MEASUREMENT,
THE BRIDGE BETWEEN
THE HARD AND
SOFT SIDES
Arthur M. Schneiderman
V isit Milliken and Company, a
1989 winner of the Malcolm
vice president of quality and pro-
ductivity improvement at Analog
Get the hard stuff right and much
of the soft staff will take care of
Baldrige National Quality Award, Devices, Inc.,2 I had also worked itself.
and one of the things you will extensively in the development and
notice is the slogans on the walls. deployment of nonfinancial per-
One in particular is eye-catching: formance metrics3 and the use of the In this article, several examples will
first balanced scorecard for their be used on how quantitative mea-
“The hard stuff is easy, the soft stuff management.4 My view was, and to sures stimulate positive behavioral
is hard and the soft stuff is more a great extent still is: changes in both individuals and
important than the hard stuff.”
organizations. However, before
proceeding, some definitions are in
To an admitted hard-sider, this one EXECUTIVE SUMMARY
order.
seemed dead wrong.
Both process and organizational
Several years earlier, I had devel-
learning can be measured by looking DEFINITIONS
at the improvement half-life and its
oped the half-life metric as a driver Generally, when people talk about
rate of change.
of the improvement process.1 As the hard stuff, they are in the
These metrics can be used to draw process view. Included here are
ARTHUR M. SCHNEIDERMAN is organizational and individual attention
process mapping, measures and
an independent consultant on process to the need for better approaches to
management. From 1986 to 1993 he metrics, improvement methodolo-
process improvement.
was vice president of quality and gies, and tools.
productivity improvement at Analog Most information at the boundary In contrast, the soft stuff reflects
Devices, Inc. where he facilitated the between hard and soft sides of the people view and is more
Quality Steering Council and chaired improvement is in the form of
the Total Quality Management concerned with motivation, reward
language rather than numerical data. and recognition, team dynamics,
Implementation Council. Schneiderman
has served on the Malcolm Baldrige Management, logic, and analytic and vision and mission..
National Quality Award as senior
practices can expand to provide tools Although these definitions paint
examiner and the Conference Board’s
US Quality Council II. He is located in for understanding this alpha-numeric the two views as black and white,
Boxford, Massachusetts and can be interface in terms of degrees of they are simply different aspects of
reached at (978) 352-2031 or at freedom rather than simple true or a total system for improvement.
art@schneiderman.com. false rigidities.
JOURNAL OF STRATEGIC PERFORMANCE MEASUREMENT
APRIL/MAY 1998
14
M E A S U R E M E N T , T H E B R I D G E B E T W E E N T H E H A R D A N D S O F T S I D E S
life? Research showed that it
MEASUREMENT PROCESS AND depended on process complexity,
ORGANIZATIONAL For any given process the which in turn had two dimensions:
LEARNING
half-life is constant over a organizational
technical
If the process view of an organiza- significant range or
tion is taken, it is as the sum of a
number of interconnected processes.
improvement. This dependency was captured in a
Each of these processes has cus-
three-by-three matrix that had
tomers and suppliers. The half-life values ranging from one
customers have requirements, and month for low organizational and
organizations strive to match the
technical complexity processes to
K is the change in knowledge, t twenty-two months for high ones. 6
process output to these
is the time interval, L is the rate of Associated with any given improve-
requirements. Results metrics
learning, and F is the rate of ment process is its own half-life matrix.
measure how well the organization
forgetting. L-F can be called the net Pre-TQM, industrial engineers were
perform this.
rate of learning, which we always relied upon to design and improve our
Internal to the process are a few
hope is positive. processes. The workers job was to fol-
key variables that drive its output.
As process knowledge increases, low the standard operating procedures
In general, they are invisible to the
the gap between the processes‟ (SOP) developed by these engineers.
processes‟ customers. Measurement
inherent capabilities and actual TQM empowered the process execu-
of these variables produces a set of
outputs will decrease. Inherent tors not only to follow the procedures,
process metrics. In mathematical capabilities represent the ultimate but also to spend a portion of their time
terms, results metrics are the depen- limit the process can achieve with- improving them. This change in
dent variables and process metrics out significant capital investment. improvement process has led to a ten-
are the independent variables asso- Consequently, any results metric fold reduction in improvement half-
ciated with a process. should also decrease, defined as a lives and commensurate increase in the
Process knowledge is the set of measure of that gap. Therefore, the net rate of process learning.
tangible and intangible operating rate of change of a results metric is Transforming intangible into
procedures used to convert the a function of net process learning. tangible knowledge will greatly
inputs to outputs. If knowledge is Conversely, the rate of change of a
reduce process forgetting while net-
constant (i.e. operating procedures results metric is itself a measure of
works and best practices databases
unchanging), and the process vari- net process learning.
will accelerate problem solving.
ables are stable, then the output of A little over a decade ago, I dis-
There does not appear to be an
the process will also be stable, sub- covered that the use of total quality
accepted definition of organiza-
ject only to statistically predictable management (TQM) tools and
tional learning. The following is
variability. This phenomenon is the methods produced a constant rate of
offered:
purview of statistical process improvement which I called the
control (SPC). improvement half-life.5 For a typical
Organizational learning is the
Process improvement requires process, the half-life for
improvement was nine months. This acquisition, application, and mas-
new knowledge that manifests itself
meant that the gap between current tery, of new tools and methods
in changed operating procedures.
performance and the ultimate that allow more rapid improve-
This change in knowledge is called
process capability could be cut in ment of those processes whose
learning. Note also that process
half in nine months. The remaining improvement is critical to the
knowledge can be lost. For
gap could be cut in half in an success of the organization.
example, undocumented procedures
that rest in volatile human memory additional nine months. Because the
can be lost when the process half-life is a measure of rate of
In other words, the half-life is a
executors move on to other improvement, it is also a metric of
results metric of the improvement
assignments. This loss of net process learning.
process. As the improvement process
knowledge is called forgetting. This For any given process the half- itself is improved, we expect to see
can be expressed mathematically as: life is constant over a significant half-lives decrease for a given level
ΔK range of improvement. What then of process complexity. The rate of
LF determines the value of the half-
Δt change of process half-life, at con-
JOURNAL OF STRATEGIC PERFORMANCE MEASUREMENT
APRIL/MAY 1998
15
M E A S U R E M E N T , T H E B R I D G E B E T W E E N T H E H A R D A N D S O F T S I D E S
stant process complexity is a metric bers on a half- or full-time basis.
of organizational (versus process) People’s time is not always This can significantly accelerate the
learning. Mathematically, it can be the bottleneck in the revolutions per month (RPM) of the
shown as: improvement process. improvement cycle and lead to
Often data collection is effectively shorter half-lives. Note
Lp t½ limited by the cycle time of that people‟s time is not always the
t the process itself.
Lo ½
bottleneck in the improvement
process. Often data collection is
t complexity
limited by the cycle time of the
Lp is the process learning metric, process itself.
Lo is the organizational learning The following section examines a
metric, and t½ is the process real example in which a sustained
development of new ones are the
improvement half-life. Lo is in fact gap between actual and expected
keys to organizational learning.
the half-life for the improvement half-life, combined with a bit of
Why does the half-life model
process itself. work? Most applications of the interdivisional competition, was the
Consider for example, a real TQM improvement methodology key in turning a TQM skeptic into a
situation7 where an organization, rely on taking a large problem and dedicated champion.
new to TQM, is working on breaking it up into smaller pieces.
reducing late shipments in their Typically, the biggest piece will
order fulfillment process. The represent 30 to 50 percent of the AN EXAMPLE
complexity of their process is such entire problem, Once identified, the In 1987, Analog Devices estab-
that we would expect a twelve- improvement team determines a lished and openly published a set of
month half-life. corrective action, implements it, and nonfinancial performance goals as
However, being new to the TQM modifies the operating procedures, part of its five-year strategic plan.9
to reflect the improved process. The half-life method, customer
improvement methodology, their
It then turns to the next biggest input, and competitor analysis were
initial half-life is forty months. As
piece, which will again represent
they master this methodology, the used to set these goals. The half-life
30-50 percent of the remaining
half-life drops continuously and was also used to set intermediate
problem. This never-ending cycle
approaches the twelve-month level. quarterly goals and to deploy these
constantly works to close the gap
The half-life of their half-life curve goals to each of our divisions as part
between current performance and
is nine months. Their Lo is nine of our annual planning process.
ultimate process capability. If the
Although a one-page summary that
months and their eventual Lp will be process is capable of satisfying
customer requirements, then that combined these with the key
twelve months. Once an organiza-
gap will also be narrowing. If it is financial goals was initially referred
tion has mastered a given
not, then a new process must be to as the “Quarterly Performance
improvement methodology, all new
designed which has the capability of Audit, “ it quickly became known as
improvement efforts are expected to
meeting these requirements. the “Scorecard.”
start at close to the appropriate
As a rule of thumb, a team will At that time Analog had a highly
half-life for that process. This decentralized organization and each
take three to five months to
model suggests that Lo will vary division was left to determine on
complete each improvement cycle.
significantly over time, as new Therefore, a roughly 40 percent their own how to achieve their
paradigms are invented, embraced, improvement in four months scorecard goals. At most divisions,
and mastered, by the organization. translates into a half-life of about a staff member was assigned
The toolsets that enable process five months. A more complex responsibility for one of the metrics
learning vary with process process has more and therefore on the scorecard.
complexity.8 For example, the 7-QC smaller pieces and takes more time At our largest division, Joe, the
tools, the 7-MP tools, and per cycle, hence a longer half-life. operations manager, was assigned
simulation modeling, span the range Improvement activities usually responsibility for the on-time
from low to high complexity. account for two to four hours per delivery metric.
Mastery of existing toolsets and the week of team activity. To accelerate
the rate, Carrier Corporation‟s
“Turbo TQM,” assigns team mem-
JOURNAL OF STRATEGIC PERFORMANCE MEASUREMENT
APRIL/MAY 1998
16
M E A S U R E M E N T , T H E B R I D G E B E T W E E N T H E H A R D A N D S O F T S I D E S
As my office was located at that
facility, and my boss (the COO) EXHIBIT 1
Late Shipments by Division*
was acting general manager, I was
asked to sit in on all of the staff
meetings. Joe was a very proud and, DIV A DIV B DIV C DIV D DIV E Corporate
Monthly Perc ent of Lines Shipped Late
100
to some extent, autocratic manager.
His initial reaction to a very
aggressive improvement goal was
“if that level of performance could
be achieved, we would already be
there ... it just can‟t be done.” Over 10
time, the division‟s performance did
not improve. My help was declined.
Pressure mounted and everyone was
frustrated, from the CEO down.
One of the displays that I used to 1
track performance against plan was 14 14 N/S N/S N/A 34
the chart shown in Exhibit 1.10 I Half-life (months)
*See reference 3
posted this chart on the door to my
office. If a division had elected to Joe held in his staff meeting able, and suggested alternative
use TQM to improve their delivery, conference room, Joe took his solutions. Often in putting the case
I also posted the starting date and a normal seat at the head of the table. together, the team would see the
Pareto Chart (rank ordered bar Needless to say the team meeting situation more clearly and reach
chart) of causes of lateness as took on the tone of a staff meeting. consensus on its own.
identified by their QIP (Quality Several of the participants said Analog Devices practiced a
Improvement Process) team. After a nothing during the whole meeting. flavor of a seven-step improvement
few months, it became apparent that With a little coaching, the next method used widely in Japan.11 The
those divisions that were using QIP meeting was in a different room and last step, step seven, is called
teams were improving while the Joe made sure that he sat at the “reflection.” Here the team reflects
others were not. People would come middle of the table. Several months on its team experiences, prepares for
by my office and look at the charts its management presentation
later, Joe turned over the team
on a regular basis. (reward and recognition), and plans
leader position to his manager of
for is next improvement cycle. In
One night, Joe passed my office production planning and changed
reflecting, the team addresses the
on his way out, returned, and came his role to team sponsor.
questions, “What went well?” and
in. He looked at the charts and then There were many occasions in
“What were our major weak-
turned to me and said, “Art, I think which finger pointing was
nesses?”
you‟re calling me a bad manager.” redirected to data collection and Typical answers to the first
Two hours later Joe agreed to start a analysis. As the team became more question might be, “Everyone
QIP team and I agreed to be its familiar with the TQM tools and participated,” “Everyone met his or
facilitator. Confronted with data methodologies, behaviors changed. her commitments,” and “Larry
that showed other similar divisions Knowing that hip-shooting would pitched in to help Sue when she was
were achieving results using TQM, no longer work, team members out sick.” Identified weaknesses
Joe‟s skepticism softened. brought data to meetings in support might include, “We spent too long
As team facilitator, I did have to of their positions. collecting data,” “We didn‟t always
overcome many soft issues. For Functional issues were resolved end our meetings with clear action
example, the team consisted of Joe using an “elevation process.” If the items,” “I wish we had someone
(as the team leader, of course) and team could not resolve the issue on from accounting on the team.”
included several of his direct its own, a case was put together that Too often, these types of
reports. At our first meeting, which described the issue, provided evaluations are done in an open loop
whatever relevant data was avail- fashion. Individual performance
JOURNAL OF STRATEGIC PERFORMANCE MEASUREMENT
APRIL/MAY 1998
17
M E A S U R E M E N T , T H E B R I D G E B E T W E E N T H E H A R D A N D S O F T S I D E S
appraisals rarely start with the pre-
Sean: “OK, Joe, how are you
vious cycle‟s findings and next
doing it. Stretching the lead
steps. Our teams were encouraged, It is the hard stuff that times? Building inventory?”
as part of their training, to review solves problems; the soft
their previous reflections at both the stuff is an important Joe: “No, no ... This QIP stuff
really works.”
start and end of each improvement enabler.
cycle. By reviewing them at the
beginning, they could plan steps to The following week, Sean called to
address their identified weaknesses; ask if I could drop by on my next
at the end, they could see how well visit. By the way, they are
they did. In this way, the reflection (CEO) and Tom Malone (president) represented in Exhibit 1 as divisions
step became closed-loop. This attend every rally, and give out the A and B.
closure greatly enhanced the teams A few years later, management
prizes at their conclusion.
learning. decided, for very complex reasons,
While the QC Circle conferences
to set a flat annual goal. Joe came to
This self-diagnosis is included in have perhaps an 80/20 focus on my office for an explanation. He
the teams‟ presentations to learning versus celebrating, the said, “Art, with no improvement
management, which addresses each sharing rallies have the converse. goal, I don‟t have any job.” The
of the seven steps. The collective Which is right? That depends on the greatest champions are converted
experience of the QIP teams current needs of the organization, a skeptics, and in this case, it all
provides the fact base that very important soft-side decision. started with data.
management can use, for example, Here again, data from the team
in setting priorities in TQM training reflections and employee surveys FUZZY LOGIC
(e.g. expanding training in data can provide the basis for the Success is like a giant magnet. It
collection and analysis or adding a decision. draws people from all corners of the
module on meeting management). organization, some of whom have
made little or no contribution to that
Analog Devices held periodic Solution Versus Enabler
corporate wide and divisional “QIP success. At the first sign of trouble
This example shows that the soft they are the ones who declare that
Fests” where teams joined together stuff can emerge as countermea- they have the solution. This
to celebrate their success and share sures to real problems teams happened at Analog when several
their experiences. In designing these experience in doing the hard stuff. It unrelated situations in 1989
conferences, two different models is the hard stuff that solves triggered a reversal in performance
were examined. problems; the soft stuff is an at nearly all divisions. Delivery
important enabler. Too often, the started to get worse.
QC Circles and Sharing soft stuff is approached as an end in A newly hired corporate support
Rallies itself. Consequently, it does not staff manager took the position that
In Japan, there are almost daily QC contribute to accelerated learning our setback was the result of a lack
Circle conferences held in various of company vision and mission
and improved organizational
statements, too many teams, and the
parts of the country. Teams give performance.
need for facilitator and teamwork
very structured and standardized As an epilogue to this example,
training. When asked for data to
fifteen-minute Presentations, the Joe‟s division became the most support her position, she answered
main objective of which is to share rapidly improving division at “I don‟t need data, this is my area of
best practices in the use or the Analog. In fact, they were right on expertise.”
seven-step improvement methodol- their annual half-life based goal Since Analog was still in the
ogy. line. Standing at the bar at one of early stages of its quality journey,
At Milliken & Company, they the company‟s manufacturing confusion over what to do next ran
have been holding quarterly council meetings, I overheard the rampant. In my view, following her
“sharing rallies” for fifteen years. following conversation between Joe recommendations would violate the
Teams are allowed five minutes for and Sean, his counterpart at a sister very principles that the company
their presentations and the division that had made a quick start was trying to inculcate in the
participants get to vote on the but had plateaued for the past organization; Analog would be
“best” presentation. Roger Milliken several months: leaping from problem to solution
without collecting data and
JOURNAL OF STRATEGIC PERFORMANCE MEASUREMENT
APRIL/MAY 1998
18
M E A S U R E M E N T , T H E B R I D G E B E T W E E N T H E H A R D A N D S O F T S I D E S
EXHIBIT 2
identifying the root cause. To act on
Example of Survey Res ults
her advice would be setting the
worst possible example.
Progress to Date
On the other hand, I did not
unweighted
know the answer either. It was only weighted
several years later, when Sterman,
Reppening, and Kofman built a
System Dynamics based computer
Division
simulation model of the Analog
Devices experience, that I was able
to see through the complexity of the
situation and understand the real
Corporat e Wide
root causes.12 (Their findings are
discussed later in this article.)
What I chose to do was to go to
those people in the organization
who should have the best
knowledge of the situation: the 0 1 2 3 4 5
division general managers. wors e neutral better
I constructed a survey that had a What have we done well? What have we done poorly?
mix of questions some of which “…selected one thing to focus on: “…have not culturally ingrained the
could be answered on a quantitative on-time delivery, and achieved a QIP philosophy”…“we haven’t built
meaningful win…” excitement among the GM’s”…“we
scale (0 to 5) and others in words. don’t have enough QIP leaders in
Questions included the following: the division staff…”
“…identified clear priorities, established “…divisions really don’t understand
How do you rate overall QIP the metrics, and kept them very visible
… (this) has led to GM buy-in to QIP…”
how customer’s needs have
changed…”
progress to date?
“…very rapid and broad acceptance
What have we done well? of QIP…” “…failed to define QIP. We think QIP is
QIP teams, it’s a subset of TQC, but
What have we done poorly?
you’re now telling me that it’s the same
as TQC. That was never explained to us…”
What problems should we
anticipate? Detailed notes were taken and all context, but as language data, could
quotes were confirmed. not be analyzed ... or could they?
How do you rate The results were presented in a In 1965, Lofti Zadeh “invented”
understanding of QIP at your written report to the corporatewide a new approach for analyzing complex
division? Company wide? QIP Council. Exhibit 2 is a typical systems called “fuzzy logic.” The science
How valuable have the excerpt taken from that report. The
of fuzzy logic legitimizes the modeling of
systems where truth takes on degrees and
following been in driving QIP? chart shows the GM‟s perception of information can be in linguistic form.
the total company as well as their Although Zadeh is credited with the
Half-life concept
division‟s progress to date. Also invention of fuzzy logic, many
Scorecard
shown are the division applications predate his discovery.
General Manager Council
revenue-weighted average and the System Dynamics, the basis of systems
meeting updates thinking, is a form of fuzzy logic that had
range of responses (double-ended
Bonus adder (based on nonfi- its origins in the work of Jay Forester in
arrows). The quotes were used to
nancial metrics)
add further depth of meaning to the the 1950‟s.13 Also, in that period, Jiro
Others Kawakita invented the K-J method for the
data. The results of this traditional
analysis of language data.14
part of the survey proved interesting
Each survey was conducted on a A short while after the survey
but lacked sufficient detail to was completed, I had the
one-on-one basis. They ranged from
produce actionable next steps and opportunity to extensively apply and
forty-five (45) to ninety (90)
priorities. The problem was that the teach the method under the tutelage
minutes. The GMs were offered
numerical data did not dig deeply of Professor Shoji Shiba. 15 Using
anonymity, but nearly all declined. enough, while the quotes provided the K-J Method with the language
JOURNAL OF STRATEGIC PERFORMANCE MEASUREMENT
APRIL/MAY 1998
19
EXHIBIT 3
A Re sult From the General Managers Survey
M E A S U R E M E N T , T H E B R I D G E B E T W E E N T H E H A R D A N D S O F T S I D E S
QIP Improvement Opportunities
Performance noncritical and are candidates for
poor ok excellent reevaluation of allocated resources.
vital The most important area is that
7 above the maintain band. These
9 Maintain
represent areas viewed by the GMs
Criticality to QIP Success
Improve as critical to success but performing
8 below what they believed was
required. Using distance from the
2 band as a numerical measure of
important
3 need for improvement, the
1 following list can be created:
Re-evaluate
Maintain Area Points
unweighted Corporate QIP
irrelevant weighted
training 60
1. CEO's Staff 4. * 8. Corp QIP function The General Managers 25
2. Corp QIP Council 5. * 9. Corp QIP training
3. Div QIP Council 6. * The Corporate QIP
7. GM's function 15
Total 100
data contained in the quotations in the interview: QIP improvement
Exhibit 2, several clear hidden opportunities, shown in Exhibit 3.
messages emerge. Although we had This matrix provided a format for
done a good job in our tactics evaluating the importance versus This list represents the fuzzy logic
(goals, metrics, focus on the vital performance for several groups and consensus view of the GMs as to
few), we had failed to create a sense individuals (items 4, 5, and 6 with where improvements were most
of urgency based on a compelling names removed to protect them) needed in the QIP infrastructure.
business need for change and had thought to be critical to the success The point scores provide an
not clearly linked our metrics to this of Analog‟s QIP efforts. indication of the emphasis (i.e.,
need. Furthermore, our The intent was to identify resources) that should be directed
self-diagnosis process had failed to responsibility for determining toward each area. As far as next
uncover several weaknesses in our corrective actions for our current steps, the bottom line was clear: we
organizational change system, situation. It was not for assessing desperately needed training in tools
including training and education blame. Each GM was asked to place and methods. There was little
and company-wide diffusion. the identification number at the mention of training needs on the
Ideally, the corporate QIP appropriate location on the matrix. soft stuff. The GM‟s also
council would have collected and Exhibit 3 is the result of recognized that they were not doing
analyzed language data around the averaging these locations for each all that was necessary and that the
issue or theme, “What are the of the identification numbers. The Corporate QIP function was not
weaknesses in our implementation solid points represent a straight providing them with enough
of TQM?” With the importance average, while the open points support.
weighted list of answers, we could represent the division revenue The report to the Corporate QIP
have proceeded to identify and rank weighted average. Points lying Council omitted the results for the
appropriate corrective actions and within the band labeled maintain” three individuals (items 4, 5, and 6).
been better able to communicate reflect performance commensurate These data were communicated only
their rational. with their criticality to the QIP to them. It proved extremely
As another example of the use of process. Points below this band valuable in highlighting issues
fuzzy logic, consider the last item in would be indicative of high
performance on something deemed
regarding their roles in advancing Should sensitive surveys like this team tool and yields a consensus
us on our TQM journey, and made be done internally or by outside rather than individual view. There is
visible conflicts of these roles with consultants? They should be done a role for an outside facilitator in
their own personal management by an internal team (in Analog‟s guiding the process, but ownership
styles and agendas. case, the Corporate QIP Council). of the conclusions must lie with the
The K-J Method is an excellent team driving the change.
JOURNAL OF STRATEGIC PERFORMANCE MEASUREMENT
APRIL/MAY 1998
20
M E A S U R E M E N T , T H E B R I D G E B E T W E E N T H E H A R D A N D S O F T S I D E S
measurement of the first, using the
The Trap half-life method can encourage (ed.) (Boston: Harvard Business School Press,
1994): 179-182.
Returning to the research done by changes in individual behavior. Can 5 Schneiderman, "Setting Quality Goals."
Sterman, Repenning, and Kofman, traditional concepts be used to 6 Arthur M. Schneiderman, "Are there Limits to
Total Quality Management?" Strategy &
their work demonstrated that the prove that the half-life metric is Business (Issue 11, Second Quarter 1998): 35-
longer improvement half-lives for directly related to learning? No, it 45.
7 Schneiderman, "Metrics for the Order Ful-
the more complex revenue- must be justified using fuzzy logic. fillment Process:' (Part 2, Exhibit 3).
generating processes (e.g. product The second deals with a complex 8 Schneiderman, "Are there Limits to Total
situation in which our traditional Quality Management?"
development and sales), coupled 9 Ray Stata, "Organizational Learning-The Key
with severe downturn in the recourse is to opinion rather than to Management Innovation," Sloan
fact. In these situations, data can be Management Review (Spring 1989 63-74).
semiconductor business cycle, 10 Schneiderman, "Metrics for the Order
meant that productivity increased collected and analyzed, but because Fulfillment Process:' (Part 2, Exhibit 3).
most of it is in the form of 11 Hitoshi Kume, Statistical Methods for
much faster than volume. Quality Improvement (The Association for
With immense pressure created language, fuzzy analysis is required. Overseas Technical Scholarship (AOTS)
by a dangerously low stock price Fortunately, tools exist for this 1985).
12 Sterman, J., N. Repenning, and F. Kofman
(felt throughout the industry) and analysis, although they are not "Unanticipated Side Effects of Successful
the resulting high labor variances widely understood. These tools Quality Programs: Exploring a Paradox of
Organizational Improvement," Management
caused by successful productivity include the seven management and Science.
improvement, Analog succumbed planning tools17 and simulation 13 Jay W. Forrester, "Industrial Dynamics: A
Major Breakthrough for Decision Makers,"
and had its first ever company-wide modeling. Harvard Business Review (July-August,
layoff. This triggered the age-old Finally, my last visit to Milliken 1958).
& Company was as a guest at an in- 14 Jiro Kawakita, The Original KJ Method
fear that improvement leads to job (Kawakita Institute, Tokyo).
loss. We fell right into the trap that house training course. I noticed that 15 Shoji Shiba, Alan Graham, and David
there were a lot more courses and a Walden, A New American TQM (Cambridge
Deming foresaw and created fear in MA: Productivity Press, 1993).
lot more talk about the hard stuff. 16 W. Edwards Deming, Out of Crisis (MIT
the workplace and a subsequent
At dinner, I asked Tom Malone Center for Advanced Engineering Study,
suspension of effective improve- 1982).
(Milliken‟s president) whether they
ment activities.16 17 See for example, Michael Brassard, The
What should we have done? The were becoming hard-siders. His Memory jogger Plus+ (Methuen, MA:
GOAL/QPC, 1989).
model suggests that a greater focus answer:
on improvement of revenue
generating processes could have “Well, „the hard stuff is easy, the
grown volume to fully utilize a soft stuff is hard,‟ but the hard
more productive workforce. It also stuff‟s important, too.” :
Reprinted from
Journal of Strategic Performance Measurement, April/May 1998,
suggested that Analog might have Volume 2, Number 2
been too aggressive in cutting Notes
prices, which exacerbated the
situation. Their findings are not
without controversy within Analog. 1 Arthur M. Schneiderman, "Setting Quality
Goals," Quality Progress (April 1988): 51-57.
2 Analog Devices is a mid-sized semiconductor
CONCLUSION manufacturer.
3 Arthur M. Schneiderman, "Metrics for the
This article has addressed two Order Fulfillment Process," Journal of Cost
related issues that reside in the Management (Part 1: Summer 1996): 30-42,
(Part 2: Fall 1996): 6-17.
space we have created for ourselves 4 Robert S. Kaplan, Analog Devices: The Half-
between the hard stuff and the soft Life System (Boston, MA: Harvard Business
School, 1989) Case #9-190-061; and Robert
stuff. The first deals with the S. Kaplan, "Companies as Laboratories' " The
measurement of process and Relevance of a Decade, Paula Barker Duffy
organizational learning and how
JOURNAL OF STRATEGIC PERFORMANCE MEASUREMENT
APRIL/MAY 1998
21