AD-A242 594

RL-TR-91-305
In-House Report
September 1991

TOTAL QUALITY
MANAGEMENT (TQM), AN
OVERVIEW
DTIC
NOV 1 8 10,91

Anthony Coppola D

APPROVED FOR PUBLIC RELEASE, DISTRIBUTION UNLIMITED

91-15748

Rome Laboratory
Air Force Systems Command
Griffiss Air Force Base, NY 13441-5700
 This report has been reviewed by the Rome Laboratory Public Affairs Office
(PA) and is releasable to the National Technical Information Service (NTIS). At NTIS
it will be releasable to the general public, including foreign nations.

RL-TR-91-305 has been reviewed and is approved for publication.

APPROVED:

ANTHONY J. FEDUCCIA, Chief

Systems Reliability Division

FOR THE COMMANDER:

JOHN J. BART
Technical Director
Electromagnetics & Reliability Directorate

Ifyour address has changed or if you wish to be removed from the Rome Laboratory
mailing list, or if the addressee is no longer employed by your organization, please
notify RL(ERSS ) Griffiss AFB NY 13441-5700. This will assist us in maintaining a
current mailing list.

Do not return copies of this report unless contractual obligations or notices on a

specific document require that it be returned.
 REPORT DOCUMENTATION PAGE Form Approved
0MB No. 0704-01 88
Pbcrepcxtrg b.,dei far &IS ohb d ri-OW~r is eST18ted toww N&" i per repre, rxc*d~-e t rfor eieri g rwsri.bpns sendrvVisrg ata sa
gatwng widfrmwwig Un cd.nreesc m-doopst ac~r d rwwrrg tin c9ad ricxrrto So-rdoofrwts regwdrlg is busdmies5toa or "v Ctr aspea off-
coc~m cIIcr
dH ,nn a CW~~dz1suggestion fcx red.ig i-s tbzder- to Was*,wcj, Hedptwes Savxas, Diemct. fo Hx 1crwn Operftn "iReprts. 12 5 .effwsor,
Davis Hq,". St-g 1204,firkigboq VA MMt24= and to tin Offloeof Mamngarwt vW1Budgn Pepwworc Redi boProjec (07U-01 88). Waet~tor DC 20503
1. AGENCY USE ONLY (Leave Blank) 2. REPORT DATE 3. REPORT TYPE AND DATES COVERED
ISeptember 1991 In-Hfouse
4. TITLE AND SUBTITLE 5. FUNDING NUMBERS
TOTAL QUALITY MANAGEMENT (TQM), AN OVERVIEW PE - 62702F
PR - 2338
Top-o0a
6. AUTHOR($)
Anthoy CopolaWU - TK

7. PERFORMING ORGANIZA-TION NAME(S) AND ADDRESS(ES) 8. PERFORMING ORGANIZATION

Rome Laboratory (ERSS) REPORT NUMBER
Griffiss AFB NY 13441-5700 RL-TR-91-305

9. SPONSORING/MONITORING AGENCY NAME(S) AND ADDRESS(ES) 10. SPONSORINGMONITORING

Rome Laboratory (ERSS) AGENCY REPORT NUMBER
Griffiss AFB NY 13441-5700

11. SUPPLEMENTARY NOTES Rome Laboratory Project Engineer: Anthony Coppola/ERSS/(315) 330-
4758. Report will also be published in Tutqrial Notes of 1992 Annual Reliability and
Maintainability Symposium.
12a-DISTRIBUTION/AVAiLABIU1TY STATEMENT 12. DISTRIBUTION CODE
Approved for public release; distribution unlimited.

13. ABSTRACT(maxk- 2w -or)

This report is essentially a slight modification of a tutorial paper prepared by the

author for the 1992 Annual Reliability and Maintainability Symposium, providing a
comprehensive overview of Total Quality Management (TQM). It discusses the reasons
TQM is a current growth industry, what it is, and how one implements it. it describes
the basic analytical tools, statistical process control, some advanced analytical tools
tools used by process improvement teams to enhanice their own operations, and action
plans for making improvements. The final sections discuss assessing quality effort-,
and measuring the quality of knowledge work.

14. SUBJECT TERMS 15 NUMBER CF Z'A-ES

Iaty, hdoil oIi ty M[og~tiQM, Tutor iol , Overvijew 1

6 PRICECODE

17, SECURITY CLASSIFICATION 18. SECURITY CLASSIFICATION 19. SECURITY CLASSIFICATION 120. LIMITATION OF AIS I IJAC I
OF
REPOR
CAS OF THI PA OF 111)
OF*'*
REPORT I- OF INI.AS i
OF13 CT 1
N'SN 7540-012M~5 Wma~
P'escdtedo A%'., S.':-8
29W-
102
 TABLE OF CONTENTS

INTRODUCTION ............................................................... i

WHY TQM? .............................................................. I

WHAT IS TQM? .......................................................... I

IMPLEMENTING TQM ......................................................

TQM TOOLS ................................................................. I0

BASIC ANALYTICAL TOOLS ............................................... Id

STATISTICAL PROCESS CONTROL .......................................... 16

ADVANCED ANALYTICAL TOOLS ............................................ 2U

TEAM TOOLS ................................................................ Z/

TEAM DYNAMICS ..................................................... 28

TEAM PROCESS EVALUATION TOOLS ........................................ 1U

TEAM PROCESS TOOLS ................................................... j1

GETTING STARTED ......................................................

ACTION PLANS ......................................................... 34

ASSESSING QUALITY EFFORTS ................................................. 16

MEASURING THE QUALITY OF KNOWLEDGE WORK ....................................i

CONCLUSION ................................................................ 43

-- - - ---. 4
 INTRODUCTION

This report is essentially a slight modification of a tutorial paper

prepared by the author for the the 1992 Annual Reliability and
Maintainability Symposium. Since it provides a comprehensive overview of
Total Quality Management (TQM), it was decided to publish it as a technical
report so that it would be available to all interested parties, rather than
just tne symposium attendees. It will discuss the reasons TQM is a current
growth industry, what it is, and now one implements it. It will then
describe the basic analytical tools, statistical process control, some
advanced analytical tools, tools used by process improvement teams to enhance
tneir own operations, and action plans for making improvements. The final
sections will discuss assessing quality efforts and measuring the quality of
knowledge work.

WHY TQM?

"If we don't change directions soon, we are doomed to end up where we

are headed," states an ancient Chinese adage. In 197U, 17 U.S. firms
produced televisions; today there is only one. Most sets are imported
because of higher quality and lower cost. In 1975, five of the six largest
semiconductor manufacturers were U.S. companies; today six of the largest
seven are Japanese. Are we heading where we want to go?

"If you always do what you always did, you will always get what you
always got." Which is not good enough, as American auto maKers found out
when they lost market share to imports, again because of quality and cost.

"There ain't no more money," says George Butts, formerly of Chrysler

Corp. He hastens to add that there is plenty of money around; there just is
not any new source. So new profits must come from the same sources of income
as present profits. However, since about 25% of manufacturing costs are
absorbed by scrap, rework and waste, there is plenty of opportunity there.

Total Quality Management promises to improve quality and lower costs.

It is therefore a means for survival, a way of increasing profits, and an
insurer of jobs. It is also a way of enhancing job satisfaction by
increasing a worker's pride in his product, and has an appeal to morality
because through quality, the customer will be getting good value, the
manufacturer enjoys a fair profit, and the worker will have a secure and
satisfying job. Everyone wins.

WHAT IS TQM?

To clarify the concept of Total Quality Management, we will discuss the

work of the best known quality "gurus," and examine some definitions used oy
various agencies. From these we will extract some common principles and some
points of disagreement.

The most famous names in TQM are: Deming, Juran, Crosby, Fiegenbaum,
Ishikawa, and Taguchi. In the author's opinion, however, TQM practitioners
should also be acquainted with some works of Townsend, Augustine, and
Drucker.
 W. Edwards Deming played a key role in spreading the use of statistical
quality control in the United States during World War II. In tne 195U's,
American industry put tneir emphasis on production, forgetting much of What
Deming taught. Japan, however, was rebuilding their industry based on tne
Deming philosophy. In 1951, the Japanese established tne Deming prize,
awarded every year for accomplishments in statistical application. It is
still one of their most prestigious awards. The Deming philosophy is
summarized in his 14 points:

1. Create constancy of purpose (for improvement)

2. Adopt the new philosophy (quality first)

3. Cease dependence on mass inspection (instead, prevent defects)

4. End awards on price alone

5. Improve the system constantly and forever

6. Institute training (of job skills)

7. Institute leadership

8. Drive out fear

9. Break down barriers Detween staff

10. Eliminate slogans, exhortations, and targets

11. Eliminate numerical quotas

12. Remove barriers to pride of workmanship

13. Institute a vigorous program of education and retraining

14. Take action to change

Dr. Deming also lists seven "deadly diseases" of American Management:

1. Lack of constancy of purpose

2. Emphasis on short-term profits

3. Performance reviews (which destroy teamwork and ouild fear)

4. Mobility of management (works against understanding and long-term efforts)

5. Running a company on visible figures alone (you can't measure the effects
of a dissatisfied customer)

6. Excessive medical costs (GM's highest paid supplier is Blue Cross)

7. Excessive liability costs (America is the world leader in law suits)

2
 Dr. Deming's advice is now in great demand in the United States. He
conducts four day seminars in quality management from which two exercises,
the red bead exercise and the funnel experiment, have become classic
illustrations in quality training.

The red bead exercise, briefly, is a simulation of a factory. Willing

workers are taKen from the audience and directed to make white beads. Their
process is to dip a paddle into a mixture of white and red Oeads. The paddle
has 50 depressions and extracts that many Deads from tne mixture. No matter
how nard the workers try, they never succeed in producing wnite oeads without
red ones mixed in. In the course of the exercise (which is far more
interesting than this summary indicates) the seminar attendees learn several
lessons including:

- Willing workers are doing the best they can. Exhortations and
threats cannot improve quality.

- Improvements will come only by changing the process. This is

management's job.

- Variation is a part of every process. It must be understood to De

controlled.

In the funnel experiment, a marble is dropped through a funnel over a

target. If it comes to rest away from the target, the location of the funnel
is changed according to a set of rules, and another marble dropped.

One set of rules moves the funnel uoay from the target the same
distance as the marble, out in the opposite direction. This illustrates the
attempt to overcome variation by adjusting a process against the direction of
error. For example, if a machine produces a rod longer than target, it would
be adjusted to make shorter rods. The result of this tinkering is shown by
the funnel experiment to double the variation in the product from that of a
process left alone. The lesson is again to understand variation and reduce
it by process changes rather than increase it by tinkering.

Another set of rules moves the funnel over the location of the marble
after each trial. This compoun6i, the errors and ultimately drives the
variance to infinity. The lesson illustrated is Deming's contention that as
worker trains worker more and more errors are introduced into the process.
It is therefore management's responsibility to provide training and
retraining in the proper methods of doing the job.

Deming also claims that quality benefits the worker as shown in the
Deming Chain reaction:

IMPROVE QUALITY - COSTS DECREASE - PRODUCTIVITY IMPROVES - BETTER QUALITY AND

LOWER PRICE CAPTURES THE MARKET - BUSINESS SURVIVES AND GROWS - MORE JOBS
CREATED.

J. M. Juran was also an advisor to Japan, and is the author of many

practical nandbooks on managing quality. His philosophy is summarized in the
"Juran Trilogy": quality planning, quality control, and quality improvement.

3
 Quality planning provides the emphasis and resources to meet the
customer's needs.

Quality control continuously evaluates the product and acts to prevent

any degradation.

Quality improvement includes creation of an infrastructure conducive to

quality improvement, chartering of project teams for specific opportunities,
and supply of resources, training, and motivation.

Philip B. Crosoy coined the phrase "Quality is Free" in his book of the
same title. He defines quality as meeting specifications, and defines cost
of quality as the expense of nonconformance including prevention, appraisal,
and failure. Since tne cost of failure is much higher than the cost of
prevention, building in quality is less costly than not. Hence, quality is
free, tnough not a gift.
Crosby invented the phrase "Zero defects," and proposed a 14 step
approach to quality:

1. Management commitment

2. Quality improvement team

3. Quality measurement (defect rates)

4. Cost of quality evaluation

5. Quality awareness

6. Corrective action

7. Ad hoc committee for zero defects program

8. Supervisor training

9. Zero defects day

10. Goal setting

11. Error cause removal

12. Recognition

13. Quality councils

14. Do it over again

A. V. Feigenbaum coined the phrase "Total Quality Control" defined in

his 1961 book "Total quality Control" as: "An effective system for
integrating the quality-development, quality-maintenance, and quality
improvement efforts of the various groups in an organization so as to enable
marketing, engineering, production and service at the most economical levels
which allow for full customer satisfaction."

4
 Feigenbaum defined quality costs as the sum of prevention costs,
appraisal costs, internal failure costs and external failure costs.

Kaoru Isnikawa in "What is Total Quality Control? Tne Japanese Way"

emphasized:

- Leadership oy top management

- Education from top to bottom
- Action Dased on Knowledge and data
- Teamwork, elimination of sectionalism
- Customer focus
- Prevention of defects by eliminating root causes
- Elimination of inspection
- Use of statistical methods
- Long term commitment

In his book he states: "TQC is a thought revolution for management."

Also, "QC brings out the best in everyone" and "When QC is implemented,
falsehood disappears from tne company."

Genichi Taguchi is noted for his emphasis on the reduction of variation

and the creation of robust designs (i.e. designs which continue to perform
well as the use environment varies). His contributions include improved
methods for statistical design of experiments to determine causes of
variation (though there is some controversy about these). He formulated
"loss functions" to quantify the adverse economic effects of variation.

Taguchi's contributions are often explained by considering a design

hierarchy: system design, parameter design, and tolerance design. System
design creates the means to accomplish some mission, and American designers
are strong in this area. Parameter design is concerned with the
specification of the system components. This is a Japanese strength, and a
Taguchi specialty. Tolerance design, the setting of limits on specified
values, is done equally well by both countries.

Though not usually listed among the TQM gurus, Robert Townsend
published a book, "Up the Organization," in 1970 which recognized many of the
Points made by the usual TQM referents. He preached rebellion against
mindless rules which accumulate in all organizations. He suggested managers
call their own offices to see what impressions a customer gets when he
calls. He noted the importance of leadership, the need for a manager to be a
coach, and the general under-utilization of people in an organization. My
favorite quote:

"If you can't do it excellently, don't do it at all. Because if its

not excellent it won't be profitable or fun, and if you're not in business
for fun or profit, what the hell are you doing here?"

Norman Augustine wrote a book, "Augustine's Laws," describing the

American aerospace industry in a way which is amusing to those who do not
realize that he is not exaggerating. Some insights of interest to the
student of TQM: "It costs a lot to build bad oroducts" (cost of quality),
"most of our problems are self-imposed," and "rules are no substitute for
sound judgement."

5
 Finally, Peter Drucker, the noted author of management books, ("The
Effective Executive" and others) states the principle that management's job
is to make a customer. He also repeatedly emphasizes that "doing things
right" occupies too much management attention which should oe devoted to
"doing the right things." Drucker advises managers to "pick the future over
the past," an excellent tenet for a TQM initiative.

Definitions of TQM.

In a draft of DoD 5000.51-G, "Total Quality Management, A Guide for

Implementation" The Department of Defense states:

"Total Quality Management (TQM) is both a philosophy and a set of guiding

principles that represent the foundation of a continuously improving
organization. TQM is the application of quantitative methods and numan
resources to improve the material and services supplied to an organization,
all tne processes within an organization, and the degree to which the needs
of the customer are met, now and in the future."

The Air Force Systems Command put out a TQM pamphlet in 1990 which
defined it as:

"A leadership philosophy, organizational structure, and working

environment that fosters and nourishes a personal accountability and
responsibility for quality and a quest for continuous improvement in
products, services, and processes.

The Air Force Electronic Systems Division's pamphlet defined TQM as:

"....The adoption of a customer-oriented operating philosophy committed

to excellence in our products, services, and relationships through the total
participation of all our employees in the constant improvement of all
processes."

The Army Material Command uses this concise definition:

"A philosophy of pursuing continuous improvement in every process through tne

integrated efforts of all members of the organization"

A Navy TQM seminar offered:

"Customer-oriented, quality focused management pnilosopny for

providing leadership, training, and motivation to continuously improve an
organization's processes using modern process control techniques"

The Federal Quality Institute's definition:

"TQM is a strategic, integrated management system for achieving customer

satisfaction which involves all managers and employees and uses quantitative
methods to continuously improve an organization's processes."

Finally, a NASA contact provided this, by Albert M. Koller:

6
 "TQM is an approach to managing work based upon (1) the analytical
evaluation of work processes; (2) the development of a "quality" culture; and
(3) the "empowerment" of employees -- all for the purpose of continuous
improvement of your product or service."

Principles and Issues

From the above, we can summarize agreement on these PRINCIPLES OF TQM:

CUSTOMER SATISFACTION

MANAGEMENT LEADERSHIP CREATING A QUALITY CULTURE

IMPROVEMENT OF PROCESSES, NOT "MOTIVATION" OF PEOPLE

EDUCATION AND TRAINING (JOB SKILLS AND TQM TOOLS, AT LEAST)

DEFECT PREVENTION IN LIEU OF INSPECTION

USE OF DATA AND STATISTICAL TOOLS

TEAM APPROACH
- HORIZONTAL (BETWEEN DEPARTMENTS)
- VERTICAL (CEO TO LOWEST PAID EMPLOYEE)

CONTINUOUS IMPROVEMENT

We can also note that TQM is not: new, a program (as opposed to a
process or philosophy), a quick fix or magic solution, spiritual guidance, a
slogan campaign, a Japanese invention, a suggestion program, or a substitute
for discipline and dedicated effort. It also is definitely not easy.

The TQM gurus are not monolithic in their opinions. There are many issues
which divide them. These include:

FORMS OF RECOGNITION: Should workers who serve on improvement teams be

rewarded with money? Is it better to acknowledge their contributions by a
public thanks or certificate? Or should the satisfaction of making an
improvement be enough'

USE OF HOOPLA: Some TQM initiatives have been kicked off with parades.
Banners and coffee mugs with TQM slogans abound. Crosby would recommend
these; Deming might consider them empty exhortations. Who is right, or does
it depend on the organization?

SETTING GOALS: Is this an essential activity or a counterproductive

exercise? Motorola's so far successful "six-sigma" program has a yearly
target for defect reduction. Deming might point out that an easy approach to
victory would be to change the definition of defect (something not unknown in
my bureaucracy). Does it take a target to sustain energy, or is it an
invitation to play numbers games?

7
 "ZERO DEFECTS": Where ZD nas not been successful, it is called an empty
slogan. As an exhortation for willing workers to do better, it would be just
that. However, it is a concise summary of the goal of all TQM effort. Since
it represents an unattainable perfection, it implies the need for constant
improvement. The difference in these two viewpoints may De merely a
reflection of the understanding with whicn a ZD program is applied. (Note
that CrosDy advocates ZD as a program rather than a pnilosopny; ne acnieves
constancy of purpose by repeating the program yearly.)

APPRAISALS: Despite Deming's condemnation, performance appraisals are

likely to be with us a while. Can we live with them? Can we make them
useful (by appraising teamwork, for example)?

TEAMS: Should participation on improvement teams be voluntary or

mandatory? Who charters a team? At the Harris Corporation in Melbourne, FL,
each activity in the plant is designated a "work cell" and its employees
assigned to a quality improvement team with a goal formulated by management.
Tnis is in addition to cross-functional teams created at higher levels for ad
hoc missions. In some agencies, teams are charted by a TQM Council, and
membership assigned to agencies, if not individuals. At the Rome Laboratory,
any level of management can charter a team, and membership is strictly
voluntary. What's best for your activity?

SUGGESTION BOXES: Often established to get the voice of the employee.

Can deteriorate into a means for venting frustration (pernaps a useful
function). However, usually a one-way street; the employee does not get
feedback on what was done with his input. Unless visible action is taken in
a timely manner, the worker may assume that no one is paying attention.
Since not all suggestions can be implemented, the lack of feedback may give
an impression of a lack of management support, even when it exists.

IMPLEMENTING TQM

The following is a summary of actions recommended to maKe TQM happen:

1. CREATE VISION (What do you want to be? Consider both goals and values.)

2. PLAN ACTION (How do you get from today to the vision?)

3. CREATE STRUCTURE

- INSTITUTE TRAINING
- ELIMINATE HASSLES
- INVOLVE EMPLOYEES
- TRUST AND EMPOWER THEM
- PERMIT RISK TAKING
- "WALK THE TALK"
- ALWAYS PUT QUALITY FIRST
- MAKE APPRAISAL/REWARD SYSTEMS SUPPORTIVE
- CREATE CROSS-FUNCTIONAL TEAMS

4. MEASURE PROGRESS

5. UPDATE PLANS AND VISION AS REQUIRED.

8
 In implementing TQM, tne chief executive officer must lead tne qay.
He cannot delegate leadership, nor can he ever put quality on tne Dack
Ourner, o,, TQM will never mature at nis agency.

If a union represents the employees, they must Decome partners in TQM

implementation. They snould De involved at tne start and all TQM information
must oe shared with them. The union should define its own role in TQM, out
it should be separate from established grievance and bargaining proceaures.

An aggressive training program is necessary to deploy the vision and

create a quality culture. job skills and oasic quality concepts must be
provided to all employees. Those involved in improvement teams should have
available to them training in group dynamics and in TQM analysis tools of
interest.

Improved quality must never oe seen as a threat. When an improved

process needs less man-hours to perform, the people released snould De
considered an asset available for other uses and guaranteed continued
employment.
TQM has been described as a cultural change. The following is an Air
Force chart illustrating this idea.

CULTURAL CHANGES NEEDED

FROM TO

BOTTOM LINE EMPHASIS QUALITY FIRST

MEET SPECIFICATION CONTINUOUS IMPROVEMENT

GET PRODUCT OUT SATISFY CUSTOMER

FOCUS ON PRODUCT FOCUS ON PROCESS

SHORT TERM OBJECTIVE, LONG TERM VIEW

DELEGATED QUALITY RESPONSIBILITY MANAGEMENT-LED IMPROVEMENT

INSPECTION ORIENTATION PREVENTION ORIENTATION

PEOPLE ARE COST BURDENS PEOPLE ARE ASSETS

SEQUENTIAL ENGINEERING TEAMWORK

MINIMUM COST SUPPLIERS QUALITY PARTNER SUPPLIERS

COMPARTMENTALIZED ACTIVITIES COOPERATIVE TEAM EFFORTS

MANAGEMENT BY EDICT EMPLOYEE PARTICIPATION

This chart can be used as a statement of goals and as a check list

to gauge progress.
9
 TQM TOOLS

Tnis section will discuss TQM tools including the basic analytical
tools, statistical process control, advanced analytical tools, special tools
used oy teams, and action plans for improving processes.

BASIC ANALYTICAL TOOLS

First, let's look at the most Dasic tool of all: tne question "wny?"
Many TQM references list a technique called "The five whys." Actually, five
is used just for convenience; the method is simply to get to the root cause
of a problem Dy asking "why" as often as necessary. For example:

Why is our mail from our customer in Osnkosn so slow?

It's held up in the mailroom.
Why?

The mail people have to look up our mail code.

Why?
Our customer doesn't put it in our address.
Why?

He doesn't know it.

Why?
We never told him. (root cause - corrective action can be taken)

Ishikawa contends that 95% of a company's problems can be solved with

the use of seven basic tools. Only one of these, control charts, is of any
great complexity. The author believes any engineer can look at a sample of
the other six and immediately understand how to use them. The seven basic
TQM tools are also popular in the literature. However, every citation seems
to differ from the others by one tool. Here's mine:
THE SEVEN BASIC TOOLS:
- flow charts
- Ishikawa diagrams
- checklists
- histograms
- Pareto charts
- scattergrams
- control charts

This list differs from Isrikawa's in that I substituted flow charts for
stratification (which I will include in my discussion of Pareto charts). A
flow chart is simply a diagram showing the inputs and outputs of all
operations in a process. (see figure 1.) Strictly speaking, it should also
10
show feedback, which is a part of every process. It is a fundamental tool in
that it provides understanding of the process under study. It is important
that the actual process be captured, not the manager's uninformed opinion of
what is happening. Very often, the actual process is not what the manager
thought it was and an accurate flow chart makes many corrective actions
oDvious.

( IREW RITE ,

RSARH kYESYE

SWRITE~EPO
REPORTa~.~PROBEMAANAGMENTENT
EI -- AGM INS

Figure 1

An Ishikawa diagram is also called a cause and effect chart or, from
its form, a fishbone chart. (see figure 2.) It simply displays the factors
which cause an effect such as a problem under study or a goal to be worked
for. The "bones" of the chart can be any set of factors considered important
as causes. Often it is helpful to start with the "4 M's" : METHOD, MANPOWER,
MATERIAL, MACHINERY; or the "4 P's": POLICIES, PROCEDURES, PEOPLE, PLANT.
Each "bone" can have any number of subordinate bones, and each subordinate
bone can also have any number of subordinates, etc. The purpose of the chart
is to isolate the factors which can then be worked on to solve the problem or
reach the goal they help cause.

FACTOR FACTOR

PROBLEM
S OR
GOAL

FACTOR FACTOR

EFFECT
CAUSES

figure 2

11
 Checkliscs are simply a means of collecting data. The idea is to
record happenings as they occur against a list of possiDle events. This
provides factual, rather than anecdotal, evidence of what is really
happening. (see figure 3.)

DEFECT M T W TH F TOTALS

A I I I 3

B II I II I I 7

C I I 3

Figure 3

Histograms are used to show the distribution of outcomes of a

repetitive event. Figure 4 snows the typical normal distribution of an event
such as the length of rods produced by a machine. Of interest would be the
mean and variation shown by the histogram. Figure 5 shows a bi-modal
distribution, which indicates that the output charted comes from two separate
processes (two machines, two shifts, two stocks of raw material, etc.) whose
products are mixed.

Figure 4

12
 Figure 5

Pareto charts show graphically the relative magnitude of output from

different factors. Figure 6 is an example of injuries separated Dy
location. Pareto charts can be nested; the largest output on one chart is
separated on another Pareto chart as figure 7 separates the causes of the eye
injuries shown of figure 6. Also, Pareto charts can be used for
stratification. The same data is plotted on several charts which separate it
by different factors. For example, figure 8 shows the number of defects to
relate to product line rather than factory or shift.

HAND EYE FOOT

INJURIES

Figure 6

13
 '04
HAND EYE FOOT CUT BURN SCRAPE
INJURIES HAND INJURIES

Figure 7

DEFECTS

1 2 1 2 3 1 2 3 4 5
BY FACTORY BY SHIFT BY PRODUCT LINE

Figure 8
Scattergrams are simply a test for correlation between two factors.
Data is gathered and for each point, the value of one factor is plotted
horizontally and the other vertically. If the resultant cloud of dots
clusters around a line, correlation is indicated. (See figure 9.)

Control charts, the last of the basic tools, will be covered in a

following discussion of statistical process control. Before that, there are
some other simple tools I would like to present.

14
 2 im n co ,

C. C .-
.w. .

PARAMETER A PARAMETER A
CORRELATION NO CORRELATION

Figure 9

A force field simply plots the forces which support and oppose some
effect. Like the Ishikawa diagram, its purpose is to identify causal factors
for further analysis as candidates for change to increase or decrease the
prooability of the effect happening. (See figure 10.)
FORCE FIELD

GOAL: CREATE THIS TUTORIAL

PRO > < CON

MISSIONARY SPIRIT > < LOTS OF WORK

WRITER'S EGO > < PAIN OF CREATION

LOYALTY TO RAMS > < OTHER COMMITMENTS

TRIP TO LAS VEGAS > < TOO MANY FOILS TO CARRY

RESEARCH DONE > < FEAR OF MISTAKES

Figure 1U

A "measles" chart is a graphic form of check list showing the

locations of some event of interest. Figure 11 shows the locations of
failures on a printed circuit card.
A run chart plots data against a time scale to show trends and
periodic effects. (See figure 12.)

15
 XX X= LOCAION OF

x DEFECT

Figure I1

DEFECTS

TIME

Figure 12

None of the tools so far described are difficult to understand.

However, statistical process control does take some explaining.

STATISTICAL PROCESS CONTROL

Statistical process control recognizes that every process has some

variation. "Common cause variation" is random and predictable and descriDes
the variation inherent in a process. Hence it cannot be reduced without a
process change. "Special cause variation" is variation outside that expected
for a process and hence due to some special cause which can be isolated and
eliminated. The process is "in control" when measurement snow only common
16
cause variation. Note that "in control" does not necessarily mean tne
process is producing products in specification. It can De in control and
also not capable of producing the desired products. If so, only a oetter
process can provide the desired product; a process in control is doing the
best it can.

If samples taken from a process in control are measured, the central

limit theorem states the overall mean of the means of the samples will equal
the process mean, and the sample means will be normally distriDuted, whicn
implies that 99.7% of the sample means will De within three standard
deviations of the overall mean. Therefore, when sample means vary randomly
within three standard deviations, the process is in control. Otherwise
corrective action may be required to eliminate a special cause of variation.

Figure 13 snows a generic control chart. The center line is the

expected process mean (or the specified target value). The upper control
limit (UCL) is the mean plus three standard deviations of the sample data
(sigmas), and the lower control limit (LCL) tne mean minus tnree sigmas.
(Note: If the LCL computation produces a negative number, the LCL is set to
zero.)

MEAN -

LCL

Figure 13

Sigma is computed from the sample data. It will be a function of the

sample size and the standard deviation of the process. The computation
differs depending on whether the parameter of inte'est is a variable, an
attribute, or a rate.

Controlling Variables

A variable is a measured parameter such as the length of a rod. For

that case, the centerline of a control chart would be the mean rod length
(population mean estimated from a series of samples or the target mean). In
addition, a second control chart, the range chart, would probably be used to
record the variations in length between rods.
IC
 using rod lengths for illustration, parameters of interest for varidole
and range charts would be:

X; =length of one rod

n = number of rods in one sample

X = mean of sample =iX;/n = one point on tne control chart

K = number of samples over a reasonably long time period

X = mean of process =EX/k (if all samples are equal) = centerline

R; range of one sample = highest X; - lowest X;

= average range =fR;/k = centerline of range chart

In basic form, the variable or X chart has a centerline of X, and the

control limits are given by tne following formulas:

UCL = X + 3 S /df, LCL = X - 3 S'/.r"

where: S = population standard deviation

S - I 2
-X.
S(X I"X)
n-1 1

OR: SR: S/c

where: S = sample standard deviation

S= -if"g)

and: c = a constant dependent on sample size, shown in table 1.

Using range data simplifies computations. When an average range (R) is

obtained from a long series of samples, the control limits on the X chart
are:

UCL = X + AR, LCL = X - AR

where: A1 is taken from table 1.

18
 The range (R) chart is concerned with the variance between parts rather
than between the sample means. For the R chart:

Centerline = R, UCL = DR, and LCL = D3R

where: D,,D 3 constants obtained from table 1.

CONTROL CHART CONSTANTS

n = 2 3 4 5 6 7 8 9 10 12 Ib 20

C= .5b .72 .79 .84 .87 .89 .90 .91 .9Z .94 .95 .96

A2 = 1.88 1.02 .73 .58 .48 .42 .37 .34 .31 .27 .22 .18

D3 = 0 0 0 0 0 .08 .14 .18 .22 .28 .35 .41

D4 = 3.27 2.57 2.28 2.11 2.00 1 92 1.86 1.82 1.78 1.72 1.65 1.59

Table 1

Controlling Attributes

An attribute is a feature of a product which is either present or not,

such as a defect. To control attributes, the centerline of the chart is set
equal to (p), tne proportion of the product with the attribute, estimated
from many samples. The control limits are given by:

UCL + LCL p -3
V n n

When sample size is fixed at (n), the centerline can be set to (np)
which is the mean number of items with the attribute in a sample of n units.
Using np as the centerline:

UCL = n + 3 n'(1- ) LCL = n - 3j/np(1-h)

19
 Controlling Rates

To control rates, such as defects per aircraft, or defects per IoU feet
of wire, etc., the centerline of the chart is set to (r), the mean rate
estimated from a lot of data. Then:

UCL = r + 3 jr LCL = r - 3w'r

For rates without constant sample sizes, such as the number of defects
per unit in a variable monthly production, the centerline would be (u) which
is the mean rate per unit, and:

UCL = u + 3J LCL = u - 3 f

n = units in sample. Note that the control limits may change from sample to
sample if n is not constant.

ADVANCED ANALYTICAL TOOLS

This section will very briefly discuss some of the advanced analytical
methods associated with TQM. Quality function deployment (QFD) and tne
statistical design of experiments (DOE) will be described. Then we will look
at some of the contributions of Genechi Taguchi.

Quality Function Deployment

QFD is based on a matrix comparing "whats" to "hows." The "whats" are

the customer's requirements and the "hows" are the organization's responses.
For example, figure 14 shows the desired attributes of a fighter plane
against the design factors which may have an impact. In the matrix are
symbols showing the relationship, if any, between each item on the two
lists. The matrix highlights the important "hows," with figure 14 showing
shape to be critical and the plant site to be unimportant in meeting the list
of mwhats."

This simple matrix can be cascaded and/or expanded to provide

additional information.

Cascading the QFD matrix means making the "hows" of one matrix the
"whats" of a subordinate matrix. In figure 15, the first matrix translates
customer specifications (top level "whatsm) into design solutions ("hows").
These design solutions become the "whats" of the next matrix matched against
"hows" in terms of the parts used. The next tier matches the parts against
the processes. The final matrix matches the processes against process
parameters. Hence, there is a defined trail leading from the customer's
specifications to such details as the size of a spray paint nozzle. The
trail followed need not be the one described in figure 15, so long as the
requirements are decomposed in a logical fashion from the top level "whats"
to the detailed process "hows."

20
 HOWS
0 VERYIMPORTANT
iN ) HASIMPACT

Z CC 0 OR

SPEED o *
SEC o __ A LARGEIMPACT

STEALTH --MODERATE iMPACT

WHATS ARMAMENT - -__ _
0 0 SOME IMPACT

AVIONICS 0 OR
RANGE 0 0 0 YOUR CHOICE

Figure 14
HOWS

0 ~0 Z~n~
W HATS 0 CL ; o o
DULLFINSHIC
0 LACK PAI 0I-H ENAMEL .0... SPA-00
SML
U VLSI I__ CMOs BULY LL
BASIC REQUIREMENTS DESIGN VS PARTS SELECTION PROCESSES VS
VS DESIGN PARTS SELECTION VS PROCESSES PARAMETERS

Figure 15

Expanding the QFD matrix means adding more details of interest. One
obvious detail is the priorities of the customer, which can De added to the
"whats". A summary rating of the importance of each "how" would seem useful
and can be added to the "how" columns, along with target values, estimated
difficulty, etc. Benchmarks of our organization's capability for meeting the
"whats" and our competitor's capabilities could be added. Since the "hows"
are not independent (one design feature can reinforce or diminish the effects
of another), the relationship between "hows" could be worth displaying.
Figure 16 shows one version of a "house of quality," so called because of the
shape resulting from expanding the QFO matrix. Figure 17 shows the house of
quality for this tutorial. Expanded QFD matrixes such as the house of
quality can also be cascaded.

QFO requires a lot of effort. Essentially, it invests time in planning

to reap a profit in the design phase for an overall shorter development
cycle. It also minimizes the need for redesign, which can dramatically
shorten the overall time from start to production. The exercise of working
with your customers to define their desires and priorities before you go
charging off to design the solution is always extremely useful, whether you
use a simple matrix or a "mansion of quality."

21
 HOUSE
OF tHOW vs. HOW IMPACT

QUALITY

0 000 0 a

WHAT
.BENCHMARKS
WHAT OR
RANKINGS
WHAT
WHAT

- IMPOHTANCE RATING

4- TARGET VALUES

- DIFFICULTY

4- ETC.

- ETC.

Figure 16

SAMPLE A STRONGLY RELATED

HOUSE OFTUTORIAL)
(FOR THIS QUALITY A U RELTEo

X OPPOSED

COMPLETE 1 (4 9 1-10 SCALE

USERS 2 HOURS 2 ( 1 2 LENGTH IN HOURS

NEEDS -
7 9 1-10 SCALE
IINusmG 3 0 o 0
USEFUL 4 0 0 11 9 1-10 SCALE

IPORTANCE 0 0 0 *
12
GL WA IF*WA 0 IMPORTANT
GOAL FA A .j HAS IMPACT

OIKC"LTY PA EASY MOO.IA~ .AIY

Figure 17

22
 Statistical Design of Experiments

Statistical design of experiments (DOE) is an organized approach to

determining the effects of process parameters on its output. For example, we
could do experiments on soldering with temperature, solder tin-lead ratio,
wave machine neight setting, etc. as test factors and the solder defect rate
as output. From this, we wish to determine the setting for each factor which
will give us the lowest defect rate. There are many different ways we could
run tests, ranging from varying one factor at a time to complex comoinations
of factors and setting. DOE procedures are designed to provide results wnicn
are statistically valid, unambiguous, and economical. The general procedure
is:

1. Select factors. It is not always obvious wnich factors are important.

Factors are often selected after a brainstorming session has filled in an
Isnikawa chart, and the team involved has ranked the hypothesized causal
factors in their consensus of priority.

2. Select test settings: Usually, a high and low setting are selected. For
something like the presence or absence of a a.ur, ti'e high setting is
present and the low factor, absent. Sone items cannot be done in two
settings. For example, we may h'w five different fluxes we can put in the
solder. This requires five "settings," at least. While DOE can handle such
cases, this discussion will e limited to two value experiments.

3. Set up an appropriate orthogonal array. An orthogonal array balances the

test settings so that, for example, factor A is tested at both high and low
settings of factor B, so that the effects of factor B will not confuse our
evaluation of factor A. In addition, interactions between the factors can be
easily determined. Figure 18 is an example of a two factor orthogonal
array. Note that the high setting for each factor is represented by d plus
sign and the low factor by a minus sign. This is shorthand for plus one and
minus one, because the high and low values of each factor are coded as plus
and minus one. (e.g. temperature of 200 degrees is -1, 400 degrees is +1;
switch off is -1, on is +1; etc.) This permits some computational
shortcuts.
FACTORS INTERACION
(TEST SETTINGS) (CALCULATED)
RESULTS
RUN A U A- B (MEASURED)

1 -- - + y1
XOW slrIIM
2 + - - y

3 (A + - 2A
Y3
4 + + + 4Y + 2AA+ a+ A-9

AVG - JiJ+ X.+i2

2 2 2
AVG + Y2+ 4 YS+Y + ..
.XY4 Y +Y
2 2 2

(AVG +) - (AVG .

Figure 18
23
3. Run the tests. The tests can be run once at each setting shown in the
rows of the array, or multiple times to gather information on the variability
of the measured results.

4. Analyze the results. The difference in output for different factor

settings can be measured and charted. For the examples shown here, it is
also easy to create a regression equation which will permit the determination
of optimum settings at points not actually tested. The regression formula is
shown in figure 18, and figure 19 shows some possible results. It is assumed
that the effects of the factors are linear over the range tested.

5. Calculate optimum settings. The regression equations in figure 19 show an

example of this for maximum, minimum and target values.

6. Do confirmation run test. Since there may be important factors not

considered or non-linear effects, the optimum settings must be verified by
test. If they check out, the job is done. If not, some new tests must be
planned.
RUN A B AeB Y
1 - " + 10 109
2 + " 6 8 7
3 " + 8 \
4 + + + 4
AVG - 9 8 7 -Y=7

AVG + 5 6 7
4 2 0 + - + +
A B AB
2A-B
Y 7-
FOR MAX Y, SET A, B TO -1 (LOW SETTING)
FOR MIN Y, SET A, B TO +1 (HIGH SETTING)
FOR Y = 5 SET A TO +1 (HIGH SETTING)
B TO 0 (MIDWAY BETWEEN HIGH & LOW)
Figure 19

Taguchi Methods

Modifications to DOE are a specialty of Genichi Tagucni, who uses a

modified array and analyzes output based on a "signal to noise ratio" which
is in turn based on a "loss function" relating variability in a product to
economic loss in its use. Figure 20 shows a Taguchi array. A notational
difference between figures 18 and 20 is that Taguchi uses I and 2 wnere
others might use -1 and +1 for the lower and upper settings. The significant
difference, however, is that the column used for interaction data has been
replaced by another factor, factor C. Taguchi assumes the interaction
between factors will usually be negligible and hence the confounding of
interaction data with factor data will be no problem. This permits three
factors to be tested where only two could be before. This economy compounds
with the number of factors. For n factors the orthodox array needs a numoer
24
of test runs equal to two raised to the ntn power. In that many runs, the
Taguchi array can test a number of factors equal to one less than the numDer
of runs. Figure 21 shows the difference in eight runs. This economic
advantage is significant; the risk is that interactions may prove to De
important. Figure '22, for example, show a drastic interaction between
factors A and B. If a Taguchi array were used, the interaction effects would
be attributed to factor C in addition to any effects that C caused itself.
(The error would presumably De found after a confirmation run failed.)

RUN A B C Y

1 2 2 1

2 1 2 2

3 2 1 2

4 1 1 1

Figure 20
(D) (E) (F) (G)
RUN A B C AB A-C B.C A.B.C
1
2
3 3 FACTORS TESTED ALL INTERACTIONS VISIBLE USING
USING ORTHODOX ORTHODOX ARRAY
4 OTHOGONAL ARRAY OR
4 ADDITIONAL FACTORS TESTED
5 USING TAGUCHI ARRAY
6
7
8

Figure 21

RUN A B AB Y 11
1 + 0 10 8 8 9
2 + 4
3 + 6
4 + + + 12 5
AVG - 8 7 Y8
AVG. 8 9 11

, 0 2 6 - + B + - +
A B A.B

Y 8 + 0 + B + 3 (AB)

Figure 22

25
 Instead of the measured output data from an experiment, Taguchi
recommends the use of logarithmic transformations which consider both the
mean values and variability. These "signal to noise ratios" are shown in
figure 23.

SMALLER IS BETTER NOMINAL IS BETTER

S/Ns = -10 Log 101 Y(Yi) S/N = 10Log Sm"Ve

1 10 n V
2
2
LARGER IS BETTER where Sm=(Yi)
n and Ve= yi in
n-1

S!NL -10 Logion

, 2

Y1 =ONE OBSERVATION

n =NUMBER OF REPLICATIONS OF TEST RUN

Figure 23

The use of signal to noise ratios (S/Ns) requires replicated runs to

obtain the variability data. The factor making the largest change in the S/N
is considered the most important factor to control in order to minimize loss,
since loss is related both to mean value and variability of the output. This
is according to Taguchi's loss functions. The basic premise of loss
functions is shown in figure 24, which compares the traditional concept of
specifications to the loss function.

NO
LOSS
ALL NO LOSS
LOSS ALL LOSS LOSS
LOSS LOSS

A B

LSL T USL LSL T USL

"GOALPOST" CONCEPT LOSS FUNCTION

ALL PRODUCT BETWEEN WIDE VARIATION IN PRODUCT IS BAD

USL & LSL IS GOOD MEAN OFF TARGET IS BAD

Figure 24

The loss functions themselves are shown in figure 25. In theory, it is

possible to compute a dollar loss from the distribution of the products.
26
This can be a loss to the producer from rejects or rework, or a loss to
society from the effects of variation. The author accepts the concept, but
would be leery of any computed dollar loss values.

SINGLE UNIT MULTIPLE UNITS

NOMINAL L = k1o 2 +
IS BEST L = k(y - T)2
2
A2= (yT)
S-SS-
ISI T USI

IS BETTER L k 2 i2 2
Y LSL

SMALLER L = k(j 2 + a )
IS BETTER L = k(y 2) US

Figure 25

TEAM TOOLS

The use of teams to improve an organization's processes is a basic

tenet of TQM. Whether called Process improvement teams, process action
teams, Delta teams, or (erroneously) quality circles, the objective of the
team is to improve quality by cnanging a process. It is possible to start a
team by merely convening a bunch of people and giving them an objective. For
best results, however, the team should have some knowledge of the quality
concepts and basic analytical tools described above. It also helps to have
some knowledge of group dynamics (how teams work), some tools to evaluate
their team process, and some tools that the team can use in performing their
process. This section will discuss teams and their tools, finishing with
some advice on getting started with team activity. Incidentally, the quality
circle was invented in Japan as a means of educating employees in the
concepts of quality, with improving processes as a secondary objective.
Hence the significance of the improvements is not terribly important and the
members of the circle are typically from a single office. In contrast, the
improvement team's main function is to make significant cnanges in an
agency's processes; hence there is more concern for the payoff and,
typically, a multi-office membership. The team is more likely to be an ad
hoc effort which disbands after the project is completed while the circle
continues withanother project. The Harris Corporation's "Employee
Im"r-9vement Teams" do seem to be a hybrid: They are single "work cell"
entities, and permanent teams with sequential projects, but the projects are
assigned by management, presumably with significant benefits the main goal.
(Harris also used multi-office "System Improvement Project" teams on an ad
hoc basis.)

27
 TEAM DYNAMICS

Figure 26, or version tnereof, is a common description of the pnases a

team goes through. It is called the team wheel or team clock.

4
SYNERGY FORMATION

3 2
COOPERATION CONFLICT

Figure 26

What the wheel indicates is that a newly formed team goes through a
formation phase wnere it tries to establish its goals and working procedures,
and each member tries to assess his (or her) value to the team and the team's
value to him. This can lead to conflict as the goals of the team members
probably will not agree. Conflict is not necessarily bad, if it is addressed
with good will and resolved to create a cooperating team. When a cooperating
team has evolved, the members have a sense of fellowship and work effectively
towards their goal. After this, there may be periods of synergy, when the
group works with high creativity, and the individual members feel a great
sense of loyalty to the group and a keen enjoyment of the group process. It
is generally the best state to be in, but there is a potential dark side,
called "grouptnink." This state is characterized by a feeling of
invincibility in the group and a tendency to submerge any individual
reservations about group decisions, sometimes leading to fantastically gross
errors. However, so long as the group keeps touch with reality and discusses
issues thoroughly, they will make tneir best contributions while enjoying the
natural high of synergy.

Team Roles

Members of a team will have differing personalities, and this diversion

is a source of team strength. Students of group dynamics have created
various taxonomies to describe the roles played by individuals on the team.
Table 2, taken from the TQM training provided to the Air Force Electronics
Systems Division by Coopers & Lybrant and Change Navigators, is one example,
describing four task oriented roles and four (team) process oriented. An
individual may play different roles at different times.

28
TASK ORIENTED TEAM ROLES:

SHAPER - Keeps team focus on objectives

INNOVATOR - Source of ideas

ANALYZER - Evaluates ideas

IMPLEMENTER - Concerned with getting things done

PROCESS ORIENTED TEAM ROLES:

COORDINATOR - Concerned with achieving consensus

NETWORKER - Connects team to outside world

HARMONIZER - Concerned with feelings

GATEKEEPER - Concerned with keeping team standards

Table 2

Each of these roles has a place in helping the team work. Without a
snaper, the team may drift. The innovator, who provides ideas, is
complemented Dy the analyzer, who evaluates them. And so forth, down to the
gatekeeper, who notices when the team is violating its established standards
of conduct and brings it to the other members' attention. One Rome
Laboratory team dramatically improved its performance after they realized
that they lacked a shaper and recruited one.

Team Decision Process

Teams should strive to reach decisions by consensus. Consensus is

reached when all the team members feel that the decision is the best possible
under the circumstances and acceptable for their special interests. A
decision reached by consensus will be supported by all concerned, while other
ways of reaching decision will have lesser "buy-in" as shown in table 3

DECISION PROCESS "BUY-IN"

NO DECISION NONE

POWERFUL MINORITY RULES SMALL MINORITY

TRADE-OFFS BETWEEN SPECIAL INTERESTS MINORITY

MAJORITY RULE MAJORITY

MAJORITY RULE - MINORITY OPINION EXPLORED MAJORITY

CONSENSUS ALL

Table 3

29
 TEAM PROCESS EVALUATION TOOLS

A team should periodically evaluate its own process. There are a

variety of nelpful tools for this. Using the information in the previous
section, a team can evaluate its position on the team wheel, assess tne roles
played by the team members (are all necessary roles present?), and discuss
the degree of participation in the decision making process. They can also
rate the team climate with a variety of surveys. Generally, tnese use a
scale rating each member's agreement with such statements as:

- We are addressing the proper issues

- We are able to openly express ideas
- We feel comfortable giving feedback
- We spend our time well
- We often stray from the issue
- Everyone is participating
- We are following our code of conduct well
- Our goals are clear
- We have a high level of energy
- We usually achieve consensus
-etc.

The insight provided by the discussion of the survey results will help
keep the team operating effectively. Surveys are available in text books and
consultant's manuals. With a little thought, the user can probably create an
effective survey on his own.

Facilitators

An objective view of a teams effectiveness, and useful feedback, can De

obtained by using a facilitator. A facilitator is someone trained in group
dynamics and team tools who monitors the team progress. He should not do any
work on the team task, but rather devote all his attention to the team
process. He may teach the use of evaluation and process tools. He will
advise the team on-line and the team leader off-line. One facilitator may
support many teams and need not attend every team meeting. The facilitator
has his own tools. One is the interaction chart shown in figure 27.

Figure 27

30
 Figure 27 snows the communication during a segment of the team
meeting. Each line represents an input from one of the team members. These
can be coded by length or width to represent the relative length of the
talk. Tne arrowhead on the line shows who received the input, with
statements directed to the whole group pointed at the center. Figure 27
shows that tne person labelled A dominated the discussion, person B did not
participate and E and F held a separate meeting.

The facilitator can also make use of cnecKlists to organize his

observations. There are many of these, often with a particular focus.
Examples are shown in table 4.

FOCUS SAMPLE ITEMS

BEHAVIOR Who: is friendly, holds back, shows tension, wants data, etc.?

BODY LANGUAGE Who: avoids eye contact, frowns, leans back, stands up, etc.?

LEADERSHIP Style of leader, effectiveness, decision process, etc.

ROLES Are all necessary roles present?

GENERAL What helped/hindered team? Does anyone dominate? Could the

team hold a focus? Was it O.K. to disagree? etc.

Table 4

A facilitator can use available lists or make his own based on what
observations could be most useful to the team.

TEAM PROCESS TOOLS

The team tools discussed above are designed to help the team work
effectively, by evaluating their team process. Another set of tools is
designed to aid their efficiency in going through the process to reach their
objective. These include:

- Brainstorming
- Interviewing
- Affinity exercise
- Multi-voting
- Nominal group technique
- Pairwise ranking
- Mental imaging

BRAINSTORMING is used to generate ideas. The team leader starts the

session by writing the topic where everyone can see it. He then explains
that the goal for the exercise will be to generate as many ideas as possible
and no idea will be analyzed or evaluated at this time. He may then proceed
to a structured or unstructured approach. In the former, he will go around
the table and each member will contribute one idea or pass, repeating the
process until all pass. In the unstructured approach, team members
contribute ideas as they come to them until all run out. Either way, the
31
ideas are written down as they come. A brainstorming session typically runs
about 15 minutes; the results are often amazing. Once the ideas are
captured, they can be discussed, organizeo, and evaluated.

INTERVIEWING is the way Lne team obtains inputs from stakeholders in

the project who are not on the team. The process starts with the
identification of the stakeholders, and the team should strive to include all
possiole viewpoints. A list of questions snould then be prepared (or a set
of lists, if appropriate), and appointments made with the people to be
interviewed. Two interviewers are recommended; one does the talKing and the
other takes notes. The stakeholder should do most of the talking and should
be encouraged to fully express nis views. The team members should never
argue or judge the input during the interview. The stakenolder should be
thanked for nis input, and the team must scrupulously perform any follow up
activity promised to the stakeholder. It is usually appropriate to give the
stakeholder some feedback of results, sucn as a copy of the team's final
report.

THE AFFINITY EXERCISE is another idea generator. It starts with a

silent brainstorming session. For 15 minutes, the participants write their
ideas on post-it notes, one idea per note, without any discussion. All notes
are then posted on a wall, and for 20 minutes the participants read the notes
and group similar idea by putting the notes together. This also is done
without discussion. Then a discussion is held to develop a theme for each
group of notes. The theme will be a noun-verb combination such as "reduce
paperwork." The themes are then grouped into action plans for making
improvements. The silent brainstorming is used as it produces far more ideas
than the usual procedures, despite the fact that people cannot build on each
others ideas.

MULTI VOTING is a way to reduce a list of ideas or recommendations to

those most important. Each member of the team is given a number of votes
equal to about half the number of items on the list. Everyone votes for the
items they consider most important. The four to six items getting the most
votes get the priority. If votes are too close to isolate the top items, the
items receiving few votes can be eliminated and a new multi vote taken on the
remainder.

THE NOMINAL GROUP TECHNIQUE is another way to prioritize a list. After

similar ideas have been combined, each item on the list is ranked by each
team member. Item I is the least important, item 2 the next least important,
etc. The individual scores from each team member are then added together,
and the item with the hignest score is the most important.

PAIRWISE RANKING assigns priorities by comparing the items on a list to

each other one by one. A matrix is used, as shown in figure 28. The team is
asked to compare list item I against 2 and the number of the preferred item
is recorded in the top box of tne matrix. The next row snows the winners of
comparisons between items I and 3 and between items 2 and 3. After all items
have been compared, the number of times each item number has been recorded in
the matrix is calculated. The item whose number is recorded most is the most
important. Should two items tie, the winner of the comparison between the
two takes precedence.
32
 1.NUMBER ITEMS 1 2. COMPARE ITEMS BY PAIRS,
CONSTRUCT MATRIX COMPARE RESULTS
2 2

3 1IBETTER THAN 2
1 BETTER THAN 3
4 4 2 1 2 3BETTERTHAN2
4BETTER THAN 1 1J3
5 5BTERHN 3 1 3 3 4BETTER THAN 3

6 44 4

I1 1111 ..
Figure 28

MENTAL IMAGIN 2 is an aid to planning. Each team member relaxes and

imagines what tr. world would be like if the best possible outcome to their
project were ,inplemented. Their ideas are recorded and combined into a group
vision. Then the current situation is assessed, and the gaps Detween the
vision an, current reality defined. The obstacles to closing the gaps are
then identified and become the targets for the improvement plan.

GETTING STARTED

It takes some effort before a group of individuals becomes an effective

team. Some never make it. Successful teams often have some help getting
started in the form of specialized training. I recommend an orientation in
group dynamics to include the team wheel, team roles, decision processes, and
team evaluation tools. (The team process tools can be provided also or
deferred until they are needed. The same applies to training in the
analytical tools.)

The first team meeting should produce a cnarter and a code of conduct.
The charter may be the objectives given the team by its sponsor, if specific
enough. If not, the team should refine it to a clear objective that everyone
understands and agrees to. For example: "Find ways to reduce the time
required to obtain laboratory supplies." (Refinements should be subsequently
verified by the sponsor.) The code of conduct is a set of rules the team
intends to abide by in their activities. It might include:

- Be on time for scheduled meetings

- All opinions respected
- All statements held private
- No smoking
- 10 minute break every hour
- No hidden agendas
- etc.

33
 With the charter adrd ?:-de f conduct, the team: ssarting in an
organized fashion, The nelr of d trained facilitdtor is probably the most
valuable in start up, to head off any Dad nabits. Some good naoits to start
with are the recording of minutes, the making and following of agendas, and
the evaluation of task progress and team process at the end of each meeting.
A simple evaluation should usually suffice (Is progress good or not, and
wnere are we on the team wheel?) with more involved evaluation used wnen tne
simple answers are not good (getting nowhere and still in conflict).

ACTION PLANS

Action plans are ways of organizing quality improvement efforts. Tne

most well known are the Shewhart cycle and the Quality Improvement Story.
The Air Force Electronic Systems Division uses one called "Chart It - Check
It - Cnange It" (C-cubed-I) whicn tracks, nomenclature-wise, with its mission
to provide Command, Control, Communications and Intelligence (C-cubed-I)
systems. Boeing uses a procedure simply called a seven step model for
process improvement.

THE SHEWHART CYCLE is shown in figure 29. It is named after Walter

ShewnarL, a pioneer in statistical quality control and a mentor of Dr.
Deming. It is also known as the Deming cycle because of its advocacy by
Deming. It starts with planning, which could oe tne planning of an
experiment, the identification of data needed to analyze a process, or the
planning of any activity for quality improvement. The next step, tne "Do"
cycle, is to follow through on the plan. The results of the Do phase
activity are analyzed in the "Check" phase. The last phase, "Act," is to
follow up on the conclusions from the analyze phase. From the Act phase, one
returns to the planning phase for the next improvement project. Each trip
around should result in a better situation for the cycler.

4. AC. 1. PLAN APPROACH

- DATA TO GET
- CHANGE GOALS
- EXPERIMENTS TO RUN
- CHANGE PROCESS
(THEN GO TO 1) 4
rACT PLAN

3 2

3. CHECK RESULTS CHECK DO 2. DO THE PLAN

- ANALYZE DATA - GATHER DATA
-EVALUATE EXPERIMENT - RUN EXPERIMENT

Figure 29

The QUALITY IMPROVEMENT (QI) STORY is a sequential procedure for

action. There are various versions. One of these is:

34
 QI STORY
1. THEME - What is the problem/opportunity?

2. SCHEDULE - The plan to finish the story.

3. CURRENT CONDITION - What are the facts? (collect data not anecdotes.)
4. CAUSE-EFFECT ANALYSIS - Determine & prioritize causes of current
condition.
5. CAUSE VERIFICATTON - Collect data to prove cause "guilty."

6. COUNTERMEASUPES - What can be done to remove cause?

7. COUNTERMEASURE EFFECTIVENESS - Does it work? (data)

8. TOTAL EVALUATION - Will countermeasures mess up something else?

9. STANDARDIZATION - Make the improvement permanent.

10. FUTURE ACTIONS - What's next? (go to 1)

The outline for CHART IT - CHECK IT- CHANGE IT (C-CUBED-I) is:

1. CHART IT:
- Identify process, put into flow chart
- Gather data on process

2. CHECK IT:
- Analyze data, isolate problems/opportunities in process
- Identify alternate approaches
- Select opportunities for improve;nent (useful alternatives)
3. CHANGE IT:
- Change process to implement improvement
- Standardize change

The BOEING SEVEN STEP MODEL FOR PROCESS IMPROVEMENT is:

1. ASSIGN PROCESS OWNER AND DEFINE BOUNDARIES
2. FLOW CHART PROCESS

3. ESTABLISH EFFECTIVENESS AND EFFICIENCY MEASURES

4. DETERMINE PROCESS STABILITY

5. IMPLEMENT PROCESS IMPROVEMENTS

6. VALIDATE IMPROVEMENT

7. DOCUMENT IMPROVEMENT
35
 ASSESSING QUALITY EFFORTS

The Malcolm Baldridge National Quality Award

The Malcolm Baldridge National Quality Award was e tdblisned in 1967 to

stimulate American companies to improve quality, recognize achievements,
esta.olsn guidelines for self-evaluation, and make available information from
successful organizations. Winners in 1988 were Glooe Metallurgical,
Motorola, and a nuclear fuel division of Westingnouse. In 1989, Milliken and
Xerox Business Products took the awards. In 1990 it was Cadillac, IBM
Rocnester, Federal Express, and the Wallace Company. Tne award is not
without detractors, who argue tnat the Baldridge criteria may not be the only
way to world class quality. They note, for example, that the empnasis on
participatory management assumes that other management approaches would not
work. Nevertneless, the Baldridge criteria is a de facto standard for
judging the quality efforts of an organization and the oasis for most other
criteria such as a growing number of regional quality awards. The 1991
Baldridge criteria and the relative weights given to each factor are shown in
table 5:

FACTOR WEIGHT

1.0 LEADERSHIP 100 POINTS

2.0 INFORMATION AND ANALYSIS 7U
3.0 STRATEGIC QUALITY PLANNING 60
4.0 HUMAN RESOURCE UTILIZATION 150
5.0 QUALITY ASSURANCE OF PRODUCTS AND SERVICES 14U
6.0 QUALITY RESULTS 180
7.0 CUSTOMER SATISFACTION i00
1000

Table 5

LEADERSHIP is further divided into the subfactors of Senior Executive

Leadership (40 points), Quality Values (15), Management for Quality (25) and
Public Responsibility (20). INFORMATION AND ANALYSIS includes Scope and
Analysis of Quality Data and Information (20), Competitive Comparisons and
Benchmarks (30), and Analysis of Quality Data (20). STRATEGIC QUALITY
PLANNING includes Strategic Quality Planning Process (35) and Quality Goal
and Plans (25). Subfactors of HUMAN RESOURCE UTILIZATION are Human Resource
Management (20), Employee Involvement (40), Quality Education and Training
(40), Employee Recognition and Performance Measurement (25), and Employee
Well-Being and Morale (25). QUALITY ASSURANCE includes Design and
Introduction of Quality Product - ,-,d
Services (35) Process Quality Control
(20), Continuous Improvement of Processes (20), Quality Assessment (15),
Documentation (10), Business Process and Support Service Quality (20), and
Supplier Quality (20). QUALITY RESULTS is divided into Product and Services
Quality Results (90), Business Process, Operational, and Support Service
Quality Results (50), and Supplier Quality Results (40). Finally, CUSTOMER
SATISFACTION includes Determining Customer Requirements and Expectations
(30), Customer Relationship Management (5U), Customer Service Standards (20),
Commitment to Customers (15), Complaint Resolution for Quality Improvement
(25), Determining Customer Satisfaction (20), Customer Satisfaction Results
(70) and Customer Satisfaction Comparison (70).

36
 Each of these subfactors is further divided into two to four areas to
address. As a result, Baldridge applications take significant effort.
However, the process gives the preparer a deep insight into nis
organization's quality efforts. Without any intent to apply for the award,
an organization can take advantage of the criteria for a self-assessment. A
copy of the Applications Guidelines can be obtained, free, from:

Malcolm Baldridge National Quality Award

National Institute of Standards and Technology
Route 270 & Quince Orchard Road
Administration BLDG, Room A537
Gdithersberg MD 20899
(Telephone 301-975-2036)

The Quality Improvement Prototype (QIP) Award

The Quality Improvement Prototype (QIP) Award was established to:

1. Recognize (Federal) organizations that have successfully adopted TQM

principles and thereby improved the efficiency, quality and timeliness of
their services or products.

2. To use the QIPs as models for the rest of government, showing other
agencies how a commitment to quality leads to better services and products.

The QIP evaluation criteria and weights are shown in table 6:

QUALITY ENVIRONMENT 20 POINTS

QUALITY MEASUREMENT 15
QUALITY IMPROVEMENT PLANNING 15
EMPLOYEE INVOLVEMENT 15
EMPLOYEE TRAINING AND RECOGNITION 15
QUALITY ASSURANCE 30
CUSTOMER FOCUS 40
RESULTS OF QUALITY IMPROVEMENT EFFORTS 50
200

Table 6

QIP applications are available from:

Quality Improvement Prototype Award

c/o Quality Management Branch, Room 6235
Office of Management and Budget
725 17th street, N.W.
Washington DC 20503

Quality and Productivity Self-assessment Guide

for Defense organizations

Designed specifically for self-assessment, the Quality and Productivity

Self-Assessment Guide for Defense Organizations contains rating scales for
37
Climate (peoples perceptions about the organization), Processes (tne
organizations policies and practices), Tools (specific techniques used to
promote quality) and Outcomes (results). The climate questions can be given
separately with the idea that a broad survey of climate and a smaller sample
of the other factors would be effective. A scoring guide includes
suggestions to help raise low ratings. The guide is available in an
automated format for use with a personal computer.

Copies may be requested from:

John Denslow
OASD/DPPO
Two Skyline Place, Room 14U4
5203 Leesburg Turnpike
Falls Church VA 22041-3466
(Telephone 703-756-2346)

MEASURING THE QUALITY OF KNOWLEDGE WORK

Measuring quality is most difficult for the managers of knowledge

workers. One reason is that there is no universally accepted standard
definition for quality. Indeed, David A. Garvin identified five categories
of definitions for quality. These are:

1. Transcendent quality: a subjective feeling of "goodness".

2. Product-based quality: measured by attributes of the product

3. Manufacturing-based quality: conformance to the specifications

4. Value-based quality: "goodness" for the price

5. User-based quality: the capacity to satisfy the customer

One should note that the categories are not mutually exclusive. In
particular, no matter which definition is used, quality is always ultimately
defined oy the customer (i.e. user-based). Let's look at these categories
and see how they apply to knowledge work.

Transcendent Quality Measures

Transcendent quality measures are merely means for capturing subjective

opinions. The most common tool used is the rating scale. For example, cake
mixes are tested by submitting their products to a panel which rates the
taste of the cake on a scale from one to five, with five being the Dest
possible. Knowledge workers sometimes use peer ratings in a similar manner.
When an attribute is actually subjective, like taste, the transcendent cannot
be challenged. In areas where other measures are possible, the more
objective measures are generally preferable. Even so, the transcendent
opinion of the customer is the most important measure of one's quality.

In the author's opinion, a useful area for transcendent measures of

quality is in the appraisal of individual performance. Dr. Deming condemns
38
the use of annual appraisals for several reasons. However, appraisal systems
will probably oe with us for a while, and the use of transcendent measures
may be one way to make them work. My recommendation is to use general
categories (e.g. shows initiative) scored by the suDjective opinion of the
employee's supervisor, on tne assumption that the supervisor's transcendent
quality judgement of the employee is likely to be an accurate measure (He
will know quality work when ne sees it).

Even when using more objective quality definitions, tne transcendent can
be useful as a "sanity check". If a measured quality value "feels" too nigh
or too low, pernaps your intuition is calling for you to reevaluate your
selection of measures.

Product-based Quality Measures

Product-based quality is measured by the amount of some desired
ingredient or attribute. For example, the speed of a computer. In Knowledge
work, one desired attribute may be innovation. The difference is, of course,
that it is much easier to measure speed.

Since innovation and other intangible features are desired not for
themselves, but for their impact on the product, measurable units such as
speed will reflect the quality of knowledge work once the worK is
transitioned into hardware or software. Under such circumstances, system
parameters can be measured to establish the quality of the underlying
knowledge work. One would select the only most meaningful measures. To be
effective as quality measures, however, the measured values must be
referenced to some benchmarks. For example, the speed of a computer is
useless for quality evaluation unless the analyst knows what other machines
deliver.

A problem with attribute measures is that trade-offs may not be

recognized. Speed may be enhanced at the expense of payload which may or may
not be an improvement overall. One way to evaluate this is the use of all-
encompassing measures such as "systems effectiveness," defined as a function
of a system's availability, dependability and capability against a specified
threat. In the simplest case, availability is the probability of a system
being operable when needed, dependability the probability that it will remain
operable for the length of a mission and capability the conditional
probability that, if operating, it will successfully complete the mission.
For this simple case:

System Effectiveness = (Availability)x(Dependability)x(Capability)

An approach between the measurement of a few selected parameters and the

calculation of system effectiveness is the use of indexes. Indexes are
artificial, but supposedly not arbitrary, groupings of measures into an
overall single measure. Examples are the consumer price index and the index
of leading economic indicators. Similarly, a quality index can be created by
identifying parameters of interest, establishing measures, weighing the
measures and combining them into one. As a simple example, Robert Gunning
invented a "fog index" for evaluating understandability of text, calculated
by computing the average sentence length, adding this to the number of words
of three syllables or more in 100 words, then multiplying by 0.4. Though
39
Gunning claims nis index corresponds rougnly witn tne number of years of
schooling a person would require to read tne text 4ith reasonable ease, an
index figure is generally not meaningful in absolute terms, out, rather,
useful for showing trends.

The more tangiole the product, the oetter product-oased measures work.
However, in knowledge work the product is often intangible, sucn as a set of
recommendations, so product parameters cannot be measured. One alternative
is to use even more indirect measures so long as they also correlate with the
the attributes desired. For example, a large number of patents held snould
indicate an innovative agency. Some other measures might be the ratio of in-
house to contracted work, numbers of papers published, resources spent on
education and training activities, advanced degrees earned, name requests for
consulting committees received, and the amount of national/international
professional activity among the knowledge workers. These are measures of the
laboratory climate or environment favoring quality knowledge work.

One could also measure the climate opposing quality in knowledge work.
Common measures indirectly showing unfavorable climates include absenteeism,
turnover percentage, average sick days taken per employee, etc. Poor
environments could perhaps be more directly measured by the number of
approvals required to do work, the ratio of overhead to productive activity,
the length of time required to obtain a part or a piece of test equipment,
etc. These could be labelled "Hassle Indexes."

Manufacturing-based Quality Measures

Perhaps the best illustration of manufacturing-based quality definitions

was proposed by Philip Crosby, who equated quality to compliance with
specifications. This, of course presumes tangible products or services,
which for knowledge work could include such items as technical reports and
briefings as well as the more obvious hardware and software end products.

The most commonly used manufacturing-based quality measure is defect

rate (i.e. the percent of the product not in compliance to specifications).
Defect rate is a universal quality measure and can be applied to knowledge
work as well as manufacturing by formulating an appropriate operating
definition of defect. Cycle time is another widely used manufacturing-based
measure which is easily applied to knowledge work.

Another manufacturing-based quality measure is the variation among

products. All products will have some variation, and the greater this is,
the more defects there will be. Variance can be measured in various ways,
such as by range or by standard deviation (sigma). The lower the value of
sigma, the more uniformity in the product. Variance, however, is not the
whole story. Both the mean and variance are important. A measure which
considers both is called process capability (Cp). It compares the mean and
variance of a product parameter to specified limits.

Cp = (upper specification limit - lower specification limit)

6 sigma

40
 With a normal distribution, a Cp of 1.0 means that 99.70 of the product
would De "in spec" assuming the mean of the product is centered between tne
upper and lower control limits. To allow for means in other locations, a
Process Performance (Cpk) Index can be used.

CpK = (minimum distance between tne mean and either control limit)
3 sigma

Using either measure, the higher the value, the Detter. Motorola's "six
sigma" program strives for a Cp of 2.0 (six sigmas between the target mean
and the specification limits) which, when the true mean is 1.t sigmas off
target, translates to a defect rate of 3.4 parts per million.

For non-structured work, the main issue with manufacturing-based quality

definitions is determining what the "specification" is. A specification for
a study on Computer Technology may specify the format, perhaps even the type
style, of the final report, which are all of secondary importance to a host
of considerations sucn as responsiveness, innovation, realism, clarity, etc.
With the exception of the fog index for understandability, I have found no
specifiable measures of these critical desires. If one assumes that meeting
the specifications for a product reflects desired intangibles like
innovation, measuring conformance is adequate. Otherwise, the manufacturing-
based measures simply will not work. One could specify that a product show
innovation, but verification of compliance would require a subjective
opinion, which is a transcendent, not a manufacturing-based quality measure.

Manufacturing-based quality figures do nave an important place in

knowledge work. A laboratory's operations include many processes and sub-
processes. It is important to note that in knowledge work, as in any other,
the final customer is only the last of a series. Each office involved in a
process is the customer for some input and the provider of some output to
another customer. Thus, even the process of creating innovations will
include such processes as publishing reports, obtaining laboratory equipment,
awarding contracts, etc., which can be evaluated by manufacturing-based
quality measures. Improving these processes must improve the laboratory
operations, even if we totally ignore intangibles such as innovation. For
example, shortening the time to obtain a needed instrument yields more time
for performing experiments with it, which in turn can produce more
innovations.

Process improvement is the heart of Total Quality Management (TQM).

Improving the process can be accomplished by radical innovations or by
accumulation of many small changes. Either way, it begins with an
understanding of the process, and depends on the measurement of quality
indicators. The process itself should tell you what to measure. If the
process is proposal evaluation, for example, cycle times and/or the number of
corrections required (defects) may be compiled to establish a baseline
against which proposed improvements can be compared.

One danger in measuring a process is that what you measure becomes the
priority, and some ways of improving one parameter may deteriorate other
critical parameters. Optimizing a process may therefore adversely impact a
larger process in which it is imbedded, or the quality of the process by
other measures. For example, improvements in the cycle time for proposal
41
evaluations can be made by taking less care in doing the work, for a loss in
quality measured by the number of errors. As always, the test of value added
is the overall impact on the customer.

Value-based Quality Measures

In value-based quality definitions, cost is d consideration. A low cost
automobile which provides dependable and reasonably comfortaole
transportation would De considered a quality vehicle even if it does not have
the features of a Rolls-Royce. In fact, the Rolls-Royce may De considered
too expensive for what it provides and hence not good value for the average
consumer. Quality is also not independent of schedule. As discussed aDo.e,
cycle time is a measure of quality, but improving cycle time can adversely
affect other facets of quality such as product defect rates. Conversely, a
good product delivered too late may be of no use to tne customer.

The author's view of value-based quality is that every product, service

or process can be measured in three dimensions: cost, time, and some measure
of "goodness," such as percent defects. Improvements wnich change one
without detriment to the other two are always worthwhile. Other changes may
or may not be worthwhile depending on the overall effect on the customer.
While the trade-offs between cost, schedule and "goodness" can be a
subjective matter, all quality decisions snould try to balance the three
considerations. For example, contracting can be measured by cycle time
(schedule), overhead man-hours (cost) and number of protests per contract
(defects). Measuring only one of these invites sacrificing the others. For
ease of reference, let's call a balanced combination of cost, schedule and
"goodness" measurements a "quality troika."

Another approach to using value-based measures is to distinguish between

effectiveness and efficiency. Effectiveness measures the "goodness" of a
product or service for its user, while efficiency considers the cost of
making it happen. To illustrate the difference, consider tne example of
supplying integrated circuits meeting the customers needs Dy making much more
than ordered and screening the output. The customer may be pleased with the
product (high effectiveness), but the cost of quality will be higher than it
should be (low efficiency). Effectiveness can be measured perhaps by sales
(or the laboratory equivalent: amount of external funding), market snare, or
one of the product-based measures. Efficiency is measured by tne cost of
quality, overhead rates, or one of the manufacturing-based measures.
The Cost of quality includes the cost of preventing defects, the cost of
inspection, the cost of rework and the cost of waste. Many companies
consider only the money spent by their quality professionals (in prevention
and inspection) as the cost of quality. In reality, a typical company may be
spending 25% of its manufacturing costs on rework and scrap. One way of
measuring quality could therefore be the determination of all the measurable
components of the cost of quality. The lower the cost of quality, the nigher
the efficiency of the quality effort.

Still another approach is the use of calculations based on the Taguchi

loss functions. The computed loss can represent actual costs for repair of a
defect, or immeasurable costs such as lost business or the "loss to society"
because of poor quality.

42
 User-based Quality Measures

As previously stated, all measures of quality must ultimately oe user-

based. The problem is translating user satisfaction to an appropriate
quality measure. The most quoted user-based definition of quality is that of
J. M. Jurdn, who defined quality as fitness for use. Juran divides fitness
for use into two categories: features and freedom from deficiencies.
Features, ne stated, cost money and attract customers, while freedom from
defects saves money and keeps customers. Knowledge work featires could
include innovations, responsiveness, ease of comprenension of ideas
presented, etc. and freedom from defects includes accuracy, legioility of
written reports, etc. Under this definition, product-based quality measures
become user-based measures for evaluating features and manufacturing-based
measures become user-based measures for evaluating freedom from defects.
Transcendent and value-based quality measures may measure either features,
freedom from defects, or overall fitness for use, depending on application.

From this discussion, there seems to be a plethora of ways to measure

tne quality of knowledge work. Table 7 summarizes these.

QUALITY MEASURES FOR KNOWLEDGE WORK

PURPOSE: RATE CUSTOMER APPRAISE APPRAISE IMPROVE PRODUCTS

SATISFACTION AGENCY INDIVIDUALS AND PROCESSES

MEASURE: Rating scales Rating scales Rating scales Defect rates

Product parameters Climate indicators Cp or CpK
Performance indexes Defect rates Cycle times
Systems effectiveness Cp or Cpk Quality troikas
Defect rates Cycle times Loss functions
Cp or Cpk Cost of quality
Cycle times Overhead rates
Sales, Market share

TYPE OF Transcendent, Transcendent, Transcendent. Manufacturing-

MEASURE: Product-based, Product-based, based or
or Manufacturing- Value-based.
Manufacturing- based, or
based. Value-based.

Table 7

CONCLUSIONS

I hope the reader is convinced that TQM is worthwhile, and recognizes

that it is not easy. Success in implementing TQM requires commitment,
training, persistence, and understanding. Figure 30 summerizes the TQM
process. As a final offering, the author would like to propose his own
formulation of TQM principles, consisting of only six points. These
principles should track well with other formulations mentioned in the TQM
literature, except that I don't believe my sixth point is emphasized enough.

43
 CUSTOMER .MISSION
DESIRES CVALUES DEPASOS
LEADERSHIPENT
GOALS
.E •STRUCTURE
•EXEC. ACTIONS
TEAMS RESULTS
SRECOGNIBRON
-HASSLE

EXTERNAL INTERNAL
Figure 30

COPPOLA'S SIX POINTS

4. LEADERSHIP. MANAGEMENT MUST: - SHOW COMMITMENT

- COACH THE PLAYERS
- SUPPORT THE EFFORT

2. PROCESS IMPROVEMENT. - PREVENT DEFECTS

- BUILD TEAMS
- BREAK DOWN BARRIERS

3. EDUCATION AND TRAINING. - INSTEAD OF EXHORTATIONS

4. USE OF KNOWLEDGE AND DATA. - IF IT'S IMPORTANT, MEASURE IT

5. CUSTOMER SATISFACTION. - ALWAYS. BUILD THE BUSINESS TO DELIGHT THE
CUSTOMERS, NOT THE MANAGEMENT.

6. REBELLION. - AGAINST ANYTHING THAT DOES NOT ADD VALUE.

Let us end with Lhis:

FINAL EXAM
TQM IS:

A. A BUZZWORD

B. A WAY OF LIFE

C. A LIVING FOR CONSULTANTS

ANSWER: YOUR CHOICE

44
 MISSION

OF

ROME LABORATORY

Rome Laboratory plans and executes an interdisciplinaryprogram in re-

search, development, test, and technology transition in support of Air
Force Command, Control, Communications and Intelligence (C 31) activities
for all Air Force platforms. It also executes selected acquisition programs
in several areas of expertise. Technical and engineering support within
areas of competence is provided to ESD Program Offices (POs) and other
ESD elements to perform effective acquisition of C 31 systems. In addition,
Rome Laboratory's technology supports other AFSC Product Divisions, the
Air Force user community, and other DOD and non-DOD agencies. Rome
Laboratory maintains technical competence and research programs in areas
including, but not limited to, communications, command and control, battle
management, intelligence information processing, computational sciences
and software producibility, wide area surveillance/sensors, signal proces-
sing, solid state sciences, photonics, electromagnetic technology, super-
conductivity, and electronic reliability/maintainabilityand testability.