You are on page 1of 6

DESIGN FOR SIX SIGMA

Straight Talk on DFSS

ew product development is the lifeblood of most organizations.


INTEGRATE
TOOLS, METHODS, N During the last decade, people working in product development
have experienced incredible changes in the expectations of their
design process.
PROCESS AND An example from the computer disk drive industry is typical of what many
experienced during product development (see A Matter of Survival).
TEAM MEMBERS AT Other industries are faced with similar competitive challenges when
attempting to develop and produce products that are competitive in both
THE FRONT END.
price and technology and truly meet customer needs.
If all industries are considered, only about 60% of new products launched
are a success. Additionally, about 45% of all resources allocated to new prod-
uct development and commercialization are spent on products that are
killed or fail to provide adequate financial return.1
Reasons for new product failures can be grouped into three primary cate-
gories:2
Market research mistakes.
Problems in design and production.
By Charles Huber, Bad timing.

Corporate DFSS Understanding Your Customers


Leader, Seagate, One big error organizations make is thinking, We know what is best for
and Robert our customers, and they will want what we develop. This assumption is fine
if you are on target. But a shift in what customers want can signal death to
Launsby, organizations relying on this market approach.
Other organizations try to understand customers but only make a feeble
President, Launsby attempt or subcontract out market research to groups who far too often do
not really understand the producer or the consumer. Constantly evolving or
Consulting newly emerging requirements late in the design cycle are also huge hin-
drances to developing winning new products.
This is very much at the heart of new product problems frequently expe-
rienced by U.S. manufacturers of automobiles. Until leading automotive
companies learn to proactively manage requirements, they will never be able
to match the development cycles of Toyota.3
Design and production problems can be further broken into the following
categories:
Sheltered product issues.
Manufacturing capability issues.
Attempts to develop the perfect product.
A sheltered product works well only in labs for demonstrations to man-
agement or lead customers. Many products can be made to look good under
limited operating ranges or only operate if constantly tweaked and tuned.

20 I A U G U S T 2 0 0 2 I W W W . A S Q . O R G
S t r a i g h t Ta l k o n D F S S

Too often products are totally


designed before thought is
given to sourcing or manufac-
turing. This frequently leads to
the selection of volume process-
A Matter of Survival
es that do not have the inherent
In 1990, Digital Equipment Corp. was about to launch a new generation
capability to produce the part as
of computer disk drives into the marketplace. Code named the RA-90, it
designed.
was the second largest development project ever undertaken by the com-
In the military missile busi-
ness, it is said some engineers pany.
would literally hang on the side Several major technological innovations were to be simultaneously inte-
of the missile as it is being grated into this state of the art (at the time) product. Key metrics associ-
launched, making last minute ated with the product were:
changes if allowed to. This need A product design cycle of five years (initiated in 1985).
is usually the result of a late Price of $15,000 at the original equipment manufacturer (OEM) level.
understanding of product re- Storage capacity of one gigabyte of information (formatted).
quirements or attempts to make Mean time before failure of 40,000 hours.
the product perfect. Compared with todays product, this seems like ancient technology. For
Initiatives as large as design
example, key metrics of todays disk drives are:
for Six Sigma (DFSS) will suc-
Product design cycles of six months.
ceed only when accompanied by
Price at the OEM level of less than $100.
fundamental change in a several
design related activities. Storage capacity of 60 or more gigabytes of information.
Mean time before failure greater than 500,000 hours.
What DFSS Is Unfortunately, because of product design glitches, the RA-90 launch was
very late in coming to market. By the time enough glitches had been
DFSS is a business process resolved to allow limited shipments, competitors had released enhanced
focused on improving profitabil- technology drives at much lower prices. What could have been a huge win
ity. Properly applied, it generates for this organization became a great failure. Coupled with other new prod-
the right product at the right uct failures, this led to the ultimate breakup of the company.
time at the right cost. Through
its use of product and team
scorecards, it is a powerful pro-
gram management technique.
DFSS is an enhancement to and fixes problems, while DFSS focuses on the
your new product development process, not a replace- design of the product and process.
ment for it. A documented, well-understood and use-
DMAIC is more reactive, while DFSS is proactive.
ful new product development process is fundamental
to a successful DFSS program. Dollar benefits obtained from DMAIC can be
Your new product development process provides quantified rather quickly, while the benefits from
the roadmap to success. DFSS provides tools and DFSS are more difficult to quantify and tend to be
teamwork to get the job done in an efficient and effec- much more long-term. It can take six to 12
tive manner. By rigorously applying the tools of DFSS months after the launch of the new product
you can be assured of predictable product quality. before you will obtain proper accounting on the
People frequently talk about DFSS as the logical impact of the DFSS initiative.
extension of Six Sigma at the manufacturing and serv-
ice level, called DMAIC, an acronym for define, meas- Roots of DFSS
ure, analyze, improve and control. This may be true,
but it is important to realize the initiatives are tremen- DFSS has its roots in systems engineering. Much of
dously different: the learning that underpins systems engineering
The DMAIC process looks at existing processes evolved under the guidance of the Department of

S I X S I G M A F O R U M M A G A Z I N E I A U G U S T 2 0 0 2 I 21
S t r a i g h t Ta l k o n D F F S

Defense and NASA. They developed a management an assessment of customer needs, functional analysis,
approach that uses performance specifications, as identification of critical to quality characteristics, con-
opposed to volumes of product, subsystem, assembly, cept selection, detailed design of products and
part and process specifications, to control the lifecycle processes, and control plans.
process. The beginning of the process centers on discovery
In the systems engineering world, management of of customer wants and needs using tools such as con-
requirements (for example, those aspects of the end cept engineering and quality function deployment
product that must meet customer expectations) (QFD). From this fuzzy front end, requirements
guides and drives the entire process. take shape. Customer issues, competitive advances,
Requirements at the senior or point of use level can technology roadmaps and disruptive influences co-
then evolve through use of a variety of techniques mingle in a stew of uncertainty.
generally described under the heading of require- Traditionally, such a mixture of needs and concep-
ments flow-down. tualizations could lead to doubt, creeping perform-
When statistical or quantitative methods are used to ance expectations and big problems with business
establish requirements between system performance results. But DFSS brings structure to the front end.
and underlying inputs, the design process methodolo- No guarantees about market success are implied by
gy transitions from a reactive, build and test mode to the preceding statement, but a systematic and repeat-
a predictive, balanced and optimized progression. able process is always better than notes on the backs
DFSS provides a systematic integration of tools, of napkins or partially understood product expecta-
methods, processes and team members throughout tions.
product and process design. Initiatives vary dramati- Giving focus to the translation of customer wants
cally from company to company but typically start with and needs into language that can drive the design
a charter (linked to the organizations strategic plan), process is essential.
When personal computer users indicate
speed is important, how is this simple idea
translated into actual performance meas-
ures? Is speed the rpm of the drive spin-
A GEMS Experience dle? Or does speed mean reduced time to
accept new data into a file? Or time to find
General Electrics Medical Systems Division (GEMS) led the a desired segment of data? Or, does speed
way for DFSS application in the late 1990s with the introduc- mean the rate at which an amateur pho-
tion of the Lightspeed Computed Tomography System (CT). It tographer can compose, edit, store and
was the first GE product to be completely designed and devel- send the kids photos off to Grandma?
oped using DFSS.
The answer is yes, but not always. DFSS
methods highlight and resolve such
Lightspeed revolutionized the capabilities of CT. It allowed
dilemmas before design teams start
doctors to capture multiple images of a patients anatomy
detailed development. It eliminates ambi-
simultaneously at a speed six times faster than traditional guity at the front end that can lead to
scanners. Doctors were able to scan more patients per shift chaos at the back.
and saw productivity double. Additionally, the images were Lets suppose the front end is clean. You
more clear. The increased speed and image quality enabled have resolved product performance
doctors to more accurately treat patients while making deci- issues, and a set of quantifiable and meas-
sions with greater confidence. urable specifications have materialized
In 1999 Jack Welch, then chair and CEO of GE, announced through agreement among marketing
all GE products designed after that time would be designed and development team members.
using the DFSS approach. Data storage and management
companies Iomega and Seagate and other organizations also
Predicting Design Behavior
have corporate level initiatives focused on DFSS.
DFSS now looks to predict how the
designs under consideration will behave.

22 I A U G U S T 2 0 0 2 I W W W . A S Q . O R G
S t r a i g h t Ta l k o n D F S S

Engineering and statistical methods provide the basis 2. Learn from the present to become stronger in the
for prediction. QFD is a terrific tool and should be future.
used in early phases of requirement flow-down. To reap the benefits of DFSS, an organization must
Obtaining the second or even third level in the house make fundamental changes in the way it develops new
of quality matrix will always enhance engineers products and processes. Ironically, while tremendous
understanding of related factors. changes have taken place in common methodologies
However, and its a big however, engineers cant use applied to improving manufacturing efficiencies (such
QFD to predict what might happen. Nor can QFD find as just in time, lean, statistical process control, design for
out if factors are interdependent. In this phase of manufacturing and assembly (DFMA), many organiza-
DFSS, engineers need to define relationships between tions still use 20-year-old fundamentals when scoping
desired response and critical dependent factors via market requirements and designing a product and its
transfer functions. processes.
Transfer functions, models, simulations and funda-
mental physics can all play a role. Transfer functions
Changes Required
can then enable engineers to inject variation into the
models to understand how the distribution of varia-
For DFSS to be effectively implemented across
tion can alter the desired performance. This allows
development, an organization must make the follow-
you to predict what will happen in actual operation.
ing six major changes:
Predictions bring another critical element into the
DFSS methodology: process, part and measurement Freeze requirements.
variation. Pay dirt from deploying DFSS stems from Allocate additional resources early.
analysis of the effect of variation before manufactur- Be aware it is all about deployment, not training.
ing begins. Such analysis leads to trade-offs between
factors, balancing outputs to optimize overall per- Develop product platforms.
formance while allowing you to rate risks associated Have management guide and lead.
with decisions about program status from a schedule, Design simplicity.
cost or profitability perspective.
Manufacturing process capability and design mar-
Freeze requirements
gin are no longer the seeds of internal conflict but
Another example from the computer disk industry
rather two limbs on the same tree. Analysis of variabil-
helps highlight why freezing requirements is impor-
ity leads to rational decisions regarding whether to
tant. In the early 80s one of the authors of this article
insist on design improvement, investment in new cap-
was a program manager for a flexible disk product
ital equipment or the launch of a massive process
line.
capability improvement using 50 Black Belts.
The technical requirements for the product were
As with the fuzzy front end, rational decisions based
well-defined (dictated by the disk drive design).
on available data are always better than educated guess-
Unfortunately, the packaging and labeling require-
es or dependence on luck. Which would you choose?
ments were not.
So, what if the predictions are wrong? What if some-
We had an extremely hard working market require-
thing unexpected happens? Technology companies
ments person, who was really attempting to do a good
know full well there are always unknown unknowns, or
job of understanding the marketplace. Every other
to quote Mikel Harry, co-founder of the Six Sigma
week he would travel to another customer site and
Academy, You dont know what you dont know.
bring back a list of 50 to 70 unique packaging requests.
At product reviews he would ask that these require-
What DFSS does ments be added to the product specification. This hap-
pened week after week.
DFSS provides a structured way to constructively use Unfortunately, there was no prioritization of these
the information learned from such events in the next requirements. Requests continued to be made late in
program. The message here is twofold: the design effort. Eventually the design and manufac-
1. There are no guarantees, even with statistical turing groups only paid lip service to the demands of
methods. marketing.

S I X S I G M A F O R U M M A G A Z I N E I A U G U S T 2 0 0 2 I 23
S t r a i g h t Ta l k o n D F S S

We would have had more success if the organization uct being designed. The design team now sees the
had worked hard to fully understand the customers immediate applicability of the tool to the design effort.
requirements early in the design cycle. Tools exist to
help organizations prioritize and fully understand the Develop Product Platforms
market of today (and tomorrow). Many organizations design products one at a time,
Unfortunately, even today this example is standard giving little thought to succeeding generations of the
operating procedure in numerous organizations we product. Developing product platforms, however, can
are familiar with, and evolving requirements are a provide strategic advantage for an organization.
major barrier to effectively implementing DFSS. A product platform is a family of products that
shares a common technology and addresses a related
Allocate Additional Resources Early set of marketplace needs. Platforms are planned so
Companies must be smarter earlier in the design several derivative products can be launched, genera-
cycle to avoid design and manufacturing glitches at tion after generation.
the pilot and launch stages. This demands: Books tend to be generated as platforms. If you
Giving thoughtful, early consideration to require- compare various generations of a book, first and sec-
ments. ond editions will typically have a great deal of com-
Selecting rugged design concepts. monality. Software is also developed this way.
In fact, life is this way. The human genome project was
Identifying critical to quality characteristics.
dedicated to characterizing human genetic makeup.
Generating transfer functions. One of the startling discoveries was that the DNA of
Setting customer driven specifications at the sys- earthworms and humans is about 70% identical. With
tems, subsystem and process levels. gorillas and humans, the likeness is approximately 99%.
Many teams are not accustomed to doing all these Product platforms are not new, either. They allow
things early onor at all. companies to shorten time to market, provide greater
Additional resources will also need to be allocated flexibility, save money and react to market shifts more
to assure all the preceding things happen in a timely quickly. Product platforms are a key DFSS strategy.
manner. How much additional resource will be
required is dependent on how aggressively you are Have Management Guide and Lead
currently funding the early design stages. As a rule of Executives must do more than just sign the check
thumb, 15 to 25% additional resources beyond your and expect results. Management must provide both
current funding level at the early design stages will guidance and leadership for a change the magnitude
need to be allocated. Wise allocation of these addi- of DFSS to happen. DFSS is a revolutionary approach
tional resources will provide a handsome payoff of to designing new products and processes, and such
minimal problems at pilot and launch. change will not happen on its own.
Unless the need for this type of change is defined and
Be Aware It Is All About Deployment, Not Training communicated effectively so every employee under-
Today, organizations are making a huge mistake in stands its benefit, DFSS experiences massive resistance.
beginning training of the masses in a multitude of Management, therefore, needs to take the time to
tools without a deployment plan. The transfer rate of understand the tools and implications of DFSS, provide
classroom training to the new product and process leadership, generate the vision, allocated necessary
design effort with this approach is poor. resources, monitor progress, communicate about the
That was interesting. Now lets go back and design improvement model and demand and reward success.
the product the way we have historically done it is the Managers should take time to talk with the new
typical response of designers who attend this type of product team members, particularly in the early phas-
training. Some refer to this type of training as the es of deployment. Most of us do what we believe our
sheep dip approach (immerse folks in the tools with- managers want us to do and what they reward. A new
out reference to their application). product team is no exception.
Better approaches are available. One is the just in
time approach to training. Teach the design team Design Simplicity
about a DFSS tool just before the tool is used, then Back in the late 1980s, a new disk drive was really a
facilitate the application of the technique to the prod- new disk drivedown to the screws used as fasteners.

24 I A U G U S T 2 0 0 2 I W W W . A S Q . O R G
In fact, the RA-90 was designed with 11 different screw
types (after using design simplicity approaches, the
number was reduced to three) to fasten the major sub-
systems to the base plate.
Thanks to the work of Geoffrey Boothroyd and Peter
Dewhurst,4 companies have been able to make great
strides in reducing costs and improving reliability.
Design for assembly (one of the DFMA techniques) dic-
tates manufacturing and manufacturing engineering be
involved early in the new product development effort.
One key is to maximize ease of manufacturing by
simplifying the design through part count reduction.
Design complexity is an important metric that can be
used as a benchmarking or baseline tool. It is a tech-
nique for comparing competitive designs (assuming
both provide the same intended functions).
Complexity is computed as the cube root of X,
where X is calculated as (the number of parts) x
(number of part types) x (number of interfaces).
Lower complexity is better. Lower complexity means
lower cost, higher reliability and fewer opportunities
for assembly errors. Lower complexity is another key
DFSS tool that must be adopted by organizations.

Long-Term Success

For the majority of organizations, long-term success


is tied directly to the new product development
process. Tomorrows revenue and growth are tightly
bound to how successful you are at launching new
products.
DFSS can serve as a mechanism to revolutionize the
way you develop new products. To reap its benefits,
you must be prepared to make major changes.
The size of the effort is formidable, but the payoff
may be no smaller than company survival.

REFERENCES

1. Robert G. Cooper, Winning at New Products (Boston: Addison-Wesley


Publishing Co., 1994).
2. Ibid.
3. From the Nexus of Lexus, BusinessWeek, Sept. 3, 2001, p. 23B.
4. Geoffrey Boothroyd and Peter Dewhurst, Product Design for Manufacture
and Assembly (New York: Marcel Dekker, 1994).

WHAT DO YOU THINK OF THIS ARTICLE? Please share

your comments and thoughts with the editor by e-mailing

Blanton_Godfrey@ncsu.edu.

S I X S I G M A F O R U M M A G A Z I N E I A U G U S T 2 0 0 2 I 25

You might also like