, more specific, tests.
situation alone might make it
it
would
because
The
we
testing
faster to rewrite the
system
completely than to make just a few of the requested changes
to the old one.
At the same time it
seemed that
it
would
be much easier to rewrite than to fix.
In February, we convinced Rob West and the other users
that we
couldn't continue to run the system without making
some improvements.
space better.
I began to make the changes to use disk
At one of our meetings, we also suggested to
the users that they should consider a complete
rewrite
the system soon, since these changes would solve
disk problems.
posal,
We explained the
reasoning
only
for
the
but at the time they were more interested in
some of the improvements they wanted
any proposal
that
would
hold
up
than
in
progress
of
the
pro-
seeing
considering
for
several
months.
The last week in February we
1978 reports.
A
decision
by
shortly to run the system on a
instead of monthly.
completed
user
the
December
management
followed
quarterly
This was based on
requirements and long time delays in
reporting
the
getting
high
basis
resource
reports
out
making any more frequent schedule wasteful.
Sharon left in March for higher pay,
-51-
leaving me as the
prompt,
for
^fortunate
and he was
programer quickly to replace her.
start
pick
but
COBOL,
any
up
what
he
and
in
the extract programs,
He took over
needed quickly.
to
able
was a good progranmer and was
PL/1
a
Brad was able to
lie didn't know
two days before she left.
find
to
once
was
notice
her
to
rsponse
D
oss's
entire CAS group.
the next few weeks he learned enough about them to begin to
For the first
question the way they worked.
looked at the programs to see what they
began to find
problems
improve the reports.
and
were
resemble those on the systems we were extracting
Instead of
the first time.
Errors
in
began
to
from
to
they had in the past, the programs were changed
and total the rejected data.
to
errors
rejecting
blindly
CAD
to
reports
The totals on our
He
doing.
changes
suggest
someone
time
for
as
report
processing
the
logic were found that had made some of the reported figures
false for the past 3 years.
the confused,
patched
documented programs.
And he slowly replaced all
with
code
internally
consistent,
When I explained the deferred project
with the table changes to him,
further changes.
of
he understood and
suggested
So we reopened the project and he rewrote
the programs to complete it.
While I was making the changes to improve disk
I was also able to include
a
couple
of
improvements from their list of requests.
-52-
other
We
usage,
requested
were
slowly
convincing CAD that we really did know what we were talking
about.
We were serious about wanting to
improve
were really giving them results for the
the task force dissolved.
more
receptive.
was
We
system,
had
they felt
were
bothered to
needed
request).
discuss
(many
of
which
familiar with the problems etc.,
groups
and
this
regular
progress
other
So, when we again
subject of a rewrite in May, both
tested
begun
meetings with them in February to monitor the
changes in the system and to
since
We again approached CAD
with the suggestion that we rewrite the
time they were
and
time
system
In May, the
with a number of changes in place.
first
CAS
changes
they
that
had
brought
were
that the other
of
never
up
the
much
more
group
was
facing.
The discussions continued for the next two months;
were now meeting once a week rather than twice
a
we
week
as
what
might
be
required for a rewrite and what advantages there
would
be
to it.
of
before.
Frequently discussion
turned
to
Not only would we be able to include all
outstanding requests in the new design, but
would benefit from the latest in
were benefits to us,
as
the
program
people
who
a
new
turn
responsive to future need for change.
-53-
we
system
design.
would
maintain the system, but also would make it easier
to modify the system, so that in
their
could
These
have
to
for
us
be
more
They could run
what
they needed,
when they needed it,
presented
was not only convinced but
The first
management officially.
In
July,
idea
to
as well.
the
CD
their
week of Aucust their man-
agement decided that the idea was promising and requested a
feasibility study for a rewrite of CAS.
In
the meantime,
it
one of our source files had not
been
correctly
realigned
OK,
The 1979 data was
after the reorcanization.
that
May
in
had been discovered
but
input
December 1978 report would have to be rerun when the
data had been reworked.
The December 1978
was finally issued in early June, shortly
pleted the March 1979 run.
as the test of our changes.)
(The March report
The second
ports were completed at the end of July.
report
revised
before
com-
we
also
quarter
served
CAS
vention from us.
without any significant
Brad and I finished the
re-
From then on, the
system was capable of being run roughly a month behind
end of the reporting month,
the
job
of
the
interputting
the extract programs into production.
From this point on, most of our efforts went into, the
new CAS.
As it existed,
CAS was still
nearly
for the users and for DPD, who had to run it.
were not being used as planned.
However,
as
unmanageable
The
the
reports
former
head of CAD told me:
We thought perhaps for a change we had a system
that we could sell to other companies. It didn't
run very well, we knew it had its problems and
its limitations --
but we also knew it was better
-54-
'We had bouc:Iht so
had.
than anything an-one ele
many turkeys from other coraraniCs that I don't
think anyone would have felt any guilt.
Besides, even though managers weren't using the
Those reports
reports, CAS did serve a purpose.
were inaccurate, but they were the best we had to
prepare the (manually prepared financial reports
to the government) from. We hadn't planned it,
but those reports had become
critical
to our
operation. We couldn't have produced any kind of
estimates --
good or bad --
without a staff of fifty.
-55-
at that
much
detail
CHAPTER 4
THE NEW CAS
The new system developed slowly from its original conThere were a number of immediate problems that we
ception.
faced.
of
understanding
First of all, we had a very poor
Although
the existing system that we proposed to replace.
CAD
we were beginning to develop some credibility with the
department, Brad was the only full-time person programming.
Yet we were proposing to rewrite in a few
system that had taken longer than
months
a
orig-
develop
to
that
time,
If we could do what we said, there was no question
inally.
that the user wanted it done.
should
But why
trust
they
us?
By now they were fairly confident that if we continued
as we were,
that we could and
their system.
would
They were bargaining
sure thing, slow progress,
against
continue
several
a
risky
to
improve
months
of
a
proposition,
rewrite the entire system in that time.
In
addition, they would have to put a great deal of their
own
that we could
effort into the project if they agreed to go with our
sug-
gestion.
Through the
twice
weekly and
summer of 1979,
sometimes
we met with CAD and
more often.
-56-
We discussed
Sue
in
detail what CAD felt :hould be in any new system, what they
thought was wrong with the old, and how they might like the
system to handle each segment that we had identified.
the
users
in
many
ideas
as
so that we wouldn't he surprised
by
goal in the discussions was both to educate
what might be
available
possible on the table,
Our
as
get
to
and
a request that ruined our design later.
In
early August,
tion was expanCed.
func-
their
CAD was reorganized and
and
Anne was hired as Rob's assistant,
began training to take over most of the responsibility
CAS.
Bill and Dick were transferred
from other parts of the
assigned to Rob to assist
accounting
with
his
for
into
the
department
area,
and
were
also
duties.
other
The
meetings became especially important, therefore, as a means
of
bringing
Rob's new staff up to date and explaining CAS
to them.
Having Anne in
the
to
make
Where Rob West had been
with
department
user
things easier for us in MIS.
began
CAS since he joined the task force, his thinking tended
to
be very much in terms of how the current system worked.
In
dealing with him, we had to always be careful
accounting terms he was familiar
with,
processing terminology confused him.
use
to
because
data
the
And we had to
the
always
old
relate our discussions to concrete examples
from
the
CAS (which we weren't able to do easily).
Anne
provided
-57-
and was quick to understand our sug-
thoughts,
some fresh
gestions for change.
She was an accountant,
terms that Rob understood,
processing than he did,
with us.
but also knew
which
data
about
more
facilitated
in
and talked
communications
As she learned about the old system, she was able
to provide the concrete examples that helped Rob understand
us.
She quickly took on the role of
between
intermediary
the two groups.
actually
Shortly before the CAS feasibility study was
approved in mid-August, we began the design of the database
files
design
Because none of us had ever tried to
files.
for a DBMS system before, we decided to use a formal design
technique.
of
Lou had seen some recent mention
a
referred to as Entity-Relationship Modelling (E-R)
computer magazines,
1977)
method
his
in
and located a copy of an article
(Chen,
that explained it.
The E-R approach involved
structure of the firm from the
separating
data
being
underlying
the
important elements were defined as Entities, and the
tionships that existed between
Any
included.
A
diagram
to
file structure was then developed from the E-R
allow data to be captured about any entity or
that fits
Rela-
them were diagrammed.
relationship that might affect reporting was
as needed.
The
captured.
This allows data to be
captured
the underlying structure of the firm
relationship
in
a
manner
instead
of
in
collecting it
t.Ie
specific
a
database
The result should be
application in question.
that is
for
suitable
only
a form
each
for
forced to duplicate the data in a different form
that we
We decided after reviewing the paper,
application.
being
of
instead
useable for any reporting needs,
out
would attempt to use E-R and see what we could get
of
it.
joined us the first
She
afternoon.
up,
and
time
she
Sue was invited
and spent all of our time in it.
room
conference
For the next few days, we reserved a
spent
what
could with us until the design was completed (several hours
The CAD
each day).
twice-weekly
joined
staff
and
meeting times,
us
especially,
Anne
they,
normal
our
during
spent a lot of additional time with us as well.
The
sessions
very
were
informal,
described as "brainstorming sessions".
best
probably
Lou and I, having a
better idea of what we were trying to accomplish, generally
Everyone
served the chairman function when it was needed.
seemed to participate freely, each taking the floor to make
suggestions, explain points, etc.
with charts and diagrams,
out ideas on.
We were
The
was
covered
and we had a blackboard to sketch
constantly
redefining
standing of the E-R approach as well as
resulting diagrams.
wall
Slowly,
a complete
gram for the whole financial area of the
-59-
our
under-
restructuring
relationship
firm
was
the
dia-
devel-
opea.
Once the E-R diagram was completed for this segment of
the firm,
the
we began the more technical job of planning
required
actual data files we needed to capture the
data.
we
had
completed it, we again net with CAD and Sue to explain
how
This task was handled by the CAS staff,
the new structures would work.
and
when
We then took the final stern
and
showing
how
of taking our actual current data
files
they would fit the new structure.
By demonstrating that we
could do that, we satisfied ourselves and our users that we
had developed an adequate E-R diagram to meet
all
current
and planned data needs for CAS.
At that point we went back to a more normal
operating
We met with the Database group to discuss the use
routine.
of one of their DBMS software packages
files we
had
developed.
to
Bill Vernon,
store
the
the
CAS
of
the
head
Database group, was present at that meeting and after
some
discussion about our plans, he announced that we would
ADABAS
on us.
and
gave us a list
use
of restrictions he "had" to, put
Since Lou's boss had already told him in a
meeting
the day before that we would have to use ADABAS for a
of that size, the news came as no surprise.
about most of
the
restrictions
Vernon's staff earlier.
as
well
We
in
had
file
heard
talking
But, this meeting served
to
both groups officially aware of the conditions we would
-60-
to
make
be
doveloping CAS under.
During
September,
feasibility study.
of
our
review
the
minutes
We looked back at the
meetings with CAD from the past few months,
to
CAS
the
on
work
continued
we
items that they felt should be important in a new CAS.
Our
reviewing
what
meetings with them continued twice weekly,
they had
already told us,
fication of some
and getting new ideas and clari-
confident of how to proceed,
fairly
The MIS team was
earlier points.
concrete
and began to develop
We laid out the
plans for a system to accomplish our goal.
basic processing steps that would have to be at the core of
any such system.
Then,
based
on
requirements,
user
decided what the initial system would logically contain
either because they were basic
required
features
user or because they would have to be included as
by
part
we
-the
of
the core of the new system anyway.
We then proceeded to set up a
basic
system and developed design rules for
design
individual
for
the
modules.
The system would be highly modular to allow for flexibility
as well as for the programming advantages.
Functions could
be easily added or changed by inserting or
modifying
ules in the basic system.
processing of individual
The system would be based on the
transactions
to
accomplish
smallest discrete piece of work possible at one time.
would offer the
mod-
advantage of making the system
-61-
the
This
interrupt-
able.
less of which outputs
and
needed,
really
were
reg-ardwith
no
instead, would
This,
chance to check interr'ediate results.
rs
hr
proqram ran unstoppably for
The old
run only those transactions that were needed to produce the
desired results,
split into as many steps as necessary.
actually began design
and
of
coding
up the file descriptions for
would need,
the
files
ADABAS
three
core
We also set
order to test out the design rules.
modules in
the
of
some
"e
we
and requested the Database group to develop the
interface modules for us.
problems
en-
we came up with a scheme
for
storing data separately from the codes for the office,
cost
' In
order to isolate our system from the
countered in
center, etc.
reorganizations,
What we did
that the data belonged with.
to establish an additional set of cross
reference
set of tables was to translate the codes
The first
was
tables.
in
the
feeder systems to a common numbering system, for processing
and reporting purposes.
translate
then
The new set would
those codes into an arbitrary set of sequential numbers for
data storage.
This way,
the
data
codes whose meaning was limited to
file.
in
would
locating
The organizational hierarchy and
the reporting codes remained separate,
changing the cross reference tables,
report in
be
we
other
data
in
with
the
information
so that by simply
could
any organizational format desired.
-62-
stored
immediately
to
The end of Sctemtber was then snent trying
lish estimates of how much
and
time
quested features into
several
difficulty, desirability, etc.
phases,
three
We
the
split
groups according
to
be
re-
need,
After establishing what the
minimum requirements would be for
split the rest into
would
resources
required for each part of the project.
estab-
an
logical
initial
system,
we
development
follow-up
for
and then made estimates for time and resources
each.
The idea here was to provide a plan that
quick initial
next.
development,
Living
with
the
would
and flexibility in what
initial
system
would
suggest some group of changes that would improve
fulness most for the next phase.
the record that this was intended
allow
to
add
doubtless
its
use-
But we also wanted it
to
be
an
on
evolutionary
development, and would not stop after the first version was
delivered.
We therefore needed to document some
plan
for
future phases and resource estimates, even if they might be
completely revised before we got there.
The last week
of
September,
the
feasibility
results were written up as a proposal to proceed
study
with
first phase of a proposed four phase development of
Cost Analysis System.
The
initial
phase
would
a
the
new
quickly
replace the functions of the existing system, and following
phases would add on the other features
-63-
desired.
A
long
rance plan
to
reiace
all
was
systcms
financial
also
sketched out.
Sue and Lou put together most of
and developed a
for
presentation
management.
user
emphasized the capability and flexibility of
design we had developed and the
design.
critical,
It
document
final
the
the
database
modular
proposed
It
program
also stressed some points that the user thought
such as auditability and testability,
that
would
A six to eight month period
be features of the new- design.
was suggested for the first development
depending on specific options chosen.
to
be
completed,
The technical issues
involved and specific content of the proposed system (aside
from the selling points) were left vague.
proposal that the design
similar rewrite of all
be
considered
financial
It did include a
as
systems
a
if
basis
it
was
for
as
successful as we expected.
The following week, Sue made the presentation
and their management.
When they
reponded
to
CAD
favorably,
she
and Lou helped CAD management revise it for presentation to
the senior corporate officers who would be
Newcomb, head of the financial reporting
presentation in mid-October.
It
varied
affected.
area,
very
Ed
made
that
little
from
the original, but emphasized even more the future potential
of the new design, and dwelled
content of the first version.
-64-
even
less
The immediate
on
the
actual
response
was
guarded, but positive.
During the early part of
the
month,
Lou
discovered
Stan at an open house recruiting session, and hired him for
CAS.
Stan had
no
data
processing
bright and had taken PL/l courses.
experience,
but
was
a
very
He looked like
promising trainee candidate.
Also in October,
face modules.
we received the first
ADABAS
We did the necessary programming
to test the ADABAS routines.
It was
inter-
and
iraediately
apparant
that there were problems, and we requested help in
solving
them from the Database group,
since the problem
be in
We were told that they
the interface modules.
get to it
when they could, as they were very
began
seemed
busy
to
would
at
the
time.
Meanwhile,
we were making some initial
progress with
design and coding of modules for the new system.
Since
had no access to
interface
the
ADABAS
problem was straightened
files
out,
developing the translation
codes from feeding systems.
until
the
we
were
concentrating
tables
that
would
match
The tables would be
for loading the data files into ADABAS, since
to use our own translated codes for data
would need them to translate the data
we
storage,
back
for
we
on
the
necessary
intended
and
we
reporting
purposes as well.
Unfortunately, Brad and Stan didn't seem to be hitting
-65-
it
but that hadn't become a real problem yet.
off too well,
the
new
Financial
Systems group as a result of the breakup of
SMD
that
Sue officially transferred from SMD to
had
little.
been announced in August, but her job changed very
we
Stan was assigned the job of defining all of the systen
We
had designed into the data dictionary.
this as a quick introduction
what
to
him
for
the modules.
Use
of
to
ability
expected to further improve our
cations to the new system easily,
of
all
was
system
dictionary
data
the
been
had
in
accomplished so far, by involving him somewhat
use
to
hoped
modifi-
make
by allowing us to quickly
find all of the modules affected by any change.
we
in
made
a
We were to follow the
rewrite of the Cost Analysis System.
recommendations that
with
proceed
to
Controller and the Treasurer jointly,
the
by
In mid-November we were officially requested,
proposal for Phase 1 of the project.
study
feasibility
the
that
requested
They
the project begin immediately and be completed by June
30,
1980.
We immediately ordered the terminals
ADABAS
The
would require for on-line use of the system.
users
our
that
interface still wasn't working, so we again asked for
to get it
fixed.
We also
group that handled CICS
requested
interfaces,
terminal session monitor and
control
-66-
assistance
which
from
help
the
would. provide
facilitics
for
the
user terminals.
(211 user dcnartment terminals run through
the CICS monitor system,
which
accepts
forwards it to processing- programs.
with
proper
authorization
to
typed
input
and
CICS only allows users
run
each
program,
which
provides a good measure of control.)
Design
and
of
coding
CAS
modules
reasonably well.
Meetings with CAD were
the specifics of
what
formats,
situation:
reports,
ie,
they
wanted
prompting,
to
etc.,
was
proceeding
now
focusing
see
in
a
on
given
and exact contents of
what fields do we need to
add
up
to
create
this reported number.
December arrived and progressed with
little
We still
major to show.
had no access to our ADABAS
files, so we couldn't put together the quick prototype that
we had hoped to provide.
Toward the end of December, there was still nothing to
show for our effort,
and Rob and Sue were beginning
to
be
visibly uncomfortable with not having any specifications in
writing.
Over the next couple of meetings,
CAD
their view of what they expected in
the new CAS.
turn,
how
presented an explanation
of
the
presented
Wd,
system
in
would
accomplish those tasks.
Finally, just before Christmas, we got some help
the ADABAS problem.
with
The solution was to give us an all new
interface, which operated
other had been intended to.
somewhat
differently
In early January we
than
the
got
some
to
successful tests accessing the ADA2AS files, but due
couple of major changes we had just made to
our
programs,
hadn't received any response from
We still
CICS group,
and approached them aqain.
a
completing
we needed to fix the existing nodules before
prototype.
a
the
They announced that
decided
Lou
they would get to us when they had the time.
that we didn't have the time, and sent Andy to a CICS class
for a week, hoping that he would be able to help us set
up
the interface ourselves afterward.
Brad and Stan seemed to rub each other the wrong
way,
with
each
and were unable
other.
in
programs
discuss
to
good
They had had several arguments for no
the past few weeks,
programming style.
reason
usually about unimportant points
communication
that program development could
between
continue
a
at
of
trying
More of my time was being spent
to provide the necessary
pace.
quietly
them,
so
reasonable
All too often, each would make assumptions about how
the other's module would work that turned out to be wrong.
The problem was not helped by the fact
Brad took
was working at different times.
and worked Saturdays,
Wednesdays
and he started mornings at
he could leave an hour and a half early in
Stan worked normal hours,
and
I
tended
evening and weekend hours during school.
everyone
that
the
to
seven
so
afternoon.
work
mostly
Because the
of the system still wasn't working, each person was
-68-
off
core
likely
and find a program critical to his
to core in
working.
person
If the
responsible
wasn't
testing
not
around,
the
options were to be frustrated, or to try and fix the
person's code.
So it
find that the code
changed,
were
last
working
on
had
been
or that someone else had caused a new problem
your testing.
At
and either
was not unusual to come in
you
other
the
Brad was about ready to quit.
end
of
January,
having
was
Lou
problems
group.
locating an experienced PL/l progranmer to join the
We decided that even a trainee
programmer,
for
would
and he hired the next
be
better
PL/l
promising
programmer who applied to the company.
than
no
trained
Carol was quiet and
pleasant and seemed that she would be able to work successfully with Brad and Stan.
But it
seemed that
between the two bothered her, and although
much better with them than they did with
the
she
each
tension
got
along
other,
her
working relationship with each of them was less than ideal.
Stan had been elected to write an interactive
for our terminals that would run CAS,
CICS help, so that we could try to get a
running for demonstration.
He was
getting that to run properly.
conflict with Stan,
started work on
in
could
prototype
still
Brad,
we
until
program
having
system
problems
an effort to
revising
the
get
avoid
extract
programs to provide data properly for the new system.
Sue had persuaded the group
-69-
to
develop
more
formal
wrote
up
the
together,
and then we reviewed then
in
the
documented in
set
complete
fairly
a
February
that way.
he understood the
couple
series of modules to
that
he
couldn't
use
con-
had
He soon discovered that once he
verted the programs,
we
of part of the system to work
modify
He then tried to
run under CICS.
him,
to explain the system to
weeks of trying
finally just gave him a copy
with.
After a
so he was coming into this cold.
design sessions,
our
of
any
on
he had never sat in
Unfortunately,
help.
much
to him before he could be of
system
the
had
But we
well enough to make some suggestions.
to explain
was
reauirements
of
When Andy cane back from CICS class,
problem
early
By
needed.
changes
any
on
meetings and agreed
Sue
CAD and
directed at finalizing those.
specifications
nowi
were
meetings
Our
system.
specifications for the
because
them
the
necessary interfaces between PL/l and CICS hadn't ever been
used in our shop.
give us help,
but
We again requested that the
this
time
limited
group
CICS
providing
to
the
missing interfaces.
However,
Andy's
suggestions
about
how
accessed under CICS, etc. allowed us to make a
our
decision
cross reference tables onto the RAMIS DBMS.
-70-
design
few
changes that would simplify the conversion later.
important of these changes was
were
files
to
The most
put
This is
the
some-
thing we had pianned Co do
wouldn't allow
the
files
between
that we get an interface
RAMIS set up.
group
it
We had access to the data in
couldn't access the conversion tables,
new
this
With
until
core
we were again without a working system
worked.
for
time
Instead, they offered to give us the infor-
mation to write our own interf ace modules.
change,
assis-
for
tance, but they wouldn't be able to give us much
several weeks.
and
programs
PL/l
our
Database
We went to the
however,
required,
This
decided to proceed immediately.
we
using,
currently
were
we
CICS
because
but
eventually,
now
but
ADABAS,
didn't
and we still
have a working monitor program.
We went to talk to OAL about getting the data that
needed from their file.
didn't seem to be any real difficulty in
needed.
what
getting
them
from
information
we
nearing
Brad's work with the extract programs was
the point where he would need that
there
but
They sounded confused,
we
soon.
Our new terminals arrived in
but the user's
terminals
continued on with nothing
March.
It
was
IIS, several months-late,
still hadn't been delivered.
show
to
for
into
effort
our
We
discovered that no one in the firm had ever
used the PAMIS interface with PL/l
would have to be generated in
interface would work.
before,
and
the RAMIS system
Again there were delays.
-71-
a
module
before
our
The terminals
vwre
finaily
and Andy began
CAD.
The
successfully.
interface were delivered by IBM
CICS
testing
in
tested
and
RAMIS interface was established
The programs for the
installed
on
Meetings with the user
those.
were changed to once a week due to lack of anything to talk
about.
Brad
Relation-s within our group remained strained;
was again talking about leaving.
Most of my time was taken
to keep the
up with trying to keep up enouch communications
group working.
By early
progress.
April
we
were
Stan's on-line monitor was
and
fixed,
of
signs
showing
finally
it
was
beginning to look like we would soon be able to edit tables
and a few other simple functions on-line.
a few problems in the programs, but
the
reaching the stage where problems were
to the function being tested.
the interface and in
RAMIS and
ADABAS
the
There were still
system
core
generally
was
isolated
CICS problems were now
actual
interfaces
programs
being
past
tested.
seemed to function well.
It
seemed that we could finally spend our time working on, our
system instead of interfaces with others.
The old system was run for February
for the
first
time
since
problem was in the allocation
program.
we
1980
and
failed
started the rewrite.
processing in the
When I found the problem,
big
it turned out to
The
PL/l
be
a
weakness in the original rules for allocating expenses, and
it
had died tryimn
happened
that
;rocess legitimate data..
to
that
occurred since
data
combination
hie system was first
had
run.
It
had just
never
before
The solution
simply for the user to change the control tables to
the illegal calculation.
again,
They made that change
but this time we ran in
We had more
attempts,
and
successful run at the end of April.
March reporting and had similar
huge resource requirements.
bypass
and
tried
to operational problems,
our files were destroyed mid-run.
problems for the next two
was
operational
then
finally
We then tried
problems
and
to
run
of
the
nearly
all
because
All told, I spent
a
of my time working on the old system for a month.
When I returned to work on the
that communications had
modules that had
been
Within a couple days,
problems.
broken
working
new
down
system,
I
found
again,
and
several
work
any
longer.
didn't
we were able to isolate and solve the
We got the table edits working
successfully
on
our terminals,
and invited the CAD staff to come up and try
them.
the first
thing we had been able to
week.
April,
This is
present
them.
The first
behind schedule,
in
the
system
and far from complete.
visible progress, and all
finally out of our way.
of
the
known
Andy finally had
run successfully under CICS and was
was very
far
But we were making
roadblocks
a
preparing
PL/l
to
were
program
convert
the working part of C r systCm.
There w-as
even
a
chance
that we might meet the June 30 completion date that we
originally proiised.
-74-
had
CliAPTER 5
PROBLEMS ENCOUNTERED WITH CAS
This chapter looks more deeply at some of the problems
we had in dealing with other groups in
is
mostly on why,
The focus here
MIIS.
these
from their point of view,
problems
might have occurred.
Technology
all
hardware and systems
programs
--
operating
telecommunications
programming languages,
for
responsible
The Systems Technology Department is
systems,
etc.
software,
This includes providing and maintaining these facilities as
well as providing technical expertise to those using them.
difference
of
opinion between the head of Technology and his boss led
to
In
the
summer
his being fired.
of
1978,
a
personal
Shortly afterwards, a group of his
responded to what they perceived as
themselves quitting.
This cut
the
an
unfair
programming
half and essentially removed the entire
staff
firing
by
staff
in
management
group.
The remaining programmers in the group were very overworked
and, to keep things running, several of them were asked
take on management responsibilities as well.
Many of
to
them
were very technically oriented and balked at being asked to
-75-
serve
as
managers.
unfamiliar workload.
",one
were
times,
extra
the
For the next couple months
of system problems resulted.
behind,
with
happy
As Technology's
a
nurber
work
the online systems slowed to unacceptable
batch jobs began to take days to get back,
fell
response
and
many
ongoing projects were simply frozen because no one had time
for then.
New equipment had been installed by IBM
in
June
July and had to be signed for soon or it would be
and
removed.
No one in Technology had had time to fully test it yet, and
no one left really had the authority to sign.
None of them
wanted the risk of signing for untested equipment to be
his or her own head.
Reportedly,
someone
finally
the problem by signing the fired manager's name
on
solved
one
night
to avoid losing the equipment.
Not surprisingly,
workload,
under
without management,
pressure
and with little
heard from other departments, one
by
one,
Technology staff also began to leave.
to fill
the vacant slots,
and
but it
an
increased
but criticism
the
remaining
The firm was
trying
was taking tine.
It was early 1979 before the system problems began
be resolved.
A management structure was slowly built
the
up, mostly by hiring from outside
clear
whether
anyone
outside
exception of two DPD people who
-76-
of
company
Technology,
actually
took
(it
with
to
back
isn't
the
positions,
werc
offered
a
Chincic
to
rove
into
department.
the
However, no such offer was ever made public).
The
remaining within Technology without exception
refused
offers made to them.
All who were offered
management emphatically refused.
gested to me were:
1) they
want to be managers,
and 2)
the chopping block.
T
rhe
were
The
staff
any
advancement
reasons
technicians
they
and
to
sug-
didn't
they didn't want their necks on
managors seemed to pick
of the blame for any problems,
of
and most
them
most
up
had
had
enough of being the scapegoats.
Toward the summer of 1979 there were rumors
about
the MIS division that a third mainframe computer was
ordered.
installed.
being
And, in fact, in early October a new machine
It was bigger and faster than the two
machines and was to
be
dedicated
to
testing.
assumed by the Technology staff that such a
in computer resources, nearly doubling
would solve many of the problems of
was
existing
It
was
large increase
existing
long
in
capacity,
online
times and unacceptable batch turnaround tines.
simplest solution to a pressing problem, and
convinced the people who controlled the funds.
response
It was
the
the
argument
(A pool was
started in APD betting on the number of days it would
take
for the new machine to be as saturated as the old.)
It helped for about a month.
began to do more work,
The programmers
quickly
taking advantage of the improvements
-
7 7-
to
Su;mit
reasonable time.
in
in
the hope of seeing results
more tts,
zith
so
The online environment improved
a
that
the same amount of time using a terminal they could rake
more requests
of
Very
machine.
the
machines were full to capacity and the
three
all
soon,
returned.
problems
New measures were called for.
There had been an informal online users group for
the
of
the
It
past couple of years.
of
consisted
several
junior managers in APD who had people trying to
program development.
They
met
had
do
with
regularly
with
the
improvement would in turn improve proqramer
had
sent
memos
and
occasionally
But, until late 1979,
of
iman
the
much enthusiasm from Technology.
group
productivity.
met
Technology group to suggest things that they
with.
goal
be
They believed that such
proving system responsiveness.
They
each
could
other to discuss ways that the testing resources
used more effectively, especially
online
had
the
with
up
had
come
not
received
The group hadn't yet been
able to push through any
significant
with the change of staff
in
pressure for solution of
the
changes.
Technology
system
and
However,
the
problems,
group was suddenly taken more seriously.
growing
the
user
They began having
weekly meetings with Technology staff towards
the
end
of
December 1979.
One of the first
results was a plan to try
-78-
puttinc-
a
terr:.inal on every progjrarmecr' s
desk.
The
would be restricted to make -only the
efficient
was felt
that
It
terminals on every
encourage
would
other wasteful testing practices (such as
single
test
at
com-
putting
prograrmers
reduce their usage of printed output and cut back
eral versions of a
usage
most
mands available to programrIers.
desk
online
to
on
many
submitting
sev-
once,
so
that
when
results were finally returned there was some hope of having
a solution among then).
The hope was, that although online
usage would be
increased,
greatly
there
would
be
more
efficient use of available resources and an overall drop in
the load on the computers.
CAS was chosen
concept.
as
a
pilot
group
for
testing
the
This was because we were beginning a major devel-
opment effort; because we were believed
to
intelligent users of the system already;
be
and
heavy
but
because
our
manager was the ringleader of the user group that suggested
it.
Our terminals were scheduled to be
end of December 1979.
A few other
installed
programmers
selected for the pilot group, on
roughly
(especially that their
were
managers
the
part
at
were
same
of
the
also
basis
the
user
group).
Due to technical problems encountered
the necessary hardware
and
Technology group was unable
software
to
-79-
get
for
in
the
everything
setting
test,
ready
up
the
to
support the new t
installed,
February
week of
hardware until the first
necessary
the
of
all
deliver
turned out, was unable to
aryway.
the test group terrminals more than
it
as
IEM,
until late Fcbruar-.
r-,Ils
When
the
doubled
number of terminals connected to the test machine.
In January, when our user's terminals
scheduled
were
out
what
the status of those terminals was and discovered that
they
to be installed, we contacted Technology to find
fact,
In
had "forgotten" to order them.
they had forgotten
other
to order terminals that had been ordered for several
user departments as well.
It appears that they had decided
the effect of the new programmer terminals
(even
though
the
user
were
terminals
until
for
called
that a moratorium on user terminals was
determined
was
to
intended
Le
connected to an entirely different machine!).
They imnediately ordered the
missing
when
terminals
they "discovered" their mistake, but by that point IBM
backlogged on orders.
It seemed
unlikely
However,
have terminals for users until late May at least.
just a few weeks later, it
that some of the
with newer models.
was noticed by
an
SMD
manager
terminals were being replaced
programmer
It was "suggested" that the
department install the old
would
we
that
was
terminals
Technology
temporarily
user departments rather than send them back to IBM.was done in early March.
-80-
in
the
This
Considerinc
the problens that Technology faced already
with extremely heavy machine usage,
it
too surprising
isn't
that they might be reluctant to install more
Where programmer terminals
user sites.
reduce the machine load,
user application would,
terminals.
might
e specially if
it
after the forgotten order was brought to
halt
the
installations.
that
postponement on the basis
new
online
required
January,
shortly
attention,
their
They
wanted
they
didn't
have
the
impact
on
information to adequately assess
in
conceivably
it was very unlikely that any
There was an attempt made in
to officially
terminals
a
enough
existing
systems, or the resources that would be required to support
them.
Although they had initially approved our request
October when we made it,
they didn't feel that
go ahead with installation of our users'
had better explained what
facilities
we
they
in
could
terminals until we
had
planned
on
providing through the terminals and what volume of usage we
envisioned.
But,
when we were able to immediately
provide
that information, they dropped the effort to halt our- user
terminals.
OAE
The OAE group,
group within APD.
like CAS,
They are
is
a
financial
responsible
of
the
company's most important internal operating documents,
the
-81-
for
one
reporting
t is
Operations 7.nalvsis E:hibit.
a 'IMnthly
that
report
matches premium income and claims against policies to
underwriting
The system accounts for many of
insurance business
and local
of agents
profitability
(such as the practice
of
por-
risk),
and
carries a lot of highly detailed information to
reports
that
it
are
generates
used
the system's size and perceived complexity,
The
CAE is
reports and their good record for on-schedule
of
fairly
significance
reports each month means that they are also
many
by
Because
managers to measure and monitor performance.
well respected by those in MIS.
provide
etc.
a number of reports showing office workloads,
The
the
of
selling
tions of policies to other firms to spread the
it
offices.
complexities
the
show
of
the
delivery
well
of
regarded
outside MIS as well.
The OAE group provides 3 tape files to CAS each month.
These 3 files, along with 3
additional
files
from
systems, together provide all of the information
other
necessary
to allocate costs within CAS.
We approached the OAE group about
the
availability
of additional information from their master file during the
course of our feasibility study.
Our needs
at
that
were fairly vague and the contact was obviously
feasibility study.
gathering
We made it
clear that
information rather than
-82-
we
intending
time
due
were
to
to
a
simply
use
the
results in the- near future.
positive;
yes,
the
from their files,
Thcir
information we
and yes,
wanted
they could
without too much effort or warning.
We
simnly
was
provide
vcry
available
it
for
Since our mission
simply to get needed facts, we did not
the conversations.
were
responses
think
incorporated
us
was
to
document
our
findings
that the required information was easily available into our
planning.
I think it was Rob West who,
during the database
de-
sign sessions, suggested that it was beginning to look like
we could take over OAE.
The files would
include
the OAE data anyway, and our design included the
much
necessary
provisions to duplicate much of the OAE processing to
duce some of our reports.
The MIS people present
pro-
probably
took it more seriously than he did, because technically
really did seem feasible.
at the time,
the following
The suggestion was
of
laughed
it
off
but was brought back up several more times in
months.
Sue, our
SMD
representative
mented that that was how empire building
was
com-
accomplished
in organizations, by gradually replacing other groups.
original suggestion was revived yet again
when we began having problems getting
the
in
March
The
1980,
cooperation
of
the OAE group.
In January of 1980,
we went back to OAE
looking
for
help in finalizing plans to use the additional information.
-83-
months
Brief discussions were carried on over the next two
between OAE and CAS.
in the meetings.
agreed on
to record exactly what was
to
bothering
was
But, as yet no one
change from day to day.
seemed
us
to
resronscs
Their
them.
from
etc.,
descriptions,
Ue needed file nar;es and
meeting
one
We slowly began to realize that from
to
would
the next the files that they recommended that we use
change, that we didn't know exactly how many files thay had
or what exactly was in
one
exclusively with
any of them.
person,
programmer in the group.
We were dealing almost
Jerry,
Jerry seemed
who
was
a
senior
to
have
a
better
knowledge of their system overall than anyone else that
Still,
had dealt with.
getting a
it
seemed
that
either
we
were
we
runaround or he didn't really understand what we
needed.
He had once suggested that we use his source file
and
duplicate all of the daily processing done by their system,
as necessary, to produce what we needed.
he insisted
that
At
it would be far more practical to use an
appropriately summarized and processed file,
or daily data.
times,
other
with
monthly
Yet other times it seemed best to him
that
we duplicate a small part of their processing to use .a file
at a much more "appropriate" level of detail.
The possible
files at that level of detail as he described
them
somehow to have the same detail as the
most
seemed
detailed
and
also as the least
cetailed.
week in March,
Finally, the first
on some of
sat in
enough was enough,
with
OAE)
and started recording her
of each meeting as it
memo to Jerry and
meetincs
our
(who
Sue
She
was held.
several managers
also
decided
that
understanding
sent
who
had
the
resulting
might
be
inter-
ested.
we had a
Two weeks later (and two memos later),
tative set of files, their contents, and a
ten-
description
which processing exactly we would have to duplicate to
our extracted information to the
etc.
right
level
of
of
get
detail,
But, suddenly Jerry was unable to find a half hour to
meet with us "because of the heavy demand"
for
his
time.
So we were again being held up; we had almost enough information to write specifications for the
new data, but not quite enough
extraction
without
talking
of
to
the
Jerry
again.
From their point of view, it
now
seems
they would not be too eager to help us.
a large, many faceted systera.
corporate reports and has
system supports a
commands
priorities
importance of its
the group
large
enjoy a
always
number
over MIS
output.
It
obvious
The OAE system was
provides several critical
performed reliably.
of
programmers,
resources
As a result,
because
and
of
the programrmers
certain amount of prestige in
-85-
why
the
The
it
the
in
MIS
orcanization.
That teir
syster
runs well is
bugs
major
have
a shop where many of the systens
tant in
i7r[or-
:lSo
and/or operational problems.
We essentially came to them
the blame.
were
The OAE group might get some credit if we
that they would get credit for.
project
unimportant
rather
and
their time
Otherwise,
credit
no
them
effort in helping us would officially get
new
the
provide
to request them to do the programming to
data, but it would be a small,
a
should fail, we take
it
success, we will get the credit; if
our system is
If
having nothing to offer them in return.
help,
their
for
asking
whatsoever.
Meanwhile, we were asking to take a
system.
their master file than at present, and
going back to their source data and
ficant part of their system.
we
even
duplicating
their
of
more
copying
be
At the very least, we would
part
of
discussed
signi-
a
That would be valuable to us,
skip
since we would get the data a month earlier and could
a number of processing steps that they
require.
Without
adding a lot of overhead to our system we would be able
start processing their inputs to
our
system
earlier, making the timely production of our
three
reports
to
weeks
more
likely.
But, in doing so,
we would not only reduce our depen-
dence on OAE, but we would have an opportunity to show them
-8G-
ready.
were
up by publishing our reports before their own
that
Because we were using advanced programming techniques
their systen did not incorporate,
we also might demonstrate
how inefficient their system is.
Although the plan to
CAS
as
a
before
test
first
financial
all
converting
it
applications had not been widely publicized,
were to fail, their system would remain untouched.
we
But
if
be
a
dupli-
be
good candidate for conversion, because we would
cating part of it
if
to
demonstrated
we did well, their system would be
quite
is
And,
possible that there had been rurors about it.
use
anyway.
The Data Base Group
for
The database group in MIS has responsibility
(DBMS) and the data dictionary.
database management systems
This is
the largest group in APD
function outside of MIS.
They
without
corresponding
a
pro-
for
responsible
are
viding interface modules, maintaining the integrity of
DBMS files, assistance
data
dictionary
Bill Vernon,
design
in
facilities to proper use,
all
efforts
to
these
put
DBMS
and maintenance of the
all
and
software.
the
manager (VP level).
group,
is
a
senior
widely considered
to
be
bril-
of
head
le is
the
liant, but most people in MIS seem to
and difficult to deal
He
with.
-87-
find
obnoxious
him
apparantly
is
a
good
problem solver to
but
explain his power in MIS),
a
in
around
h. c
The
more time at lunch than he does workincg.
spend
group,
like
of
few
very
most of MIS has had major turnover problems;
them
to
tends
otherwise
r-aV
(which
crisis
have been with the company for more than one year.
Our dealincs with the database group have been
pretty
of
delays
because their workload is heavy,
There was one
program that we needed.
problem
decided
1979'when Vernon himself
to
friendly
have been
period
in
September
write
an
interface
It didn't work,
wouldn't
he
and
to the rescue,
suggesting
necessary
was
unable to get the information from him that
were
We
fix it or give it to one of his staff to work on.
for us to fix it ourselves.
that
but the individuals
we have dealt with (aside from Vernon)
and helpful.
lot
a
There have been
successful in general.
However, one of his staff came
that
we
try
a
interface that she wanted a chance to test.
of
Although there
we
than
was a longer learning period involved for us
hoped for, this averted a major holdup, and
type
new
by
new interface we were able to deal directly with
had
using- the
her
from
with
the
that point on.
Some of the more
interesting
database group have been indirect.
ADABAS DBMS to store our data was
our senior management.
interactions
Our decision to use the
essentially
ordered
Apparantly the basis for the
-80-
by
deci-
the past
(the
data-
base staff members I asked were unaware of any
such
study
sion was a study dcne at son'-e tim.e in
in
two years --
the last
at Worldwide --
are updated
appropriate
but agreed that the choice was
from what they knew.
But,
been
has
the longest any of them
since both the available DBMS's
several times each year, the results of such a
study could be quite wrong by now anyway.)
Of the restrictions placed on
of
use
our
ADABAS,
to
some, such as tie number of files we were allowed
use,
were seemingly arbitrary decisions offered by our managers.
head of the database group,
about them from him.
we
though
even
from
originated
Most'often such rules seemed to have
earlier
allow
would not
hear
didn't
For instance, we were warned
by one of his staff that Vernon
the
to
us
use more than 3 or 4 files, but the actual restriction
was
placed on us in a meeting with the manager above him.
The important thing here is that these decisions
were
we
were
made for us without any real
trying to do or how.
knowledge
of
what
They were handed to us
as
mandates.
later,
Yet, when we asked the people in the database group
there seemed to be logical explanations for
(some made with incomplete knowledge however).
not been
offered
explanations
at
the
management decisions as far as we were
simply had to design around them.
-89-
decisions
the
time;
But, we had
they
concerned,
and
were
we
6
C17PTER=
CONCLUSIONS
We tried a new (to the corpany)
design.
supplemental
necessary).
work
with
project
SMD
to
develop
requesting
the
request
(and
department
when
project
the
formal specifications bascd on
in
for
was
The standard procedure
approach
Then the programning group would work directly
project.
from the completed specifications to complete the
We,
however, attempted to begin
work
specifications to restrain the design.
to
was
idea
The
preset
any
without
give them a "quick and dirty" prototype system that
solved
some of their problems as soon as possible, and then to use
that as a basis for further
discussion.
We
then
could
and
evolve a design that met their needs, and then go back
"clean up" the programs
in
each
section
as
the
design
became finalized.
This was attempted for
several reasons.
First;' the
project was not well enough understood by either
us,
SMD,
or the users to provide detailed specifications.
Rob
West
the
had the best knowledge of the old system, and probably
best idea of what was wrong with it from the user point
view.
of
But his long exposure to the old system had narrowed
his perception of possible improvervents.
-90-
Anne,
and
Rob's
of
staff, had little understanding
cost allocation.
improve the interface with
technical
We
improvements,
complications
the
'IS
of
staff had many ideas aLout how to
The MI.
easier to work with.
the
like
but,
hacd sore good ideas
other new people,
user,
the
also
ideas
had
specific
about
disk
improve
such using a DDMS to
usage, to solve known operational problems.
system
the
making
really
But we
No
didn't know what the processing; steps in CAS were.
one
person had the complete picture.
etc.
still
needed
available
for
handling
The mechanisms of processing data,
to be worked out,
and the options
inputs and outputs needed to be explored.
finalizing
our approach would help speed up the process of
the system design.
On our side, it seemed
that
hoped
We
like
only
the
able
way to give the users an idea of what we might be
to
produce so that they in turn could make reasonably informed
choices between options we might suggest.
By giving them something to work
with,
we
hoped
to
encourage insights and suggestions before the design of the
system became fixed.
to show;
Rather than explain, we would be able
we could test their ideas instead of forcing then
to live with them.
It would also be a vehicle to encourage
user participation in the design.
It seemed otherwise very
difficult to develop much user participation in
requiring as much technical sophistication as
-91-
a
this
project
would.
This would force the issue.
There
7culC be no
them to be frustrated by a
"fixed"
design;
for
chance
design
the
and
it,
would not be fixed until they were satisfied with
there would be no finalized specifications put into writing
Only the broad
until the function was tried in the system.
study description were
to
start.
the
from
required
be
feasibility
oricinal
system requirements contained in the
approach
Besides benefits of their learning, with this
we
would also be in a much better learning position ourselves.
have
We could hope to
an
needs
their
of
understanding
before we comitted ourselves to impractical goals.
accepted the
The users
They
well.
pretty
approach
were uncomfortable with not having any preset guidelines to
but they did like the idea of par-
measure the results by,
ticipating in the evolution of
a
complete
for beginning
without
reasonable, if
not ideal.
Because of
the
many
technical
on
through
as
teing
our
part.
encountered
problems
the
didn't manage to develop a prototype in
hoped.
argument
specifications
But, we were unable to follow
met
truly
expediency
Management accepted the
their needs.
that
system
time
The system itself wasn't too difficult,
had
we
but
there
were many untried interfaces with other systems which
the
necessary to connect together all of
wanted to utilize.
Unfortunately,
we
'ran
pieces
into
we
that
were
we
problems
long
This causei r-any
with every one of those interfaces.
delays.
the many problems,
March.
In January,
getting edgy about
results.
with
Noverber,
Although we had been trying for early
we were unable to deliver anything until
it
SMD and Sue (who by that time
to
began
push
to
had
left
SMD
for
rore
written
The minutes of some
specifications to resolve the problem.
of our meetings were reworked as the basis
plete written documentation.
or
specifications
either
havinT
not
take her new position)
was
became apparant that manager-ent
almost
used
the meetings between the two departments were
up
work
exclusively to discuss problem areas and
February
and
During January
com-
more
for
written
requirements for the system.
By the end of MLarch,
the
users
management
and
detailed
more comfortable with our progress because fairly
if
we
didn't
But we still had
essen-
progranning
team
idea in many cases of how we should approach meeting
needs.
a
now
much better idea of what to hold against us
tially nothing working, and our
had
They
system requirements had been completed.
deliver a working system on time.
were
The programmers were beginning to grow very
fortable with our progress, just as we
were
had
no
those
uncom-
beginning
to
resolve all of the many interface problems so that we could
finally begin to work on the system itself.
idea
In this case, we successfully introduced the
the change.
/devel-
We bypassed most of the forral design
opr-ent recuirerents.
But by failing to deliver, we jeoparWe ended
dized our chance of doing so again in the future.
up reinforcing their preference
instead of the change
Clearly, it
were
we
we had done
introduction.
to
introduce.
more
promising
attempting
job
better
a
its
planning
of
beforehand,
problems
By investigating the
system
official
the
for
would have been seen as a much
technique if
of
(or
schedule
we might have come up with a more reasonable
more reasonable promises).
We would have had trouble here if
tried
we hadn't
to
I don't
change the standard methods of system development.
believe that we could have completed the design work
alone
the
book.
in
six months,
if we
done
had
everything
Aside from the problems inherent
cations before providing any
by
feedback,
we
specifi-
the
fixing
in
simply
didn't
know enough when we started to write them.
The users were hampered by having a lot of
experience
with a poorly designed system, and very little idea of what
might be done to replace it.
both new and non-technical,
being
replaced
replacing it.
or
the
And the APD
being
Our SMD representative,
technical
staff
understanding of the old system.
-94-
the
understand
didn't
issues
was
system
involved
likewise
short
While we understood
in
on
what
features we could offer
the
if
user,
had
we
comoletcd
programming specifications without any experimentation, the
vative system design is
from
suffer
to
likely
inno-
Any
resulting systevm probably wouldn't have worked.
such
an
But
if
initial shortage of specific knowledge and skills.
that
unlikely
we failed to meet our promises, it was very
we would be given a second chance at using this development
technique.
place required
some
understanding
They'had had poor experience with
past, with unfulfilled promises and
the
resistance.
their
of
problems.
unexplained
with
been told repeatedly that they would have to live
from lack of auditability,
ational problems.
poor
relatively
up
standing of data processing, but had picked
underlot
MIS had never been able to solve the problems that CAD.
as
with the system, apparantly using technical "facts"
We
excuse for anything they couldn't program.
in,
offering to do things which seeried
more
we offered
things
couldn't be done.
that
they
had
been
then
difficult
the users than projects others had failed at.
told
saw
an
came
to
In addition,
explicitly
We clearly had to establish our
-95-
of
with.
dealt
had
"facts" from previous MIS people that he
a
--
oper-
to
to difficulty of use,
Rob West had a
had
they
that
The existing system had technical problems
the
in
staff
MIS
first
the
in
user
the
Just selling the project to
believ-
ability.
At the same time,
posed a threat.
CAD,
the original systen,
use the system well.
time we arrived,
our suggestions
for
particularly Rob,
a
new
had helped
develop
and: had worked hard to understand
and
by
the
It
was far from
ideal,
but
they were comfortable with it.
Our proposals to replace the complicated
controlled the CAS processing
mechanism scared the,.
the machine,
system
with
a
tables
greatly
that
sirplified
They were used to doing things
instead of it doing things for them,
really didn't trust us.
and
They "understood" how the
was performing its part now.
they
machine
By simplifying their interac-
tion with the programs, we reduced their perceived
control
over the system, even though their actual control would
increased.
for
But also, we were taking
away
their
job
be
se-
curity, saying in effect, "now anyone can use the system".
None of us at the time, I'm
sure,
explicitly
nized any reasons for the users' reluctance to
old system.
But we were building
performance on other
education.
projects,
on
and
trust
met
Our meetings served to give
their system.
the
good
resistance
with
to
exposure
to
improve
on
We explained many of the technical issues as
best we could in non-technical
value to them,
change
through
them
many of our ideas about what could be done
recog-
as well as to us,
terms,
and
explained
the
of such. things as database
The meetings
systens and modular decsigjn.
accepted.
point
We would,
an example for the
wasn't
being
fully
in
create
try to
Then
question.
we
demonstrate how it micht be handled, and relate
to how it
or
understood
whenever that happened,
issue
recog-
to
get to know each other, so that we were quicker
nize when some
us
helped
also
would
that
back
would work in the old CAS.
We also tried,
as a general practice,
importance of the users to the project.
to emphasize the
write
We couldn't
the system without incorporating their knowledge about
it should work.
how
Likewise, they wouldn't be able to use the
system effectively
without
fully
could do and how it
would do it.
understanding
The
wanted to provide them with a better
idea
was
tool,
so
what
it
that
we
that
they
far
more
might do their job better.
Internal Problems
Our dealings with the
users
were
clearly
successful than our dealings with other groups within
We planned our work with the users pretty well,
taking heed
of the current literature on CIS implementation.
failed to make the logical
extension
about the problems of interaction with
and
worry
other
MIS
Yet the problems of dealing with the user were for
least,
far easier than those of dealing with MIS.
-97-
MIS.
But
as
we
much
groups.
us,
at
Once we had convinced thenl
that a
new
CAS
worthwhile, we had something our users wanted.
to gain from naking the project successful,
would
They
be
stood
obviously
and
understood that, since they requested us to do the project.
CAS itself would be their reward.
needed support fron
Our system, however,
groups as well.
We needed
the
cooperation
MIS
several
of
the
five
systems feeding CAS, and specializecl support
for
each
of
the technical features, like ADABAS, that we
incorporated.
But there was no incentive for the MIS groups
to
aid
us.
The reward system in
us
for
the
MIS
will
completion of our project,
reward
only
regardless of how many people in
other groups contributed to that success.
it is only we who fail.
for them to help.
tial
There was
In fact, in
no
And if
we
inherent
a case like OAE,
fail,
incentive
the
poten-
threat posed by our success provided an incentive
for
them to help us fail.
Jerry, in OAE, whether consciously
or
seems to have found a perfect strategy for
system from encroachnent from other
protecting
systems.
everything look easy while we are in the
schedule
we
By
planning
he encourages us to plan for an easy task.
are under the optimistic
unconsciously,
were
his
making
stages,
Later, when
deluded
we
into
setting up, he delays until we fall behind schedule and our
project fails.
If
that delay is
-98-
not enough to hurt us,
we
fail
may still
lecause
likel7
is
the actual work involved
to turn out to be much more.than we estimated.
We were fortunate that the inforration we cot from OAE
was not required for Phase I of the CAS rewrite, but simply
possible.
When
we'll have a much better idea of
what
something we had hoped to throw in
Phase II
is
planned,
is
involved in
using the rest of
deal
their system. (and of how to
needed everything
we
then,
Jerry).
with
Had
we
successful
for
for,
from
want
we
data
the
looking
went
if
completion of the initial project, we would have been
even
We simply hadn't planned for encountering
this
worse off.
kind of problem.
be
A stock solution to this type of problem would
make formal project
assistance.
of
requests
This gives them. the
help,
and
completion
can
to
incentive
covers us if they don't, since our project
be made contingent on the completion
all
of
needed
for
group
each
others.
the
should
But we would be left to suffer with any delays they
choose to put into their schedule, and we were
speed as well as success.
Rather
this method merely avoids
blame
for
failure.
doesn't solve other problems, such as our
as a threat.
hoping
ensuring
than
to
being
for
success,
It
also
perceived
It would take extremely good project specifi-
cation to nake this solution work if
us to fail, since results of
the
-99-
vants
the other group
request
depend
on
its
interpretation.
going
No reward systen is
cooperate fully if
make
then
they perceive us as a threat.
Despite the fact
that
in
everyone
supposed to be on the same team,
it
MIS
isn't
is
iCeally
more
any
the firm are
than saying all of the d'epartrcnts in
same team.
to
true
the
on
give
The organization and the reward structure
each group its own goals
that
aren't
truly
with
meshed
those of other groups.
A Model for Analvsis
In order to adapt the work
of
and
political
I
risk analysis to the problems of CIS impleirentors,
like to draw an analogy.
Let us assume that
each
national firm.
country.
implement
a
CIS
is
The MIS department will then
In order for our firm to
do
would
depart-
ment will be a country for purposes of the model,
MIS group setting out to
social
and
an
be
business,
have a product that someone in the other country
the
inter-
the
home
it
must
wants
(a
laws
of
system).
While there, they will be forced to obey the
that country.
(Since the staff is not moving
but only working there,
country laws as well.)
they
had
better
permanently,
obey
the
home
To help the firm become successful,
and remain around for a while, they probably don't want
simply obey the laws, but to try to fit
-100-
in to the
to
environ-
ment.
They will want to pay attention to the Lolitical and
social background they find themselves in, rather than just
the financial
and econor-ic.
this foreign firm
to cover a lot of
problem.
an 1IS group gets involved in.
It
interesting points, such as the
(data processing)
of
of
a pretty sound model for much
The MIS citizens
the citizens
mold
to be a model citizen.
I think this is
interaction that
They will perhars try to
all
speak
the
the
extends
language
same
language
although perhaps different dialects, but
the
host
country
(department)
usually
don't. Too often, MIS-ers forget to use the language of the
host country when they are away from MIS-land on
business.
Many of them don't even bother to learn the language of the
country they are working in,
communication problems
languages of all
which
can
cause
(even though once upon
these
countries
sprang
some
a
major
time
from
a
the
cormon
tongue).
For programmers or others in IIIS to deal with
tants, they should know some accounting.
They
don't
have to become accountants, but they must learn the
nology well enough to communicate..
understand questions and be
understood.
And while it
firm knew some data
able
to
make
be
have
termi-
able
their
MIS
is
on them to be
-101-
most
able
often
to
to
replies
would be nice if everyone in
processing,
intruder, so the burden is
They must
accoun-
the
the
explain
the necessary cata irocessing concepts.
The analogy holds for many of the problems that we ran
W'e were so Lusy making sure that we
into in rewritingj CAS.
CAD's
by
accepted
were
and our product
government
and
in
our
course,
was
that we forgot to worry about suppliers
citizenry,
The help we needed at home,
home country.
of
technical assistance with interfaces and feeder systems.
My intent is
not to push the model, but to suggest the
efforts.
to
application
international firms, for
developed
being
validity of borrowing the techniques
CIS
for
development
Recent work in evaluation of investments in other
can't
you
countries has begun to nake people realize that
limit project evaluation to economic factors alone.
It has
incorporate
social
to
been necessary to try to find ways
in Iran have pointed out the
tions like the recent upheaval
need for
a
better
Any U.S.
issues.
of
understanding
investment in
unprofitable
probably in an
Situa-
decisions.
and political factors into investment
those
at
Iran
position
that
non-economic
point
regardless
was
of, the
economic value of the project, because of purely political/
social reasons.
If your company is shut down or taken over
by a foreign government,
doesn't
it
how profitable you expected it to be.
of expropriation corresponds
model
--
a
takeover
of
to
the
-102-
make
much
(Again,
difference
the
decentralization
facility
by
the
threat
in
the
foreign
government.)
Structured Method of Pnalysis
in
Political scientists,
response,
have been trying to
develop structured methods for looking at the non-quantifi-
quantify
been based on attempts to
variables, usually through polling
and
social
opinion.
The
political
expert
provide
more interesting work, however, simply attempts to
the
of
all
about
enough structure to force you to think
important issues.
has
Some of the work
able areas that can irmpact the firm.
list
ideal
The idea is not to find one
of problems to avoid, but to identify how to structure your
An
planning to be sure and consider all the issues.
particular
vidual analyst might then use that to develop a
scheme that works well for
her
or
basis for an initial
will
This
himself.
table
or
probably amount to some kind of checklist
indi-
as
a
analysis.
begin
we
With data processing projects like CAS,
to
see the need more and more for some such political/
social
Although the specific variables they are
ooking
analysis.
scientists
at are different, the approach of the political
is applicable to a similar
tation.
analysis
of
a
implemen-
CIS
As with foreign investment, until recently evalua-
tion of CIS projects has rested almost entirely on economic
cost/ benefit analysis.
I don't hope to offer
-103-
a
complete
solution,
but perhaps making7 use of their work will give us
some insights into CIS implementation problems.
the analysis r-ight be done in
For a CIS,
First, we might list
the actors
each
in
parts.
four
three
of
cateof
the
system are those who receive copies of reports or have
on-
gories:
users,
line access.
suppliers,
as
They might be further classified
(major) users (usually those
important secondary
data),
Users
support groups.
important secondary users.
who
access
control
and
the primary users being CAD,
the
less
or
in
CAS was structured
such
a
this
anal-
secondary
users
way that only two groups would be identified in
ysis --
to
tertiary
and
users,
primary
being all of roughly equal importance to us.
of
We should separately identify suppliers
the system if they are not users.
direct suppliers,
They may be
and indirect suppliers,
to another use which
CAS had known problems in
this
then
The
supplier to CAS was the primary user, but
secondary sources
was
already
known
Validity of CAS results depends on the
mation extracted from other systems.
to
divided
to
basis
this
feeds
area.
to
on the
whether they provide the information only to
or provide it
input
system,
this
only
the
be
quality
of
one.
direct
input
from
inaccurate.
of
infor-
Improving the
inputs
from these suppliers could thus improve CAD's product.
For the third category, we need to list the
-104-
technical
crouns vho will be
feeder system support groups,
(ADjBAS,
CICS,
rAMIS,
CAS
Effected.
etc.),
from.
support
nceded
from technical support groups
data
fror
and
processing
operations to run online.
Is
Next, the system should be evaluated on its own.
it
important to the organization?
machine or people resources?
Is
its
Does
affect
operation
And do other systems
likewise
affect
What kind of information base does it
contain?
Is
other systems?
large or
small
amount
of
Is
data?
information or a closely guarded secret?
Will
it?
that
a
known
widely
it
on
drain
a major
it
system
the
affect the way the data is distributed in the organization?
Will someone have access to more information as a result of
this system?
to run? to maintain?
system
How will the new
use?
to
Is the systen simple or complicated
compare
on
these points with whatever it replaces?
inherent
The answers to these questions establish the
impact of the system.
Since data
is
a
major
source
of
power in an organization, the data content and access to it
hit
to
system
can largely determine the potential of the
of
rela-
tively unimportant information can bring on problems,
put-
political obstacles.
While even a small
ting the complete financial reports of
amount
the
firm
online system with unlimited access is a sure-fire
maker.
into
an
trouble
Thirdly, we neec
to go biac;
of actors,
to the list
Ind
using the previous analysis, decide whether the system will
have a positive impact, a negative
each actor.
or no irmpact
impact,
Some of the results may be mixed.
may require more effort of a
department,
additional
more power in terms of
major issues this way,
plan for them.
spots,
it
By
bring
to
data.
Or
importance
of
access
them
identify
But if we can
the
we are a lot closer to being able to
specific
isolating
potential
trouble
can alert the implerentor before problems arise.
Finally, we should go back and look
the actor affected is,
at
but his boss will
If
actor A is
benefit,
factors acting to balance the added
of
something else for B.
trade
If
of
likely
there
burden
overall, actor A gains and actor B loses,
arrange some kind
powerful
how
and how much he will be affected,
determine the overall issues.
unhappy,
systen
but
access may be restricted to some due to the
the data now being collected.
A
on
to
may
be
on
him.
perhaps
benefits,
with
the actor who gains
be
other
If
we
A
isn't
to
can
doing
in
a
position of power over the loser, the system is probably in
trouble, unless we can find some mechanism
for
equalizing
benefits.
In the case where there are just too many or too
important
losers,
the
system
example, if we had examined
with OAE, we could
have
in
stands
no
advance
recognized
both
chance.
our
the
For
relationship
threat
we
posed,
and the lack of incentive that they had to help
We might then
have
our
modified
approach
to
us.
them
and
improved our chance of successful dealings with them.
I believe that this type
of
analysis
structured
of
political and social issues, done beforehand, could prevent
a lot of CIS implementation failures,
to plan around
problems
before
because it
they
project the best possible chance.
arise,
While
allows you
giving
CAS
was
the
not
a
complete failure, we would have been alerted to many of our
oversights,
and could probably through better planning have
averted many of the delays we experienced.
with little
chance of success due to political
factors within
the
atteripted at all.
this way,
Many
firm
might
be
systems
and
redesigned
social
or
By lowering the failure rate of
we might
further
departments and make still
improve
relations
never
systems
with
user
more gains on future projects.
Conclusions
The computer system designer/
many problems.
They tend to center more
the "people" aspects of the
points out
many
implementor today
of
the
and
system.
Current
potential
stumbling
dealing with users, but gives the analyst
more
little
faces
around
literature
blocks
to
in
help
organize and plan around people problems.
A structured approach to analysis will help to uncover
-107-
pitf-alls in
advance or avoid thei
Published
altogcethcr.
structured design methodologies cover technical design well
and economic valuation to sone
are
issues
for
the
most
extent..
part
But
the
people
I
propose
untouched.
At an
political and social issues into such a methodology.
of
the effects
early phase in the project life cycle,
of
evaluation
integrating a more comprehensive structured
the
proposed system on all of the actors should be evaluated.
What we have to be particularly sensitive to,
with
CAS,
is
problems
at
hone.
problems
than
There is,
however,
a
to
potential for such intra-MIS conflict
like Worldwide's.
has
literature
user
sensitized analysts much more
intra-departmental problems.
The
seen
as
in
an
to
large
organization
So we must be at least as aware of those
problems as we are of the ones with users.
The structure of many data processing
departments
is
amount
of
With
the
increasing
interdependent systems and the
use
of
like that at Worldwide.
tools,
this structure is
advanced
becoming inadequate.
longer able to reward people appropiately
software
It
for
their
because the systems have become more complicated
organization.
adopted.
A matrixed organization should
justified
on
the
same
basis
forces have been set up across departmental
-108-
than
which
the
be
cross
that
lines
no
work
probably
A task force structure for projects
system lines is
is
in
task
the
rest of the organization.
allow a 'eward struc-
This woul
ture that provides incentives for intra-group assistance.
Hopefully by doing upfront political/ social analysis,
we will not only uncover problems, but be able to find ways
to avoid
them.
Whenever
should
changes
possible,
be
planned, along with any incentives needed to elicit cooperation.
such
For some problers, simple steps
tation may be sufficient to greatly improve
with OAE, interviews may become much more
is
you let it be known that the discussion
results.
As
satisfactory
if
into
going
a
The memo provides
memo-, and will be distributed to others.
exposure --
documen-
as
the
good or bad depending on their response at
time.
The critical point is
able
to be
expect problems, in order to be
able
Being caught off guard, especially
to
to
during
know
plan
early
where
for
them.
developThe
ment, is apt to cause a major setback to your system.
way systems in general are headed, it will increasingly
political and social issues that pose the biggest
A structured analysis can prepare
database group that isn't
you
for
an
willing to cooperate.
be
threats.
OAE
or
a
Hopefully,
forewarned you will be able to develop a solution in
time.
And more successful impleentations will be the result.
-109-
to
APPENDIXN
SOME NOTES ON. THE ENVIRONENT
At this point, I would like to try to
While
points that seem to require further explanation.
I
very
is
Worldwide
at
don't believe that the environment
some
up
clear
unusual, it perhaps fits less well into a description of
a
than
"typical" data processing shop in some specific areas
in others.
Production Systems
"production"
The process of putting a system into
probably not too clear.
involves
It
a
that would probably best be done by a
instead of a programmer.
oriented job, that must
It is a
be
time
completed
number
well
for
things
clerk
trained
consuming,
(In
clerk to be able to run the system.
of
an
some
is
detail
operations
organiza-
tions, in fact, the position of data processing "librarian"
has recently begun to appear as a
opment teams.)
part
of
system
devel-
There are a large number of standard
forms
to
allow
that have to be filled out.
These are
intended
the clerks in DPD to correctly plan and schedule jobs,
set
They
are
up inputs, verify and distribute reports, etc.
necessarily quite detailed, including
-110-
file
names,
report
names, and names and locations of support
programmers
and
those who are to receive reports.
Completing the forms and getting them typed
least).
(mentally,
challenging
is time consuming and not very
Because MIS typists turn
over
long
programmers, they seldom remain
correctly
even
enough
faster
than
to
very
be
familiar with data processing terminology and formats.
instance,
IBM Job Control Lancuage instructions
spaces in the right places.
unaware that there is a
Most typists at
difference,
treat
(JCL)
The result is
JCL
that
forms
are
test runs.
often
times
along
After 3-6 weeks,
accept the programs to
being
like
and
all
better
be
with
DPD
run
copies
is
in
corrected
before
After a complete set of forms is
sent to DPD for review,
and
look more like English.
brought back for retyping several
acceptable.
must
first,
programmer scribbles, using lowercase, spaces,
punctuation to make it
For
letters
be typed exactly as written, with all capital
at
they
ready,
of
it
and
are
is
successful
required
to
production
either
tests,
return the documentation with an explanation of why it
rejected (this is usually missing forms, mistyped
JCL,
or
was
or
instructions that are difficult to follow).
CAS is even more of a problem than most systems.
several incoming files from
other
tation must be updated every time
systems,
the
other
its
With
documen-
systems
are
Program chng:-ices and cianing TIS
changed.
regularly
caused
times,
temporary files, program run
of
sizes
descriptions,
in
changes
have-
standards
etc.
titles,
report
until
Even such things as program names changed frequently
we set up standards for the system.
Contract Procgrar:ers
The use of
cited as a problem area,
at
prograrmers
contract
Worldwide
pro-
Contract
perhaps unfairly.
is
grammers are widely used in the industry to meet short term
staffing needs.
Programmers are often a scarce resource in
firms, so a project will be forced to go to outside help to
This is
meet temporary staffing needs.
confined
normally
to the actual programming phase of a project,
there
where
is a large volume of work and little knowledge of the
pany is
com-
required.
Contract programmers in
general
expected
aren't
have the dedication to the firm or to the specific
that a company staff member might have.
short time they are assigned to
a
firm,
don't care about the long term viability of
solutions.
Because
they
to
project
of
the
supposedly
their
problem
The fact that they will probably be assigned to
a project only during the development stage is
said to make
them indifferent as to how well the completed program
will
short
term
run.
Their concern is thought to be only with
-112-
specific currnt goals.
successes and moeetinc
or
to
write
to
neccssary
skills
learn the non-data processing
of
know
to
expected
their short stays, they are not
Because
finan-
good systems (such as good accounting practices for
cial systems).
But,
IS staff.)
remain longer on average than
better paid by their own
Therefore,
better
are
people
quality
definitely do care about their projects.
in
the
sell
that
firms
staff.
MIS
than
there
about
care
They
them
of
some
reputation
their
to
important
Call-backs and referrals are
follow them later.
standing
generally
are
found
and want a good
producing good work anyway,
their
and
And
through the personnel hiring process.
to
available,
regular
than
firms
can
They
skills.
chosen for specific needed
Contractors
are
that
several
often be selected from
actually
Worldwide
at
(Contract programm-ers
don't hold.
arguments
these
of
most
permanent than the contractors,
more
no
at a firm where the "permanent" staff is
services.
Computer Service Chargeback
others
and
As Gibson and Nolan (1974)
have
pointed
begin
out, chargeback of computer services should normally
very early in the life of
Cost information is
a
critical
data
to
processing
the
department.
evaluation
of
the
relative merits of any two proposed solutions to a problem.
-113-
Without cost chargeback,
the oranization has no
to decide which projects should be
undertaken,
computer systems will tend to ref lect the
rather
of various user departments,
fair
:ay
and
their
political
power
reflecting
than
the
value of particular applications to the firm.
Worldwide has
managed
somehow
charging for computer services.
to
They
continue
have
without
developed
a
respectable set of computer applications, but no one really
knows what the cost has been.
SMD attempts to judge systems by their dollar value to
the company,
cost
data
services.
but their analysis suffers from a lack of real
because
there
is
no
for
computer
Where other firms record usage of the
computer,
storage space (tapes and disks),
etc.,
billing
and
supplies
of
paper,
and bill the users in detail, Worldwide has chosen to
ignore this possibility.
Instead, they rely on the limited
ability of CAS to allocate the cost of data
departments.
MIS expenses show up
item (some $15 million a year),
as
a
processing
single
allocated on the
to
overhead
basis
of
some departmental workload measures in user areas.
Instead of bills,
the
user
managers
summarized reports of computer usage,
CPU usage in hours and total number of
quarter.
No attempt
is
made
figures or to charge for usage.
-114-
to
highly
giving only the total
jobs
relate
In
receive
fact,
run
cost
our
for
to
user
the
those
Rob
West,
complains when his "share"
used.
totals
compares his rerort w;ith other managers'
of
resources
available
When we were trying to convince him
and
isn't
rewriting
that
CAS was necessary, one of his arguments was that we weren't
using nearly as much CPU time as CAE was (at that point his
figures already showed us with several
for CAS for the year --
bill
other places).
CPU
hundred
which would amount
to
a
Ie didn't feel that OAE was
important to Worldwide than CAS.
Therefore,
sizeable
any
more
in his opinion
we deserved to use at least as much resources as they
It
hours
did.
didn't matter to him that they ran 30 times as many jobs
as we did and produced many more reports.
of
computer
ser-
vices makes SMD's cost-benefit difficult to
estimate
(due
This lack of awareness of the cost
to lack of figures) and still more difficult to demonstrate
(what is
the cost to a manager of a free service?).
Sum-any
T4
Armin
cronyms
and Terminology
ADABAS
DDMS used for our large data files.
APD
CIS
Applications Programmring Department, a part
of MIS.
Cost Analysis Departatent, p rimary users of
the Cost Analysis System.
Cost Analysis System, used to refer to the
systen ancd the rocramrrinc group assicgned
to it.
On-line terrinal monitor program available on
user terminals.
Computerized Information System.
COBOL
Computer programming language used in extracts.
DBMS
Data Base Management System.
DPD
Data Processing Department, a part of MIS.
E-R
Entity-Relationship method for database design.
Extracts
RAMIS
The front end of CAS, programs collect and
convert feeder system data for allocation.
Manufacturer of most computer hardware at
Worldwide, also provides much of the systems
software.
Management Information Systems, a division
of Worldwide Insurance.
A CIS and the support group for it in APD.
Provides three input files to CAS.
DBMS used by CAS for smaller data files.
PL/1
Programming language used in the new CAS.
SMD
Systems Management Department,
a part of
Technology
Systems Technology Department,
a part of MIS.
TSO
On-line terminal monitor program available on
programmer terminals.
CAD
CAS
CICS
IBM
MIS
OAE
Ficure 8
-116-
IS.
BIBLIOGRAPHY
"The Entity-Relationship A proach to Logical
Chen, Peter;
Q.E.D. InformaData Base Design"; Wellesley, Mass.:
tion Sciences, Inc., 1977.
Gane, Chris, and Trish Sarson; Structured Systers Analysis:
Improved- Syste-s TchYork:
0'1e
Tools and Techniaus;
nolog-ies,
Inlc.,
19L7
Gibson, C. F., and R. L. .olan; "Managing the Four Stages
Jan-Feb. 1974.
of EDP Growth", Harvard Business Review;
Jackson, Michael;
Press, 1975.
Principles of Program Desian;
Academic
Does Political Instability ReKobrin, Stephen J.; "When
Columbia Journal
sult in Increased Investment Risk? ",
of World Business; Fall 1978.
"The Multinationals Got Smarter About PoliKraar, Louis;
March 24, 1980.
tical Risks", Fortune;
Markus, M. Lynne; "Understanding Information System Use in
Unpublished
A Theoretical Explanation";
Organizations:
doctoral dissertation, Case Western Reserve University,
1979.
The Politics
Pettigrew, Andrew M.;
Tavistock,
London:
Decision-Making;
L.
Yourdon, E., and L.
Yourdon, Inc., 1978.
Constantine;
-117-
of Organizational
1973.
Structured Design;
POS T SCRIPT
On May 5, a memo was released in MIS announcing a reorg-anSMD was
Among other changes,
ization of the departLIent.
officially resurrected, but as a part of each APD group.
Existing SMD representatives were reassiened to the APD
managers that they supported (an effective downgrading of
they were considered cquals of
their status, since forrrly
At the same time, we heard unoffithose group managers.)
cially of a "minor" corporate restructuring to take place
soon.
On May 6, Rob West announced that he was leaving for a new
job. By the time he had left, Anne had given notice as
well. Dick, also on Rlob's staff, transferred to the Financial Systems group that Sue was in. He was to support some
But, after less than a week,
of Rob's work from that area.
he too announced his intention to leave Worldwide.
in the RAIS and
At the end of May, our original contacts
ADABAS support groups left within a week of each other.
Brad and
week in June.
Andy moved to Technology the first
I left Worldwide
Stan are both actively looking for jobs.
at the end of June, on the planned completion date of CAS.
numThe project to develop a consistent corporate-wide
Chances are
"being considered".
bering scheme is still
that it will be "considered" until everyone involved has
left
or lost interest.
I expect that CAS will be completed by the end of the
but CAS has
schedule,
sunmer.
This is somewhat behind
already proven enough to the firm that they will procecd
with our long range plans.
An advanced development group,
headed by Lou, has been set up to begin plans for conversion of the next financial systcm. CAS has essentially no
users at this point,
but it
has the support of top
management, so it is expected that Rob's replacement will
have an interest in making use of the system.
-118-