Document 180785

HOW TO MAKE A SUCCESS OF Q U A L n r
FUNCTION DEPLOYMENT
K G Taylor
LBM UK Laboratories
Quality Function Deployment (QFD) is a technique
that was developed in Japan to match produd design
against the needs and wants (in terms ofquahty and
m o n ) of the customer. It also mcludes 0 t h pspeas
ofqualitycontrolnotcoveradhen,foragencral
description, see King [ll. A good deal has ken
written about bow to &sign produfts using QFD, but
recent rrports (Grithn [2]) iadirrtr that not all
attempts mat with SUCCCSS, particulprty in theareas of
compdtrsoftwareandmces.
The suaxsfd use of QFD depends both on the 00operation of the software development team, and on an
un’
of how to tackle somc of the problems
involved This paper suggests 8omc prpaical ways in
which to make QFD wodq includinga way to estimate
the time to ckvclop a matrix, a d how to reduce this
time.
-
PUNNING FOR QFD HOW BIG SHOULD MY
HOUSE BE ?
The i”Kz of the lqlw” gatbenng a d
QFD phases for a produa cannot be owx wmed Not
understanding the overall wants pad & of the
customers mansthatsoonerorlaterthcpoduawill
fail; witness the 8dadc a d o caruidge and the electric
tricycle. If it does not fail because of illoonceptioq it
may be Overtakur by t c c h n o l ~ ,s was the swiss
watch and the slide rule. Both of these piW can be
avoided,as will be shown
one of the reasons why QFD is not l“&
accepted is that it is p e r c a d t o &k
too long or cost
too much It also relies on knowmg what the customers
wan& which again adds to b e ast. Tht first part of
QFD is sometimes called a house of quality. This
section on planning for QFD describes bow to estimate
the time raquired to genetate a house of quahty, and
how to reduct this time when the estimate proves too
long.
Figure one shows an Wervicw of the PrOQd design,
not including code development and tcaing. The
rcquircments gathering phase is beyond the scope of
this paper, but a ftw points arc included here. Like
many things in compting, QFD is largely a ‘garbage
in, gahlagc out” proass; so Wcll-mXaIcbd customs
wants and needs (which lllt the inplt to QFD) arc
mandatory. This may involve additional project cost
and t h e in design schedules Typlcelly this type of
rcstafch tabes three to six I n O andcosts0~or
~
two
perant of the project krdgd. Som companies
regularly collect suggestions fkom customus about
PrOQa %
arc mly genaal
enough to replace a survey.
The requirements used in QFD must be wcU-chosen In
practia.,$bk means that they cannot be arbitmy
comments mackat a trade show. They mustarrathe
whole scope of the p ” s
usc. and be rcprtstntativc
of a large Propomon ofthe customas. This almost
certarnly meansthat they must be gathmd i n a b u ~ t y
or workshop, and not merely colleacd as they are sent
in, perhaps by a small group of customers. Another
mason why they must be coUcctfd is that they must
have a Customer importance fahng asaiakd with
them, and this can only be found in a survey.
f3yfouowingthroughthetwo phasesafQFD shownin
Figure one, the produd design is d m d y cvaluatcd
against “ e K ’ w a n t s pad needs. Difkml! design
proposals can be compared, and the whole p ” s
merit can be evaluated productpricingcanbe hctoled
into the method by aslnng the arstomns to rate ‘Low
cost” ag;unst other wants and rids, and only
including the wants and needs that come above it in
importance.
The word processing system. Figure two introQccs
the example usedin this m,a word proassing
*em The customer wants ami nceds arc shown on
the left of the figure, together with the rating of
importancetothecustomeronascaleof l(a 10.twelve wants and needs are typical of those from
surveys we have con&ct& but are not specifically
from a m e y on word procesSing.
In practice. many more wants and needs are collected,
and these form the ”leaves” on a tne of requirements.
These are sorted into groups, which form the branches
of the tne.The twelve w a ~ t and
s needs in this example
w thebranches at a &&stance
down the W.
The impoxtance rating is found (from a survey) once
thegroupsofwantsandneeds(twelveinthiscase)
have been found It may be necessary to correlate the
survey results with particular market areas with a view
to exploring the need for two or more products.
The choice of (groups of) wants and needs is crucial to
thesc~ocessof the ex&There should not be more
thanabouttwenty, as thiswouIdleadto excessive time
neeckdtoamsiderthe QFD amtrix, since this time is
roughlyproportionaltothe " b n o f wantr and needs
squared.
marketplace. The design objectives are selected by the
technical planning, and development groups to
rcpreswt measures on the customer wants and needs.
Units of m e a " should be agreed, but absolute values
are not needed at the system planning stage. Typically,
one or two design objectives come h m each want and
need, although there are likely to be more functional
objectivesthanthis, even though these maybe statedat
a high level. Functional objectives typically relate to
which emimnments the product should run in, and
additionalfeatures nquested
Inmoslprojectr,moredetailcdI#earchwillbealso
needed. This is likdytobe most or all of the following:
c u s t O " S w a n t 0 r ~ . Thaccanbefapanmus
tothis: Strong, I
"
weals of me&a Tbest me
1)MsrketrescarcbmthebepacrgClan
. dourprocfuasin
weighted 9,3.1, .ad0 in the adysis which follows.
In this pm of the aMlysi& chae a n be no nep;stivlt
effects; these are dealt with in theseandmatrix.
comparisonwiththeoampdition
2) Market m
h on the pMathl market areas for
OurpoQCLs, iachding new markffarras.
3) Technicel rrstarch on recent advances and
Next, for each combination of design
and
want andneed, thequestion isaskad"Howmuch does
meeting this design [email protected] help satisfy the
4) Detailed studies
S
c
o
w
. For each objedk, a technical importonce
mting is Calarlatad by multiplying the custonrer
importance rating by the weight of &ea dscnbed
above,and summing this down the column.
theskinclicatcd,andtbcotherrcscarchfactoredintc
it in the way tobe &scribed.
The resultdthis calculation is shown at the bottom of
Figurc &.This is a first pass at the
ofthe
design~cctive.
Thne further measurcs can be brought in to imprwe
the usefulnm of this mrc by taking into account
QF'D Phase one
competitiveperformance:
1)
Customerpaceptionoftbtstrcngthsand
wealwsscs of OUT produas ci"d with
iruwvations.
on one or more critical aspects of
thepoauct
QFD can beusedfor the last ofthesc, and is closely
tinkdwiththeatherthree.Howevla,the"SystemmidPQFDchart describtd haq should be limitedto
In the 6rst phase of the QFD, the strategy is to develop
a good rmdtrstanding of the abstrucfion of the ideal
produA Tbis abstraction is in the form of product
otjecms that an.gcmral enough to mQrc for the
Lifetimt ofthe pmduct and beyona The wants and
DecQ sbould fonn a 'Vision" for the ideal pmduct, as
indiceted in Figure one. The long-term nature ofthis
vision will be demonstrated by the fact that most (but
not au) OrFigure two relates to the ObJcdives that
d d have ban dcvised for typewriters in use a
century ago.
-
Figure two shows the first phase of QFD for the word
Wants and needs go down the I&hand side ofthe
while o cross the top go the
design obgectivcs used to mawre those wauts and
necds. These design otyctms have betn grouped
under headings of function, usability, perfontlance,
reliability,
imtalMon,
"rice.,
and
documentation. This is to ensure that these important
aspeus of function and quality are not overlooked, and
to compare these aspects with survey data from the
piocessor example.
competition.
2) Maritet trends in "desirable" features to boost des.
3) Performance comparisons between our POQds and
those of competition.
These work by [email protected] the customer importance
rating for each want and d in the following way:
The customer perception (on a d e of 1 to 5) dour
produa now is oamparui with that for the competition.
A suallty plan is derived from this by taking time
wants and needs for which the cuslbmers tbink wt MI
short of the annptition, and giving them a higher
customer importancc rating. This can only be Qnt
wherewchaveancxistingproQain the marketplace.
The SeODnd way that the rating can be modified is to
give it a 20??or 5oo/o increase when the want and need
is perceived as being a featun that everyone wants (at
the moment), or is one that will cxpaad the potential
market for the produa.
The third way to refine the method is to compare ow
product's &hnical specifications (either measured or
prtxkted) with those of the Oampetrtion. However, this
4/2
To predict such a collapse is a key role for produa
planners. It could be &ne by compafing Ihe KduCbg
costs of new Lechnologies with the merit of old o w .
and watching for when the new technology passes the
old one in t e e of its cost-performance.This is part of
the QFD phase two process which follows.
is best done during the second phase of QFD,
especially in a system-wide study of the rvpe discu~sed
here.
Note that,unlike the rest of the phase one matxi& the
lifetime of these competitive measurcs may be only a
Few months;thus they should be checked continually.
DEPLOYING QFD INTO A PRODUCT DESIGN.
Detailed studies. Where a customer want and need
raquirrs added &tail, further investigation will be
aecsslry.Inthisacample,thearact wantsandneeds
for colour printing need further investigation to find
the rmmbadcolours and types of paper q u i d . .
This couldbe the subjea of an additional QFDstudy at
a dctaiIed level.
QFD often stops when the initial h o w of qual~lyis
c
o
m
p
w
l or else the two phases of QFD ciimscd
here are merged into one. This is not to be
" m e n d e d , since the ikst phase af QFD rdates to
long tum
wtomer
wants and
scumd phase relates to ever-
whaars the
produas. It is
~lythat#veralvasionsofthcphasetwomptrixwill
P U " G CUSTOMERWANTS AND NEEDS
INTOCONTEXT.
exist ova the product lifetime.
The relationship between the two QFD matrices was
shown in Figure one. Figure thret shows a phase two
matrix for the word prooessing example. Across the top
are product candidates (in this case componcn& of a
word proassing system), and down the si& axe the
&sign objectives derived from phase one .
Although customers know what they want, they are
probably not awan of certain faaors, such as the
impad of new technology. They may also take certain
things For granted, such as "Good size and number of
keys on the keyboard" which has becm added to the
design ob~cctivcshere. Note that the ways that these
addadobjecbvc arc bcond is ihtical to thase derived
d m d y finmcustomer wantsand aeeds;so thcy are in
DO way "cheating" by being added by the planner.
.> .,
The phase two chart is scored in a similar way to the
phase one chart, but with the added pos?iiity of*
items making a negative contributon to a &sign
obpzive. These negmve contribrdions are not
recorded in phase one because they are due mainly to
technology htations that could be short-tcnn. Evcn
Conflicts. The xxalled "roof" of the house of quality
(indicated in Figure one) is used to identlfy points at
which two design objeaives conflict with each other.
tal conflicts in phase one do no3 have
negahve soores there, but instead carry through to
phase two because correct objectives must be spaSed
for botb sides of the conflict.
For example, the objective for "Low weight and size of
the systcm" oonflicts with the one for "Good size and
number of keys on the keyboard". A small system is
likely to have little space for keys ! These fundamental
design problems are toplcs for research programmes
and innovative solutions. This particular problem
might be solved by the recent range of pen-based
computers.
Figure three shows the score for each candidate item at
the bo".Two xts of scores are shown; the white
sets are for a fixed system (mains powtradandnot low
weight), and the shaded sets are for a portable system.
Notice that the word processor with a grapiucal user
interfixscores highly in both cases, whereas the highfunaion hardware is both heavy and power-hungry.
Sincc a!l laser printers need mains power, they are not
The area of portable word processors leads to many
intemting tra&+i& in system design, and needs
separate consideration in the second phase of QFD.It
is an example of a situation where a single prchct
cannot s
a
w the whole market with today's
technology. Howeva, when colour LCD scl.eens have
got cheaper and clcarer, and the power a"ption
of
various components has beem reduced (or batteries have
Impmved), the whole market may be open to a single
product When this happens, the market for mainso w e d compters could suddenly collapse, just as it
has (almost) done for mains-operated calculators.
dircdcandidatesfortheportablesystem.
scorcS for each component may be compared with Lhe
cost of tbat component Looking at the scorn for a
fixed system, it is notcworthythat the printersall score
highly, men though they typicairy cost I C s than half
the price of the system itself. This tells us that
investing a small additional amount in a printer will be
more cffeaive than investing it in a higher rcsoLutjon
display, for example. A slmilar argument applies to a
word processor with a graplucal user interFace.
4/3
Misunderstandings in the original requirements may be
found by doing this.^
The phase two matrix can also be uscd to find gaps in
the poposcd solution In this case, there are three
"weak m a
(ones with low or lregatin scores);
reliability, installation, and maiatenance. This
indicates a need for a new plan item in this area, such
a s a " m a c " c i
Conclusions
This paper has described how @ty
function
deployment can improve the @ty
and costeffeaivlmess of pPoducts and systems. The two-phase
approachdcscribedhas several unique pmperties in the
way it is used Emphasis has been placed on how to
limit the work involved,yet at the same time i&[email protected]
problem areasthat xuay needfurther investigation.
Multiple matrices The phase one matrix captures
whut functions thc customers want, and how they want
them to work; it tries to be general enough'to avoid
discussion of d o are the users and where they are. The
phase two matrix rep" an attemp at a practical
solution to the djocths in phase one, for a particular
sd of users ina pamcular situation.
By the use of these techniques, &signers cm avoid
producing produds that do not mMthccaIaom"
needs. or Which do not match up to the competitiop
In the umaqAc, by aualysing the two cases for lixed
andportablequipment, we have really been looking at
t
w QFD matrioes; one for a portnblc system, and one
h a fixed one. To Q a compidt study for two market
segments like thest, two se& of customer priorities axt
reaty needed These can be pwessed through the
same phase one matrix to yield two sets of technical
importance ratings for the design objeaks. In this
casethetwomaktshavebccnsfsumedtobe similar
e x q t for the d B e m raquircments on mains power
~ t h y - p r a d r a c e ~ ~ s a v i a s , u d
systems whilst still keeping within budgets.
Rcfe"es
andweight.
[ 11 Kmg, B."Betta designs in half the time"
Customer fadback In all cascs it is wise to check
each poposedpodrM with a sample set of customers
before
pzweuhg
with
devtlopment.
[2] Gri&,
"Eement
A 1992 Journal of
D innwation
Quality Function Deployment
:onf licts
n
Customer
Importance
Desian Conflicts
+
"
A b
Customer
Perception
Plan Item M e r i t
FIGURE ONE
4/4
Phase One
Deployment
Matrix
\\
\
Candidate
Design
Object
Phase TWO
Deployment
Matrix
FIGURE'IHREE
Z 1994 The Institutionof Electrical Engineers
Printed and published by the IEE, Savoy Place, London WCPR OBL, UK