RAO
ASEI\4EI\rr'
.'y.' '
^.^XCHU^^X
/tts)
DEWEY
^^^"^^"
Massachusetts
Instituteof
Technology
Sloan School
of
Management
Working
Paper
HD28
,M414
%
The
Coupling
of
Product
Architecture
and
Organizational Structure Decisions
Rosaline K. Gulati
Steven D. Eppinger
May
28, 1996Working
PaperNumber
3906
Abstract
This
work
is motivatedby
ourinformal observation thatcorporations re-designtheir productsand theirorganizations quite separately.
We
find,however, thatthe relationshipof productarchitecture toorganizational design isan intricate one. This study provides a rudimentarybasis
forunderstandingthe linkages between productarchitectureand organizational design,
which
may
allowmanagers
toimplement
theappropnate organizational orarchitectural structures. Inaddition,athorough understanding ofthe reliance that productarchitecture has
upon
organizational designand
vice versacanaid managers increatingan environmentinwhich
productarchitecturecan exploit theadvantagesofthecurrentorganizational design
and
inwhich
theorganrzational designcan enhancetheeffuiency ofthepersonnel interactions ref^'iired ^o
implement
a product's architecture.We
discuss severalobservationsaboutthedimensionsby
which
these attributes are coupled.Acknowledgment
Funding
forthis research has been provided bytheMIT
International CenterforResearchon
theManagement
ofTechnology.We
appreciate the assistance ofthecompany
thatwe
studiedindevelopingthe
examples
presentedin this paper.Contact
Professor StevenD. Eppinger
MIT
Sloan School ofManagement
Cambridge,
MA
02142-1347
Eppinger@mit.edu
MASSACHUSETTSINSTITUTE
OF T'-^wvoi.OGY
SEP
le
1996I. Introduction
This paperexplores theinterdependence between
two
keydecisionareas inthedevelopment
ofnew
products: product architectureand organizational design.We
conducted afieldstudy ofaudiosystem
development
inamajorAmerican
automotive firm. This firmcompetes
inglobalmarkets
and
utilizesnew
technologieswhich
give risetoseveral p)ossibleproductarchitecturesand
organizational structures.By
separating thesedecisions forourdiscussion,we
are able toexplore
some
ofthefundamental issueswhich
couple them.Our
results suggestseveralways
inwhich
decisionsand plans forproductarchitectureand
organizational designmust
be integrated.Inorderto explore thecoupling betweenthese
two
issues,we
focusedourstudyalong thefollowing dimensions: product andproblem decompositions, integration
mechanisms,
communication
patterns,supplier relationships, and reporting structures.We
define productand
problem
decomposition to be thepractice ofsplittingacomplex
engineering challenge intoseveral simpler ones, while integrationis thechallengeofmergingsolutions to theseseparate
problemsintoanoverall system.
Communication
patterns (i.e.how
informationmight flow)depend
on
the type and structureoftheprojectteam [Barczak andWilemon
1991]. Supplierrelationships can takeamyriad of forms. Therefore,
we
were particularlyinterested ininvestigating the characteristicsofthese alliances
which
might haveinfluenced productarchitectures. Reporting structures
we
examined
were thosewhich
affected orwere
affectedby
Figure 1.
Problem
Decomposition and Organizational Assignment.The
organizationmust
decompose
thecomplex problem and
all(x;ate tasks to the product
de\clopment team
(solid arrows).Additionally, information
must
be ccx)rdinated across individuals,teams, and suppliers (dashedarrows).
Pastresearch in product
development
hasshown
thatacompany's
capability toconceive anddesign avariety of superiorprcxiucts and bring
them
tomarketfasterthan its competitors can bea sourceofsignificant competitiveadvantage [Wheelwright and Clark 1992J.
Wheelwright
andClark
emphasize
the imptxtance oflearning in the organiziition.Companies
that follow a path ofcontinuous
improvement
inproductand processdevelopment willmore
consistently "design itright thefirsttime" and yield a head start in gettingtheir productsto market.
An
understandingofthe key relationships in product planning can facilitate "designfing] it rightthe first time."
Priorresearch by Clark and Fujimoto [1991] explores the impactofstrategy, organization,and
management
upon
product development.They
claim thatmanagement
direction and thedevelopment
organization both playcritical roles iii providing the integrated effortandleadership needed tosuccessfully execute a product's architectural plan
and
tomove
that prtxluctefficientlyand quickly tomarket. Rosenthal [1991J suggests that, inorderto besuccessful, a
firm
must implement
a managerial view ofthedesign anddevelopment
process thatattemptstohelp catch design flawsearly, correct mistakes,andavoid long
development
delays.Henderson
and Clark [1990] indicate that"architectural innovation has the pcHential to offerfirms the
opportunity to gain significantadvantage," in thecontextof understanding that a well-entrenched
Clark [1987] indicates thata corporation's problem solving structuremirrors thetechnical
and
conceptual structureofits product(s). Innovative changesin the productcan expose
discontinuities inorganizational knowledge, informationflows, andprocedures. Ultimately, the
natureof these breakscan determine the styleof competitive response [Clark 1987].
This research connectsarchitectural choicesandorganizational choices. Itexpands
on
ourprevious research projects
which
haveexplored system integrationincomplex development
environments.
McCord
and
Eppinger [1993] highlight theimportance ofdeterminingtheneedsfor integration and coordination
by
studying theunderlying technical structure ofaproject.Pimmler and
Eppinger [1994]show
thatan understanding ofthe "system engineering" needs,which
arise becauseofcomplex
interactions betweencomponents
of a design, is useful todefinea product's architecture and toorganizedevelopment teams. Finally, Morelli, Eppinger,
and
Gulati [1995] propose thatfor the
management
ofproductdevelopment projects, certainaspectsof organizational design can be planned byanticipating the technical
communication
linkagesrequiredforprojectexecution.
Ina
complex
product, theco-dependencies between architectural andorganizational choices can beimportant considerationswhen
decomposing
aproblem. Alexander [1964] states thatthereisan important underiying structural correspondence between thepattern of a
problem
and theprocessofdesigning the problem's solution In ourcase, itis feasiblethat the architectural or
organizationaldecompositions
depended
on
thenumbers
and distributionsoftheirpotentialintermediatestable forms '. Thatis, thedirection in
which
theorganization orarchitecture mightevolve can besignificantly influenced by it's priorform.
Complex
systems will evolvefrom
simple systems
much
more
rapidlyifthere are stableintermediate formsthan ifthereare not.Furthermore, the
components
of a technological system (suchas acomplex
organization) willinteract. Therefore, theircharacteristics will derivefrom thesystem [Bijker 1987]. Bijker,
Hughes, and Pinch 1 1987| give ihe
example
of themanagcmcnl
siruclurcofan cleclriclight andpcnver ulilily depending
on
ihc character ofthe functioninghardwareor artifacts in the system.They
alsostates thatthemanagement
structure reflects theparticulareconomic
mix
ofartifacts inthesystem, and the layoutof the artifact
mix
is analogous to themanagement
structure.Simon
[1990] argues that in nearly
decomposable
systems (such as these) the shorl-run behaviorofeachofthe
component
subsystems is approximately independentoftheshort-run behavior oftheothercomponents. In the longrun, the behavior ofany oneofthe
components
depends only in anaggregate
way
on
the behavior ofthe othercomponents [Simon
1990J. In lightof thefact thatcomplex
problemsinvolvecommunication
among
many
people,von
Hippel [1990] proposes thatfirms specify tasks in order toreduce the problem-solving interdependence
among
them
bypredicting
w
hich tasks are likely tobe importantnew
information sourcesandwhich
tasks affecteachother.
ProductArchitecture
We
defineproductarchitecture tobe thesetoftechnical decisions (theplan) for the layoutoftheproduct, its modules,and for theinteractions between themodules. Productarchitecture is the
scheme
bywhich
the function of a productis allocated tophysical components. Itcan be a keydriverof theperformanceofthe manufacturing firm and relates toproduct change, product
variety,
component
standardization, productperformance, andprcxluctdevelopment
management
[Ulnch 1995]. In
some
companies,the product architectures offlagship products might evenguide decision prcx:esses.
At
Sony, design isdone
verydifferently dependingon
the product.For the
Walkman,
generational changesare led byengineering, with heavy involvement from topmanagement.
In other products, marketingand sales lead certain classesofchanges. In productsthat
do
not fit in eitheroftheabove
two
categories,the industrial design organization plays aheavy
role [Sanderson 1995].The
physicalelementsofa productare the parts,components, and subassemblies that ultimatelyimplement
theproduct's functions.The
chunksare the collectionsoftheseelements and somay
implement
one
orafew
functional elements intheirentirety [Ulrichand Eppinger 1995].A
modular
architectureisone inwhich
chunks implement one orafew
functional elements in theirentirety,
and
where
the interactions between thechunks arefundamental to the primary functionsoftheproduct.
An
example
of amodular
architecture isacar radiowhich
is utilizedin severaldifferentaudio systemsacross vehiclelinesand is a stand-aloneproduct.
An
integralarchitecture is theoppositeof a
modular
architecture. It isoneinwhich
asinglechunk
implements
many
of thefunctional elements, andwhere
the interactionsbetween
thechunks
arenot well-defined and
may
be incidental to thefunction ofthe product.An
example
ofan integralarchitectureis the integrated control panel developed forthe 1996 Ford Taurus audio
and
climatecontrolsystems.
Orf^cinizdliondlDesign
Organizational design is thedecision process thatbrings about acoherence between the goals
and purposes for
which
the organi/iilionexists, the patterns ofdivision ol laborand interunitcoordinationand the people
u ho w
illdo
thework
[Galbraith 1977].We
focusourattention onthe creation offormal managerial processes and
communication
channels that facilitate theorganization's decision process.
One
importantdimension of organizational design is the typeofstructure thatgives rise toits capabilitiesand coordinationabilities.
A
pure functionalorganizationencourages long-term technical specialization.
However,
physical andorganizational distance between sub-functions increases.
A
pure projectorganization focuses anorganization's energies major
development
projectsand encourages cross-functionalcommunication.
However,
by doing this, itmay
sacrificesome
functional expertise[Wheelwright and Clark 1992].
A
matrix organizationintegrates the specialized resources oftheorganizationswithoutorganizing aroundaself-contained product orproject [Galbraith 1977]. Although a matrix
organization solves
many
ofthe problemsof pure functional and pure projectorganizations, it isimportant torecognize that ithas otherdrawbacks
which
arebeyond
thescope ofthis paper(e.g.problems caused
by
p)oor relations between units) [Galbraith 1994].Otherorganizational design
mechanisms
include the creation of slack resources (i.e. addingadditional resourcesand reducingeach individual group's required level ofperformance),
creating .self-contained tasks(i.e. assure thateach group hasall theresources it needs to perform
its task), investing in vertical information systems(i.e. invest in
mechanisms which
allow theorganization toprocess information acquired
dunng
task performance withoutoverkxiding thehierarchical
communication
channels), or creating lateral relations (i.e. selectivelyemploy
lateralCouplingArchitectureto Organization
While
other research hasanalyzed thedimensionsofdecompositionof organizationsand productarchitectures independently [Sanderson 1995,
Uzumeri
1995,Meyer
1993],this paperhighlightsthe
ways
inwhich
organizational competenciesandframeworks
arecoupled to architecturalinteractions
and
theirfunction in theorganizational structure.We
draw
the conclusionthat thisdistinct relationship meritsspecial considerationinmanagerial decision making.
A
thoroughunderstanding ofthe reliance thatproductarchitecture has
upon
organizational designand
viceversacan aid
managers
increatingabeneficialenvironment inwhich
productarchitecture canexploit the advantagesofthecurrent organizational designand in
which
theorganizationaldesigncan
enhance
theefficiencyofthepersonnel interactions requiredtoimplement
aproduct'sarchitecture. Additionally, organizational designcan assisttheexecution of a product's
technology
by
facilitating the integrationof variousdisciplines, technologies,components,and
systems intoaproduct.
The
next section ofthis paperoutlines ourresearchmethodology
and introduces theaudio-system design focus ofourfield work.
Then
we
presentexamplesofarchitecture affectingorganizational designand organizational design affecting architecture,
and
discuss the couplingofthese decisions.
We
conclude with asummary
ofthe implications for practitionersand
directions for future research.
II.
Audio System Case
StudiesThe
data forthis case studycome
from interviews andobservationsofaudiosystemdevelopment
teams in
two
verydifferentorganizations (oneAmerican and
one European)in amajor
US
automotive manufacturing firm. Although underthe
same
parentcompany,
thetwo
sites aredifferent in
many
ways,includingculture,language,work
habits, vehicle programs,management
style, technical capabilities, supplierrelationships,and scope oftechnical responsibility.
We
interviewed personnelin engineering, managerial,andbusiness functions
from
globaldevelopmcnl
teamson 26
dilTcrcnl car lines and 63 audio systems. Thjsparticularcompany
designs and manufactures about3(X) different radios.
The
intersiews were conducted overafivemonth
period in 1995.Our
study concentrated largely on factory installedautomotive audiosystems. Finally,
we
extracted theexamples from the case studiesand re-framedthem
intomore
general issues forour presentation here.
An
audio systemconsistsofall thecomponents
ina vehiclewhich
aid in providing audibleinformation.
These
components
include, butare notlimited to the radio, amplifier, speakers,wiring harness, andcellulartelephone.
r^
IF Speaker CellJar Anlema TeOaT Phone lU AM/FM Anlema LB [Speaker^^:^
A Tuner Anlema Mo(or Power Ampdfterm
Dsp(ay Cassene—
CD—
Changer FronT Conlrols I Mear IControlsFigure3.
Audio System A
n-hitcci'ire.lakec JJ Speaker Hear
4
m
Speak^
At
first glance,an audio systemmay
appearto be a very simple groupofcomponents. In reality,however, it is quitecomplicated.
An
audiosystem can involveanywhere
from 4(X) to600
components, six totendesign engineers, and three to fiveoutsidesuppliers. Itcan also take three
years to fullydevelop.
The company
we
studieddevelops ten totwenty audiosystems atone
time. Outside suppliers are often in\ol\cd in \arious functions including: integrated circuit
Some
ofthecomplexity inan audio system arisesfrom the technical interactions and couplingeffects from nesting. Nesting refers tothe idea that
components
within alargersystemareself-contained, such as theaudiosystem within avehicle [Christensen and
Rosenbloom
1995].Both
modular
and integral architecturescan be arrangedina nestedfashion. Ina nested hierarchy,each
component
can alsobeviewed
as asystemwhich
comprises sub-componentswhose
relationships toeachotherare alsodefined
by
aproductarchitecture. Similarly, the productmay
alsobe
viewed
asacomponent
within alargersystem, relating toothercomponents
withinadefinedarchitecture. Simply, products
which
atone levelcan beviewed
ascomplex
architectedsystems actas
components
insystems ata higherlevel [Christensen andRosenbloom
1995].Subsystems
aredefinedas systems-of-use within a nested hierarchyofproductarchitectures.Figure
4
illustrates anested hierarchyofproductarchitecturesfrom vehicleaudiosystems toentire vehiclelines.
At
the highest level, the architectureof a vehicle platform iscomprised ofall ofthe differentvehicle lines thatthis particular
company
makes.At
thevehicle platform level, pricing isdetermined,manufacturing issues are dealt with,
modular
components
are specified,and
theservice
and
repairrequirements arelaid out.The
nextleveldown,
thearchitecture of a vehicleline is
where
liaisons can be createdwith marketingand sales,ergonomics are taken intoconsideration,
and
noise, vibration,and harshness issuesare oftenuncovered.At
the thirdlevel,the audiosubsystem level,orany otherelectronic subsystem forthat matter, theaudio-specific
customer requirementsare uncovered.
The
interfacesfjecifications and requirementsand
thesubsystem designspecifications are delineated. Lastly, the architectureofthe radio, in turn,can
itselfbe analyzed as asystem
composed
ofintegratedcircuits, speakers,and
fasciadesign,forexample.
Archilfcturv ofa Vehicle I'latform
PricingDetennination ManufacturingIssues
Architcflurc ofuVehicle Line
l.iais«n witliMarketing
l-.rgononiii.!-Architecture ofanAudio Subsystem
("usiomcrRequirements
Antenna Requirements SpeakerRequirements
r 1
'ArchitcctureofaRadio ,
I Integrated CircuitSpetificatuins '
I
SpeakerRequirements |
ControlsandDisplays i
I Tuners andAmplifiers
I
FasciaDesign I
InterfaceSpecificationsandRequirements
Subsystem DesignSpecifications
Liaison withSales Noise,Vibration,andHarshness
LXtemiine ModularComponentSpecifications ServiceandRepairRequirements
Figure4.
A
nested hierarchyofprtxluct architectures.The
interactions alongwith thenesting ofthe audiosystem in theautomobile forcea couplingbetween theaudio system and thevehicle.
By
making
design decisions ofoneor theothermdependently,a firm
would
sub-optimize pieces ofthe entire automobile.III. Discussion of Findings
In this section,
we
will presentourfindingsbyfirstdelineating themanner
inwhich
and towhat
extentarchitectural choicesarecoupled to theestablished organizational capabilities
and
structures. Next,
we
investigate the natureofthecases inwhich
organizational design drovearchitectural decisions.
A. Architecture choicesdictate organizational design choices.
Developing audio systems forautomobiles is a surprisingly
complex
task.Not
only is thearchitectural layoutofa singleaudiosystem extremely complicated, but the interactions
between
a vehicle and its audio system
may
involve largenumbers
of peopleand physical parts.Decomposition
determinesteam
assignments.Products that are
decomposed
into architectural chunks encourage theassignmentofateam
toeach chunk. Productsare usually
decomposed
until a team, individual,or suppliercan beassigned responsibility foreach
chunk
[Rechtin 1991]. Traditionally the groupsoffunctionalelements ina product have had afunctional team assigned tothem.
An
example
ofan effectivechunk-to-team
mapping
is theassignmentofautomotive systems todepartments/teams suchas aclimatecontrol systems oraudiosystems. Forstatic
modular
architectures inwhich
theinterfaces arevery well understood, thisapproach
makes
sense.However,
when
the architecturechangesor ifthe interface parameters are not well defined, this approach
becomes
lessappropriateas it
would
require theorganization tochange.A
change intheorganizationwould
probably be verycostly for justonearchitectural generation.
Inother design
domains
(e.g. software),complex
systems arebroken intosmaller,more
manageable
tasks.The
complex
detailsofeachofthe smallersub-systemsarebelow
theabstraction barrier. Often there is anagreement(or contract) tospecify the detailsof the
interfaces
and
parameters passed between thecomplex
systemand
each ofthesub-systems.Effectively, ateachlevel, the architectureis theplan ofall these abstractions
and
contracts[Moses
1995]and
teamsare assignedto thesubsystems. See Figure5.Complex
Problem
toplevel decomposiljon ABSTRACTION BARRIER ABSTRACTION BARRIERFigure5.
Problem
Decomposition and AbstractionThe
organizationin ourstudytraditionally has engineered strictlymodular
architectures.So
people have been grouped notaccording to theirtechnical specialties, butinstead according to
thephysical
modules
of the product. Since today's audiosystemarchitecturesarebecoming
more
integrated, theorganization has adapted toaccommodate
cross-moduleteam
structures.Incidentalinteractionscatalyzetheformation of
problem
solving teams.Often incidental interactionsoccuratthe intersectionofthe
decomposed
elements. Inourstudy,couplingproblems
were
more
difficult toanticipate in themore
novel technologies.These
coupling issues giverise to integrationproblems
and
cause thecreation ofad
hoc systemintegration teams to handle theseissues.
A
solid analysis ofwhere
interactionscan facilitate theclusteringofthe high frequency interactions within subsystems can minimize the interactions
acrossdifficult barriers
[McCord
and Eppinger 1993]. Furthermore, pre-development planningand product integrityenhance theperformance ofprtxluct
development
teams (andalso minimizeunplanned incidental interactions), especially ifthe organiziition is verysystem focused, since
lead time and prcxluctivity
become
much
more
predictable(Brown
and Eisenhardt 1995].Inourcasestudy, theformation of
ad
hoc teamson
asmall scalewas
notuncommon. Though
we
did find thatad
hoc teams had not been planned inadvanceduring the productdevelopment
planningprocess. This lack ofanticipation caused the programdelaysand extracosts. Inone
instance, serious system problems had notbeendiscovered until the entireaudiosystem had been
integratedin thevehicleduring theprototyping process.
Due
tothe urgencyincorrecting theseissues (i.e. the vehiclecould notbe sold with a malfunctioningaudio system
and
a majordelay inthe manufacturingofthe vehiclecan be extremelycostly), aformal troubleshooting
team
was
established.
Architecturedetermines
communication
patterns.The
layoutof the productarchitecture's fundamental and incidental interactions impliesaspecific patternof organizational communication. Ifthere exist barriers to theexecution ofthis
pattern, these barrierscan catalyzedelays in theproductdevelopment cycle. Thisissue
becomes
particularly relevant toco-locatedteams, especially ifonlyasubsetofthe larger
development
team is being co-located. Additionally,
knowledge
ofspecific types and patterns ofcommunication and
the ability topredictcommunicationsmay
allowmanagers
toimplement
appropriate organizational structuresbased
on
aproject'stask structure [Morelli, Eppinger,and
Gulati 1995].
On
theother hand,when
one canidentifythattwo
people ortwo
groups need to shareinformation,it is important toassurethat thecorrect information
exchange
takes place.Sometimes,
aninformalmethod
suchasco-location byitselfisnotsufficient. For example, it isimportant notto
make
theassumption thattwo
engineers withdesks inclose proximity willcommunicate
the necessary information toeach other. Additionally,we know
from
Allen'scommunication
vs. distancecurve [Allen 1977], thatitis unlikelythatengineers willcommunicate
withotherengineers intheirdepartment ifthey are locatedseveral floors apart.Furthermore, theavailability, transfer, and useof information areall distinctconcepts.
Co-location increases the availabiliiy ol the information, while transfer implies that theinformation
isappropriately disseminated. Use implies that theactual information is utilized. Co-location
does not necessarily ensure that the correct information will be transferred
and
used. Inparticular,in novel situations,
where
itis imperative thattechnical information be transmitted toa group relati\ely unfamiliar with the
new
requirements, firms mightconsiderutilizing multipleinformationchannels or gathering mcthcxls inorderto insure transferofthe correct information
and
effective utilization.Architecture determinesthe feasibilityofco-location.
Ifasystem is simple enough, theneed forhigh-frequency interactions can be fulfilled
by
effectively co-Icx;ating the entire team. This approach is valid and successful forsmall projects
and teams.
When
thesystem iscomplex, thesame
rulesdo
notapply. Ifthe team is large,thereasons forco-locationare
no
longervalid.Subsystem
dc\elopersdo
notneed to interfacedirectly \\ithall of thevehicle developersatall times. For example, an audio system team
may
need to
work
more
directly with theinstrument panel designteam duringsome
part oftheproduct
development
cycle;during other partsofthe process theaudiosystem designteam
may
need to interface withthe wiring harness orcliirate control design team. Often, there exist
interactions
which
require technical expertise forashort time during the productdevelopment
cycle.
Another
way
todeal with thesecomplexities is bytemporarily co-locatingone
ortwo
keyengineers from the subsystem
development
teamon
the vehicleengineering team. Thisapproachmight be a feasible solution ifonlyone vehicle teamexisted.
However,
in realityeach of themany
vehicle teams needsthe expertise ofasubsystems engineer. In addition, the subsystemsdepartmentitself
may
have itsown
needsandbe unwilling topartwithits engineer(s) foranextended period oftime.
Given
that multiple types ofintegration exist(within productdevelopment teams, within systemteams
which
aremade
up
ofseveral productdevelopmentteams, and betweenexternal teams andproduct
development
teams),we
recognize thateachintegration ismost
difficult atthe largersubsystem level
[McCord
and Eppinger 1993]. Additionally, past research hasshown
thatdecomposition
becomes
easierattheinternal team level.At low
levels, integration iseasybecause interactions occurin high frequencies over asmall cross-functional team.
At
higherlevels, interactions
become
more
occasional and less frequent as the team gets bigger.Integration iseasy
ina
sma
1 team.As
interactionsbecome
more
incidental and lessfrequent, integration
becomes
more
difficult.Figure 6. Interactions inanorganization
Lastly,awell understood interface between architectural chunks minimizes theneed for
ad
hoccommunications. Ifthe architecturecan bepre-planned
and
allthe interfaces can bepre-specified, co-location
may
not benecessary. Furthermore, ifcomplexity isaccurately defined,tradeoffs
between
complexity,quality, and productdifferentiationcan be thoroughly considered.In ourstudy,
one
team dedicated toreducingcomplexitychampioned
thecomplexity issuewithout
due
regard tootherfactors, causingmany
decisions tobe re-examinedand resulting indelays in theproduct
development
cycle.B.
The
establishedorganizational capabilitiesand
structures dictate architectural choices.The
organi/ations thatmanufacturecomplex
products such asautomobilesexhibit complexity inmany
facets.Each
division, eachdepartment, each prcxiuctdevelopment team consists ofmany
different internal andexternal networks.
Each
ofthese networks in turn helps todefine theorganizntion.
The
capabilities and structures ofthatorganization then influence architecturaldesign.
Sialic organizations give rise to staticarchitectures.
Results
from
our study validatedourpriorobservations thatfixed organizational structuresgenerate products
whose
architectures remain fairly ngid.We
believe that this generalizationmay
hold acrosssome
ofthemajororganization types(i.e. pure product, pure functionalormatrix),
however
this hypothesis remains tobe testedin afuture study.The
architecture is oftenareflection oftheorganization. Iftheorganizational structure isstatic, the architecture is likely
to bethe
same
overmany
product generations. Integratingand changingthe architecture alsobecomes
difficult.The
longerthis cycle remains to be true, thestrongera prediction it becomes.Gi\ en this premise, one might suppose splittingintoproject organizations, forexample,
would
facilitate having each productarchitecture slightlydifferent from the others, because each
organization isaproject team (i.e. eithera matrix or a
hybnd
thatcuts across all the traditionalfunctions, butdoes notnecessarilyutilize cross-functional expertise). It appearsthat the project
organization might be
more
adaptable tonew
architectures, because each project team issoseparate. Innovations inone architecture
may
be difficult toimplement
inthe otherproject team.In fact,each particularorganization
may
be staticand therefore produce astaticarchitecture.Additionally,pure projectorganizations lend to produce architectures reflectiNe ofthat particular
organizationratherthan thatofthe
company
as a whole.The
productsmay
also requireengineeringefforts thatare duplicated in othersegments ofthe corporation, hence
becoming
more
costly than necessary [Galbraith 1973]. Purefunctional organizations might bemore
dynamic
than pure product organizations (because theproduct organizationsare pooledtogether), butthey are
more
likely to losesightofthe goalsofeach individual product'sarchitecture. Matrix organizations cansolve
some
oftheseproblems, but they have otherproblems
which
arebeyond
the scope ofthisresearch.When
considering organizational designs,it is importantto recognize thatthereis
no
one bestway
toorganize andnotall theways
toorganizeare equallyeffective [Galbraith 1977].
Organizationalskills
and
capabilities affectarchitecture.An
organization with specific skillssometimes chooses its architecture suchthat the impactofthoseskills are
maximized
inordertogain acompetitive advantage.As
Meyer
and Utterback[1993] have
shown,
product families canbe a resultoftheunderlying corecapabilitiesoftheorganization. Furthermore, thecross-functional skills of a successful product
development
organization
and
effective synergies with thefirm's existingcompetencies can lead toaproductadvantage
[Brown
and Eisenhardt 1995].For
some
corporations, thismethod
hasproven tobequiteeffective. For example, in 1933Toyota's founder,
Toyoda
Kiichirooannounced,"We
shall learnproduction techniquesfrom
theAmerican
method
ofmass
production. Butwe
will notcopy
itas itis.We
shall use ourown
research andcreativity todevelop aproduction
method
that suitsourown
country'ssituation"[Ohno
1988]. During the 1970sand 1980s,Toyota
challengedGM
forthe title ofthe world'slargest automobile manufacturer [Pine 1993].
Supplier relationshipscan affectarchitecture.
Managing
supplier-customerrelationships is a verycomplex
task. Ignoringtheinterdependencies in thispartnershipcan have direconsequences later inthe product
development
cycle[Kim
1993].Some
ofthemajorways
inwhich
automotivesuppliers havebeen organized in ihc past have been either tobe dedicated supphersor ifthey weren't dedicated
suppliers, long term relationships had been established such that both the supplierand the
manufacturer underslcxxlone another's
mode
of operation and the manufacturer had the abilitythen to anticipateand
know
w
hich supplier tocallwhen
the time forearlysupplier involvementcame
around.The
advantage ofsucha relationshipis thatpurchasing did not have to beinvolved toqualify a
new
supplierand timewas
savedduring the prcxiuctdevelopment
cycle.In ourstudy, the suppliers
were
not dedicated, andwe
noticed that ifthe suppliers did not havethe
same
incentivesas the prcxiuctdevelopment
teams, thequalityand the timelinessof theinterdependent tasks
was
poor.We
surmise that thismay
be thecase because each of thesubteamsof the larger prcxiuct
development
team hasitsown
organizational allegiances orpersonal agendas thatprecede their
commitments
tothe goalsofthe project. Furthermore, theseallegiances
may
exacerbate theproblem ofsub-optimizing thesmaller systemsby
creatingdisincentives to globallyoptimize. In thecaseoithe suppliers, they
wanted
tomaximize
theirprofits.
Sc^metimes,this can result in higher than necessarycnerall \chiclecosts if thesupplierdevelops
anarchitecture that reaches
beyond
the specifications ofthe original architecture.An
understanding ofthe supplier'sagenda can allow oneto position the product such thatthe
suppliercan cither get
more
salesvolume
orutilize t'leireniiinecring expertise in otherprofitmaking
\enturcs.Lack
ofcommitment
and/orlackofgoal alignmentfrom all segments oftheprcxiuct
development
team can bedue toa desire toappease uppermanagement
outside thesubsystem groupat the expenseof a prcxluct'sdesign, a desire fora vehicle
program
manager
tomake
aradicalchange
late in thesubsystemdevelopment
cycle, cultural differences ingeographically disparate departments,differentcompensation systems in different countries,or
different
employee
motivators in differentcultures.Organizationaldesign of agloballydistributedteam affectsarchitecture.
Sometimes
there arecompelling reasons toglobally distributeamulti-national productorganizationwith global sales (e.g. keeping intouch with customers, beingclose to
manufacturing facilities, etc.).
Given
this global distribution,itwould
beideal to clusterthearchitecturesuchthat the high- frequency interactionstake place within eachsiteand
low-frequency interactionscan occuracross sites.
We
haveyet todemonstrate thatvirtualco-location (using collaborative technologies suchas video conferencing,electronic mail,
distributeddatabases, Internet,
CAE,
etc.)iscomparable tophysicalco-location.For example,
we
observed thataglobally distributed organizational structure can causedifficultyin scheduling video-conference or tele-conference meetingsdue toverylittle overlapin the
workday
and difficultyincoordinating team projectswhen
the team itselfis geographicallyseparated. Co-location is not always thecorrectanswer tothese problems.
Good
reasons forco-location inalarge
team
include situationswhere
thereis alotofhigh-frequencycommunication,situations
where
interface specifications are notfullyspecified,orsituationswhere
communication
needsare unpredictable.Team
A
TeamC] sutvlciim"^ ^iib- learn (sub-lcam ftub-li-arrTlt
well specified interface ^ub-teani (sub-lcam"] (Tulvleam^ poorlyspecified interface (^b-te;iin sub-tciim^ (fub-team"Team B
TeamD
Figure7.
A
well specified interface betweenTeams
A
andB
allows for lessformal co-location methods.
However,
co-locatingTeams
C
andD
would
facilitatethe high-frequency interactions.Virtual co-kx;ation canea.se theproblem, but not alleviate itfully.
Although
some
research(Hamen
and Nihtila 1995] indicates thatthe useofelectroniccommunication
reduces the effectof physical distance
on communication
activities, face tofacecommunication
is still necessary.The
useofinformation technologiescan beenhanced ifthey are seenas atool to facilitate thebusinessand integrated into the business.
Davidow
andMalone
[1992] recognize that, "in yearsto
come,
incremental differences in companies' abilit''?s toacquire, distribute, store, analyze,and
invokeactions based
on
information will determine the winners andlosers inthe battle forcustomers."
Goldman
[1995] affirms this research by mdicating that,"sharing theexpenseofprecompetiti\'e technology, facilities,and resources leaves
more
resources tospend oncustomizing prcxluct features andservices thatprovide forcompetitive ad\antage.
The
goal is tounite
complementary
core competencies in ordertoserve customerswhom
theseparatecompanies
could not serveon theirown.Each
member
ofthis type of \iilual organi/ation ischosen because it brings
somethmg
unique that is neededtomeet
acustomeropptirtunity."In theory, the virtualorganizationcan give a
company
access tomore
specializedcompetenciesthan
any one
organizationcanafford tomaintain and henceconcurrently engineermany
more
products.
However,
more
compelling studiesaboutthe effectiveness ofvirtual co-location inisolation haveyet to becompleted.
Inourstudy,
one
ofthe vehicleprogramsutilizedresourcesfrom
NorthAmerica
as well asEurope.
The
satellite organization did notalwayshaveaccess toallofthe resources ofthehome
organization tocomplete theirprojects. Ifthis is the case, the situation is easily rectified ifthe
home
organizationcaninsure thatits satelliteorganizationsareassignedmdependently
executableprojects.
However,
theremust
be adequate information flowbetween
the parentorganization and thesatellite inordertoutilizeeachorganization's resourceseffectively.
Lastly, co-locationshould not beconsideredthecompletesolution tothis problem.
Tyre
andvon
Hippel [1995] suggestthat thelocation itselfplays an integral roleina projectteam's efficacy.
They
believe thatmanagers
should considerwhat
types ofknowledge and what
forms of searchare
most
important toateam's progressat anygiven time,and selectthework
locationaccordingly.
Importance
of
effectivecommunication
Automotive
audio systems havetraditionally been designed bylargelyautonomous
organizationseitherwithinautomobile manufacturingfirms orat supplier firms.
Sometimes
this typeoforganizational structurepromotes the visual styleofan audio systeminterface tobe very
different
from
the style ofthe vehicleitself. Recognizing this problem, theUS
auto industry hasplaced greateremphasis
on
organizing vehiclelineteams. Inthis fashion, theyhope
to createnatural
communication
patterns between subsystem designers(e.g., audiosystems) and thevehicle interiordesigners.
Organi7ati(>ns lend lo(Jc\clop languages ol ihcir
own
as people aho
shareacommon
set ofproblems lend loshare shorihand
ways
olrelemng
loactiviiiesand technologies. Technicaldcparlmcnls hire people
who
have been irained andknow
ihe languageofIhc deparlmcnl'sspecially.
Such
a language pcrmils people locommunicalc more
efficiently by transmittingmore
information with fewersymbols.However,
despite the fact that specialized languagesincrease efficiency withinadepartment, they decreaseefficiency betweenorganizations
[Galbraith 1973].
Many
ofthe interactionswe
studied involved these types of formal organizational ties.However,
we
did observethaimuch
ofthe necessarytechnicalcommunication
occurred throughthe informal organizational networks. Krackhardt and
Hanson
11993] indicate thatalthoughthese informal networks canexpeditedelayed initiativesandaid in meetingdifficultdeadlines,
theycan alsobl(x;k
communication
andevoke
opp<isilion unlessmanagers
know
how
loidentifyand direct them. Moreover, Granovetter's [1973] research has
shown
thatsmall-scaleinteractionscan be translated into large-scale patterns, that is the strengthof mtcrpersonal ties
can affect the political
makeup
of theorganization.The
underlyingphilosophies ofan
internationalcorporationoftenassume
globalease ofcommunication.
Architectural direction canbe very difficult to
communicate
acrossorganizations. For example,even thesimplest explanationsofarchitectural characteristicscan be misinterpreted ifagroup
speaking
American
Englishconveys themessage
toa group speaking British English or agroupspeaking English translated from
German.
In the
company
thatwe
studied, the European di\ision and the NorthAmerican
di\ision sharedonly
human
capital until recently,now
they alsoshare products.Only
now
is thiscompan\
experiencing the
growing
painsofglobalization as heightened bythedifferences in eachorganization's decision models andunderstandings oforganizational processes.
As
Allison [1971]showed
through hisanalysisoftheCuban
missilecrisis, the actsofcomplex
organizations cannot be simply understood by analogyas the purposiveacts ofindividuals. This
simplification obscures the neglected factthatcorporate policy decisionsare not
made
by
onedecisionmaker,but ratherbythe bureaucraticresults of aconglomerate oflarge internal
organizations.
Social researchhas
shown
thatthecreationoflateral relations inan organization oftenprovides amechanism
which
reduces the quantityof decisionsreferredupward
in thecorporatehierarchy.Itis
assumed
thatinformal processes are necessaryand inevitable inacomplex
organization.However,
when
alarge productdevelopment team iscomprised ofdiffering attitudes,containsmembers
from
differentcountriesand is geographically dispersed, the effective useofjointdecision
making
may
also requireaformally designed process [Galbraith 1973].Identificationof internally efficientdepartments
which
arehindered bybarriers toexternalcommunication
may
allow anorganization toselectaproductarchitecturesuch that thehigh-frequency interactionsoccur inside departments ratherthan across departments.
Effectsoforganizationalculture
on
architecture.Organizations as a
whole
show
patternsofbasic assumptions thatareinvented,discovered, ordeveloped
by
given groups asthey learn tocope withtheirspecific problems.These
problemsinclude, butare notlimited to external adaptationand internal integration [Schein 1985].
Groups
withintheorganization
may
also exhibit theirown
distinctgroup
culture.Wc
observed that theshared experiences, knowledge, and understanding oftheorgani/iitioncaused each group lo bring dillcrcnlassumptions with it tothe larger prcxlucl
dc\ciopmcnt
teamand as aresultaffected architecture decisions.
In one example, the
company
wanted todesign a singleaudio architecture forglobal use.However,
engineering decisions regarding interlacesandconnectionsbecame
ver>'difficult toresolve.
We
laterfound thatthe Europeanengineers regard mcxiulanty (and upgradability)asfundamental loa prcxluct's architecture, whereas the North
American
engineers f(x;usmore on
specific prcxluctand vehicle line features. Interestingly enough, this finding reOects notonlya
dilference in organizational cultures, butappears losuggestthat this difference is
due
todistinctdifferences in the regional customers. This difference
may
affect thefirm's ability tomanage
thediscontinuity
between
present productsandunknown
future productsandhamper
athoroughunderstanding oftheirmarkets
and
customers [Dougherty 1987J.The
managerial implications ofunderstanding a group'sculture includeenhancingmanagement's
abilityto utilizetacit, highly situatedknowledge
ofemployees
suchasmachine
operators, secretaries,orcustomers [Tyre and von Hippcl 1995]. Tyre and
\on
Hippel [1995]suggestthat observingthese kindsof
employees
in their normalwork
environment (i.e. theirsub-groupculture or shared values) allows
managers
todevelopa "contextualized appreciation" ofissues that these
employees
may
face.Lastly, Foster [1986] has
shown
thattechnological change necessitates significantorganizationalchange.
He
states thatcompanies
lose theirleadership notonly becauseofweak
strategies, butalso because of strongcultures.
IV.
Conclusion
This paperdescribesourexplorationofthe linkagesbetween productarchitectureand
organizational design
by
bringing forthspecific examples solidifying the premisethat productarchitectureand organizational designare notindependent. Infact, theyare interdependentand
affecteach otheras
we
have found throughourcase studies. For this study,we
interviewedmembers
ofaudio-system teams intwo
very differentorganizationsofalargeAmerican
automotive manufacturing firm.
We
found that the technical decisions relating todecomposition,architecture
and
integrationare tightlycoupled toboth the capabilitiesand thedesign oftheorganization
which
must executethedevelopment process.We
observedin ourcase studies thatorganizationsdo
simultaneously exhibitmechanisms
ofproductarchitecture affectingorganizational structure as well as patternsof organizational
structure affecting productarchitecture. Hence,
we
assert thatthetechnical architectureof aproduct co-evolves with theorganizational design.
The
notionof co-evolution is adynamic
extension of
Conway's
earlierobservation thatorganizations create technical systemdesignswhich match
thecommunications
structureofthe organization itself[Conway
1968].Decomposition
Product
Architecture
Organization
Design
Figure 8.
Problem
RelationshipsWe
proposea possiblemechanism which
may
help toexplainthis fundamentalcoupling:architecture
and
organization arelinked through theprocess ofproblem decomposition andsystem intcgralion. Dc\clopcrs handle
complex
system design challenges bydccomposinu
thelarge system \n[o simpler(incs
which
can then be designed or specified forouLsourcing. (ThisdecomfX")sition process is repealed until thesubsystemsare simple
enough
to tackle.)At
some
point, the
development
organization'schallenge is tomtc^ratc the various pieces togetherintoacomplete
workmg
system. In fact, decomposition and integration arc generalized inverseproblems.
We
are ot" theopinion thatestablishingand following prcx'edureswhich
take advantageof thebenefitsachie\ed through recognizing thecoupling ofthese decisionscan facilitate the product
development
process. Inappropriateor inllexiblearchitectural or organizationalframeworks
may
becurtailed. For example, ifan organiz.ation has evolved from being tcx)closelyaligned toa very
modular
prtxlucl architecture, an understanding of co-evolutionmay
catalyzeamore
balanced
move
towardsamore
integrated organizational structure. Perhaps amore
effectivemalnx
structure can be formed.Over
time the product architecture willchange atadifferentrate than thedesign of theorganiziition.
Even
though these changes happen slowly, it isessential toacknowledge
thecouplingat thedecision level inordertoachieve thebenefitsofrecognizing their
interdependence. Itdoes notappear feasible for
new
generations ofarchitectural changes to farexceed thatof thepace oforganizational rc-slructurng.
While
we
were
able tomake
stringobservations basedon
ourfield study,many
ofourconclusions arc merely speculative sincethis study representsonly a single classofprcxluct
design. Inorder tostrengthen theconclusions thatmightbe drawn, it
would
be useful toconductstudies ofseveral different prtxlucts toconfirm the robustness ofourfindings. Additionally,an
exploration ofthe following research questions might al.so be infomiativc:
How
can the timeconstantfor theorganizational change smaller thanthetime constantfor architectural change?
What
attributesof product plansand organizational design plans lendthemselves toenhancingstrategic planning?
What
types of organizationscanmore
easilyadoptnew
productarchitectures? Is anorganization
where
themajorskillsand
capabilitiesof the peopleset theparametersof theproductarchitectures
more
successful than oneinwhich
the productarchitectures dictate theorganizational structure?
Can
requirements forhigh-frequencyandlow-frequency
communications
beidentifiedby
aprioriknowledge
ofinterface specifications?To
what
extentdo
collaboration technologieschangethenature oftechnical interactions inproductdevelopment?
Bibliography
Alexander, C. (1964). Notes
on
the SynthesisofForm,
Harvard UniversityPress,Cambridge,
Massachusetts.
Allen, T.J. (1977).
Managing
theFlow
of Technology,MIT
Press, Cambridge,MA.
Allison, G. T. (1971). Essence ofDecision:Explaining the
Cuban
Missile Crisis, Little,Brown,
and
Company,
Boston, Massachusetts.Barczak, G.,and
Wilcmon,
D. (1991)."Communications
Patterns ofNew
FYcxiuctDevelopment
Team
Leaders."IEEE
Transactionson
EngineeringManagement
,38(No. 2).Bijker,
W.
E., Hughes,T. P.,and Pinch, T. P. (1987). The Social Construction of TechnologicalSystems:
New
Directionsin theSociologyand
Historyof Technology,The
MIT
Press,Cambridge,
Massachusetts.Brown,
S. L.,and
Eisenhardt, K.M.
(1995). "PrcxiuctDevelopment: Past Research, Present Findings, and Future Directions."Academy
ofManagement
Review,20(4),343-378.Christensen, C. M.,and
Rosenbloom,
R. S. (1995). "Explainingthe Attacker'sAdvantage:Technological Paradigms, Organizational Dynamics, and the Value Network." Research
Policy(24), 233-257.
Clark, K. B. (1987).
"Knowledge,
Problem-Solving, and Innovation in the Evolutionary Firm: Implications forManagerial Capability and Competitive Interaction."Working
Paper,Harvard BusinessSchcx^l Division of Research, Boston,
MA.
Clark, K. B., and Fujimoto, T. (1991). Product
Development
Perfortnance: Strategy,Organization,
and
Management
intlieWorld
AutoIndustry,Har\ard Business SchoolPress,Boston,
MA.
Conway. M.
E. (1968)."How Do
Committees
Invent?" Datamation, 14(4), 28-31.Da\
idow,W.
H.,and Malone,M.
S. (1992). The Virtual Corporation, HarperBcx^ks,New
York.Dougherty, D. (1987).
"New
Products in Old Organiziitions:The
Myth
o\' the BetterMousetrap
in Search of theBeaten Path," Doctoral Dissertation,
MIT,
Cambridge,MA.
Foster, R. (1986). Innovation: TheAttacker'sAdvantage,
Summit,
New
York.Galbrailh,J. R. (1973). Designing
Complex
Organizations, Addison-Wesley, London.Galbraith,J. R. (1977). Organizational Design,Addison-Wesley, London.
Galbrailh, J. R. (1994).
Competing
with Flexible Ixiteral Organizations, Addison-Wesley,London.
Goldman,
S., Nagel, R., and Prciss, K. (1995). Agile Competitorsand
Virtual Organizations,Van
Nostrand Rcinhold,New
York.Granovetter,
M.
S. (1973)."The
StrengthofWeak
Ties."American
Journal ofSociology,7S(6),1360-1380.
Hameri,A.,and Nihtila, J. (1995). "Distnbuted
New
ProductDevelopment
ProjectBased on
Electronic
Communication
-A
Case
Study." Helsinki University ofTechnology,Finland.Henderson,R. M.,
and
Clark, K. B. (1990). "Architectural Innovations:The
Reconfiguration ofExisting
Systems
and theFailure of Established Firms." AdministrativeScienceQuarterly, 35(9-30).
Kim,
D. H. (1993)."A
Framework
andMethodology
forLinking Individualand
Organizational Learning: Applications inTQM
and Product Development,"Doctoral Dissertation,MIT,
Cambridge,
MA.
Krackhardt, D., and Hanson,J. R. (1993). "Informal Networks:
The
Company
Behind
theChart."
Harvard
Business Review.McCord,
K. R.,and
Eppinger, S. D. (1993)."Managing
the IntegrationProblem
inConcurrentEngineering."
Working
paperno. 3594,MIT
SloanSchool, Cambridge,MA.
Meyer, M., and Utterback, J. (1993).
"The
ProductFamilyand theDynamics
ofCore
Capability."Sloan
Management
/?^v/>vv(Spring),29-47.Morelli,
M.
D., Eppinger, S. D.,and Gulati, R. K. (1995). "PredictingTechnicalCommunications
inProductDevelopment
Organizations."IEEE
Transactionson
Engineering
Management,
42(3), 215-222.Moses,
J. (1995).System
Designand
Management
Semmar
Lecture,MIT,
October 16,1995
unpublished.Ohno,
T. (1988). Toyota Production System:Beyond
Large-ScaleProduction, ProductivityPress, Cambridge,
MA.
Pimmler, T. U.,and Eppinger, S. D. "IntegrationAnalysis ofProductDecompositions."
ASME
Conference
on Design
Theoryand
Methodology
,Minneapolis,MN,
343-351.Pine,J. B. (1993).
Mass
Customization, TlieNew
Frontier inBusiness Competition,HarvardBusiness SchoolPress, Boston.
Rechtin, E. (1991). SystemsArchitecting: Creating
and
Building Cotnplex Systems, PrenticeHall,
Englewood
Cliffs,N.J.Rosenthal, S. R. (1991). EffectiveProductDesign
and
Development:How
toCut
Lead
Time
and
Increase
Customer
Satisfaction.,BusinessOne
Irwin,Homewood,
Illinois.Sanderson, S., and Uzumeri,
M.
(1995)."Managing
productfamiliesThe
case oftheSony
Walkman."
Research
Policy, 24,761-782.Schein, E. H. (1985). Organizational Culture
and
Leadership:A
Dynamic
View, Jossey-BassPublishers, Washington, D.C.
Simon,
H.A. (1990).The
SciencesoftheArtificial,MIT
Press, Cambridge, Massachusetts.Tyre,
M.
J.,and von
Hippel, E. (1995)."The
Situated Nature ofAdapted
Learning inOrganizations."
Working
paper no. 356S,MIT
Sloan School ofManagement
working
pap)cr.
Ulrich, K. (1995).
"The
Role ofProductArchitecture in the ManufacturingFirm." ResearchPolicy, 24{3),419-441.
Ulrich, K., and Eppinger,S. (1995). ProductDesign
and
Development
,Mc-Graw
Hill, Inc.,New
York.
Uzumcn,
M., and Sanderson, S. (1995)."A
framework
formodel
and product family competition." ResearchPolicy, 24, 583-607.von
Hippel, E. (1990)."Task
Partitioning:An
Innovation Process Variable."Research
Policy79,407-418.
Wheclwnght,
S.C,
and Clark, K. B. (1992). RevolutionizingProduct Development,Quantum
Leaps
in Speed, Efficiency,and
Quality,The
Free Press,New
York.3^27
U68
3