Customer Retention Paper Dec 2005

You might also like

Download as doc, pdf, or txt
Download as doc, pdf, or txt
You are on page 1of 38

Sustainable Competitive Advantage Through Customer Retention:

The Roles of Learning and Governance


SHANTANU DUTTA
Marketing Department
Marshall School of Business
University of Southern California
RAMUMAR !ANA"RAMAN
Marketing Department
Marshall School of Business
University of Southern California
#$% !& MA#%R
University of Southern California
Marshall School of Business
Department of Management an' (rgani)ation *BR" +,-.
$os Angeles/ CA 0,,102,1,1
Tel3 *45+. 14525565
7a83 *45+. 96,2+:14
e2mail3 kmayer;marshall&usc&e'u
Decem<er 5-/ 4,,:
Sustainable Competitive Advantage Through Customer Retention:
The Roles of Learning and Governance
ABSTRACT3
Customer retention is an important issue for strategy scholars/ <ut one that has receive' little
attention& =hile organi)ational scholars have e8amine' firm survival as a 'epen'ent varia<le in
a num<er of stu'ies/ the survival of an inter2firm 'ya' has <een neglecte'& =e 'ra> from
agency theory/ transaction cost economics an' the kno>le'ge2<ase' vie> *B?. of the firm to
e8amine the 'eterminants of customer retention& Using a uni@ue panel 'ataset of over a hun're'
customer relationships of a large information technology firm/ >e are a<le to assess ho> the
history of the transaction characteristics <et>een the supplier an' their <uyers impact customer
retention& Specifically/ >e fin' that customer retention is 'riven <y past actions at least as much
as actions that occurre' in the current transaction& By 'isentangling the effects of characteristics
of past an' present transactions <et>een the supplier an' their <uyers/ >e are a<le to offer a
nuance' insight to literature from agency theory/ transaction cost economics/ an' the B?& 7or
instance/ our fin'ing suggests that <uyers in this market are 'iscerning an' punish their suppliers
only >hen they o<serve a pattern of governance misalignment <ut 'isplay for<earance for one2
time mistake& This fin'ing highlights the importance of evolution of relationships on
organi)ational performance that coul' <e incorporate' in future TC% research& Similarly/ the
a<ility to reuse kno>le'ge gaine' from one proAect to other proAects enhances customer
retention/ as firms are utili)ing their learning from prior e8perience& These fin'ings also suggest
that managers nee' to carefully assess the nature of the transaction characteristics >ith their
customers to she' light on future customer <ehavior/ in particular retention&
5
"n high technology in'ustries/ the increase' pace of glo<ali)ation has le' to a rise in the
intensity of competition& As firms 'evelop skills in reverse engineering an' other means of
replicating technological innovations/ the a<ility to retain customers has <ecome increasingly
important for sustaina<le competitive a'vantage *Reichel'/ 500-B Cet)/ Blatt<erg D Thomas/
4,,5.& =hile there is a rich literature that has e8plore' ho> capa<ilities an' kno>le'ge *e&g&/
Argyres/ 500-B Hoetker/ 4,,:B ale/ Dyer D Singh/ 4,,4B Macher/ 4,,:B Henis) D Macher/
4,,6./ learning *e&g&/ Argote/ 5000. an' governance *e&g&/ Nickerson D Silverman/ 4,,+. impact
inter2organi)ational e8changes an' 'ifferent 'imensions of organi)ational performance *ale/
Singh D Eerlmutter/ 4,,,B Anan' D hanna/ 4,,,./ there has <een virtually no
conceptuali)ation an' evi'ence on ho> these factors impact customer retention&
=e 'evelop a conceptual frame>ork that suggests ho> the inter2organi)ational literature
on kno>le'ge/ learning/ an' governance impact the a<ility of a firm to retain its customers&
"nterestingly/ these theories are >ell suite' to she' light on customer retention in high2
technology in'ustries/ <ecause of the nature of transaction <et>een the <uyer an' their ven'ors
in these technology intensive markets& 7irst/ these customer2supplier transactions often entail
reliance on intangi<le kno>le'ge& Secon'/ the comple8ity of these activities often makes it har'
for customers to evaluate the @uality of the output& Thir'/ suppliers often ten' to outsource some
of the >ork to su<contractors& 7inally/ most of these <usiness customers often have repeate'
interactions >ith their suppliers&
This research makes three contri<utions to the literature on strategic management& 7irst/
>e 'isentangle the role playe' <y characteristics of prior transactions <et>een the customer an'
supplier as >ell as the role playe' <y characteristics of the current transaction to 'etermine their
effects on customer retention& Civen the repeate' nature of interactions <et>een suppliers an'
4
customers in these markets it is important to separate the effects of characteristics of the current
transaction from the characteristics of previous transactions& (ur panel 'ata of customers ena<le
us to 'o so& =hile prior >ork on transactional governance *see Shelanski an' lein *500:. for a
revie>. has focuse' on ho> the characteristics of the current transaction affect ho> it >ill <e
governe' *see Argyres an' $ie<eskin' *5000. for an e8ception./ >e sho> that the characteristics
of past transactions/ inclu'ing ho> they are governe'/ also play an important role in customer
retention& "n some cases >e fin' that characteristics of prior transactions play a greater role in
customer retention than the current transaction characteristics as customer concerns accumulate
over time& Customer retention is an important performance metric in <uyer2supplier
relationships an' >e kno> very little a<out it& This stu'y takes one step to>ar's filling this
important gap&
Secon'/ our fin'ings are consistent >ith our conceptuali)ation of the impact that
kno>le'ge <ase' vie> of the firm an' learning have on customer retention& "n particular/ >e fin'
that greater kno>le'ge ac@uire' through e8perience >ith a particular customer as >ell as greater
e8perience in the in'ustry enhance customer retention& 7urther/ >e fin' that >hen the supplier
completes more proAects >ith a specific customer that generates reusa<le kno>le'ge/ it provi'es
the supplier >ith very strong incentives to perform >ell in or'er to a'' to their kno>le'ge <ase
an' thus enhances customer retention&
Thir'/ our fin'ings are consistent >ith our conceptuali)ation of the impact that agency
theory an' transaction cost economics have on customer retention& "n particular/ the customerFs
a<ility to verify the @uality of the proAect also plays a key role in customer retention& 7urther/ >e
e8amine >hen a supplierFs 'ecision to utili)e a su<contractor can hurt customer retention&
=illiamson *509:/ 501:/ 500-. focuses on integration in the presence of contracting ha)ar's in
+
or'er to re'uce governance costs/ <ut >e e8amine circumstances >hen a supplier may hurt the
relationship >ith a customer <y using a su<contractor *i&e&/ effects of governance misalignment
*Mayer D Nickerson/ 4,,:..& Specifically/ >e fin' that >hen the supplier has a history of
outsourcing proAects that are critical to the customer/ then it has a negative impact on customer
retention& "nterestingly/ outsourcing Aust the current transaction 'oes not impact retention& This
fin'ing offers a nuance' vie> of the role of governance on inter2organi)ational relationships& (ur
fin'ing suggests that <uyers in these markets are strategic an' punish suppliers only >hen they
o<serve a cumulative pattern of outsourcing of critical proAects& To our kno>le'ge/ >e are the
first to e8amine the impact of governance choice on customer retention&
=e are a<le to test our hypotheses using a 'ataset of customer2supplier relationships from
the information technology *"T. services in'ustry& (ur 'ata consist of 6,: proAects from a large
"T services firm *hereafter kno>n as Compustar. in Silicon ?alley/ an' inclu'es a variety of
transaction attri<utes an' e8perience measures& The contracts 'ate from 501- to early 5001 an'
thus ena<le us to e8plore separately past as >ell as current transaction characteristics& These 6,:
contracts represent all the "T contracts <et>een Compustar an' 565 customers/ >hich allo>s us
to e8amine the 'rivers of customer retention arising from the characteristics of the transaction
an' the prior relationship <et>een Compustar an' the customer&
The follo>ing section 'iscusses prior >ork on customer retention an' >e then 'evelop
our hypotheses on ho> it is influence' <y learning an' governance& =e then 'escri<e our 'ata/
statistical metho's/ an' varia<les/ >hich are follo>e' <y the results/ 'iscussion an' conclusion&
BACKGRO!" A!" L#T$RATR$ R$%#$&
=hat 'ifferentiates companies in to'ayFs hyper2competitive an' 'eman'2'riven markets
is their a<ility to a''ress their customerFs preferences an' priorities& =ith increase' competition/
6
firms are reali)ing the importance of loyal customers an' a'opting strategies to create an'
sustain a loyal customer <ase *Cet)/ Blatt<erg D Thomas/ 4,,5.& The main o<Aective of the
customer relationship management *CRM. approach is to increase the life time 'uration of
customers <ecause of the un'erlying assumptions that long2term customers are more profita<le&
Reichhel'Fs *500-. stu'y >as one of the first to empirically 'ocument the relationship <et>een
lifetime an' profita<ility& He reporte' a significant increase in profits from small increases in
customer retention rates& 7or e8ample/ he sho>e' that as little as a :G increase in retention ha' a
significant impact on the net present value of the firm ranging from 0:G in the case of
a'vertising agencies to +:G in computer soft>are in'ustry&
%8tant literature in marketing has looke' into the various aspects of customer
relationship& =hile some stu'ies are concerne' >ith formulating metho's to mo'el customer
retention *Schmittlein/ Morrison D Colom<o/ 5019B Schmittlein an' Eeterson/ 5006B Reinart)
an' umar/ 4,,+./ other stu'ies are intereste' in stu'ying the 'rivers of customer retention such
as satisfaction *Bolton/ 5001. an' competitorsF offerings *Rust/ $emon D Heithaml/ 4,,6.&
Although these stu'ies help us un'erstan' the factors that influence customer retention/
these stu'ies have <een in <usiness to consumer settings *B4C. an' are focuse' on factors that
are most applica<le to customer retention an' management in a B4C setting& These stu'ies 'o not
account for important ven'orIcustomer transaction characteristics that coul' influence customer
retention& "n many B4B markets the same ven'ors an' customers ten' to >ork together
repeate'ly over time/ hence it is important to capture the characteristics of <oth the current
transaction as >ell as prior transactions& "n this stu'y/ >e focus on the characteristics of prior
transactions is important as it helps us un'erstan' ho> firms learn from their overall in'ustry
e8perience an' their e8perience >ith specific customers to offer <etter pro'ucts an' services&
:
7urther/ our stu'y also provi'es a sharper focus on ho> o<serva<ility of proAect performance an'
governance of previous proAects influences customer retention&
(ur frame>ork thus she's insight on the literature at the intersection of kno>le'ge *i&e&/
capa<ilities an' learning. an' governance& Buil'ing on e8aminations of alliance capa<ilities
*Anan' D hanna/ 4,,,B ale/ Dyer D Singh/ 4,,4. an' contracting capa<ilities *Argyres D
Mayer/ 4,,6./ this stu'y 'evelops a frame>ork to assess customer retention is also an aspect of
inter2firm governance in >hich firms can 'evelop capa<ilities&
TH%(R# AND H#E(TH%S%S
The Role of Observabilit' and Governance
Many of the transaction attri<utes that transaction cost research has sho>n to influence
vertical integration an' contract 'esign 'ecisions may also influence customer retention& (ne
transaction attri<ute that shoul' influence customer retention is the 'ifficulty of 'etermining the
@uality of the output of the proAect& Customers are more likely to return if the supplier performs
>ell on the proAect& =hen the @uality of output can <e easily measure'/ it creates strong
incentives for the supplier to perform >ell <ecause the customer >ill promptly o<serve any
failings& Measurement costs create a moral ha)ar' <y generating noise in the relationship
<et>een effort an' outcome *Holmstrom/ 5090.& "f any key 'imension of the output of the task
is 'ifficult to measure/ the supplier may have an incentive to shirk <ecause any failings in the
final pro'uct >ill <e 'ifficult to o<serve *An'erson/ 501:.& "n such cases the customer >ill
likely un'ertake a''itional monitoring of the supplier to compensate for the >eak outcome2<ase'
incentives *Holmstrom D Milgrom/ 5005B Mayer/ Nickerson D (>an/ 4,,6.& %ven >ith this
-
a''itional monitoring/ ho>ever/ the supplier >ill have some uncertainty a<out >hether the Ao<
has <een complete' correctly <ecause the @uality of the final pro'uct is 'ifficult to o<serve&
Thus proAects that result in output that is 'ifficult to measure are likely to result in
customers that are not completely sure ho> >ell the supplier has performe'& This shoul' result
in a lo>er likelihoo' of repeat <usiness compare' to proAects >ith clearly measura<le
performance criteriaJ<ecause such proAects provi'e stronger incentives for the supplier an'
allo> the customer to feel more secure >ith the outcome& %ven if the greater o<serva<ility
makes it clear that a mistake has <een ma'e/ the supplier can make the necessary corrections
until the customer is satisfie' >ith the result&
Hypothesis 53 Customer retention >ill 'ecrease >hen the output of the current proAect is
'ifficult to measure&
%ven if the customer is satisfie' >ith the proAect upon completion/ the 'ifficulty of
o<serving @uality in'icates that pro<lems may arise in the future& The pro<lem is compoun'e'
>hen the supplier completes multiple proAects >hose output is 'ifficult to measure for the
customer& The supplier kno>s that shirking on @uality is highly unlikely to <e 'iscovere'
imme'iately an' even if it is 'etecte' in the future/ there is often am<iguity over ho> the
pro<lem arose& Civen these >eakene' incentives/ the supplier may <e more likely to shirk the
more they continue to >ork >ith the customer& %ven if the supplier >ants to maintain a close
relationship >ith the customer/ their employees/ >ho are actually 'oing the >ork/ may <e a<le to
shave @uality to get the proAect complete' more @uickly an'Kor at lo>er cost& Thus measurement
costs create a pro<lem not Aust for the current proAect/ <ut represent an ongoing cumulative
pro<lem as more proAects >ith potentially latent pro<lems are complete' for the customer&
Hypothesis 43 The greater the num<er of prior proAects that a supplier has complete' >ith
a specific customer >here output is 'ifficult to measure/ the less likely the customer is to
return&
9
A customerFs perception of the supplier is 'riven <y more than Aust the a<ility to verify
the @uality of output& Transaction cost economics *=illiamson/ 501:/ 500-. hol's that using
employees is preferre' in situations >hen contracting pro<lems such as asset specificity are
present& There has <een some research on the governance choice2performance relationship *e&g&/
Masten/ Meehan D Sny'er 5005B =alker D Eoppo 5005B Silverman/ Nickerson D 7reeman
5009B Eoppo D Henger 4,,4.& Using market2<ase' governance/ such as su<contracting/ can hurt
performance if contractual ha)ar's are present *Mayer D Nickerson/ 4,,:B Nickerson D
Silverman/ 4,,+.& =hile >e 'ra> upon transaction cost economics to e8amine the role of
governance in customer retention/ our focus on customer retention causes us to focus on an
e8planatory varia<le that is very germane in the conte8t of <uyer2supplier relationships in these
in'ustries& "n particular/ >e assess ho> the nature of governance of pro'uctsKservices that are
critical to the <usiness of the customer impact customer retention&
5

Some proAects have limite' visi<ility an' mistakes that arise can <e fi8e' <efore they
cause serious pro<lems for the customer& Conversely/ pro<lems >ith other proAects can have an
imme'iate an' costly impact on the customer& "f a representative of the supplier *either a
su<contractor or employee. acts opportunistically or incompetently on a proAect >ith little
potential to harm the customer/ then the results >oul' <e em<arrassing for the supplier <ut the
customer might <e more forgiving since there >as no maAor harm 'one& "f the same type of
mistakes occur on a mission critical proAect that is very important to the customer/ ho>ever/ the
'amage to the supplierFs relationship >ith the customer is likely to <e much more severe& =hen
a customerFs critical applications or systems are involve' in a proAect/ suppliers may prefer to use
their o>n employees rather than su<contract the proAect to an outsi'e company&
5
=e certainly ackno>le'ge that asset specificity is important in governance/ <ut >e <elieve that since our focus is
on retention of <usiness customers/ governance of pro'uctsKservices critical to their <usiness may <e even more
important to these customers/ in mo'erating the relationship <et>een governance an' customer retentionJa
relationship not e8amine' to 'ate <y transaction cost researchers&
1
=hen a supplier >ins the <i' for a proAect >ith a customer/ the supplier must 'etermine
ho> to organi)e an' complete the proAect& The supplier can use employees or turn to a
su<contractor& EroAects that involve critical customer applications shoul' <e complete' >ith
employees for t>o key reasons that involve customer perception& 7irst/ the customer may
interpret the use of a su<contractor as a sign that the supplier 'oes not have any employees that
are @ualifie' to complete the task an' thus may 'ou<t the value a''e' <y the supplier& Secon'/
the customer may also feel that su<contracting their critical proAects is a sign that the supplier
may not take the customerFs important proAect very seriously&
"n a''ition to the customer perception issues/ there are t>o more reasons >hy a supplier
might >ant to avoi' su<contracting a customerFs critical proAect that relate to the likelihoo' of
pro<lems arising 'uring the proAect& 7irst/ su<contractors are unlikely to care as much a<out
future <usiness opportunities >ith the customer as the supplierFs employees/ >hose entire income
is linke' to the supplier& Secon'/ suppliers are <etter a<le to verify the training of employees an'
thus the risk that the >orker >ill <e incompetent is lo>er& "t can <e more 'ifficult to verify the
skills that a su<contractor has on his or her resume/ relative to its o>n employees/ as the firm
'irects the training of its employees an' has more e8perience >ith them than >ith outsi'e
su<contractors& Thus firms that try to use su<contractors >hen the proAect is critical for the
customer are more likely to encounter costly mistakes an' thus less likely to have a satisfie'
customer that >ants to continue >orking >ith the supplier in the future& %mployees are more
relia<le an' more likely to lea' to <etter proAect outcomes >hen contracting 'ifficulties cause'
<y mission critical proAects are present&
Hypothesis +3 =hen the supplier utili)es a su<contractor to complete a current proAect
that is mission critical for the customer/ the customer is less likely to return&
0
Hypothesis + assumes that customers have a clear an' imme'iate negative reaction to
su<contracting the current proAect an' 'onFt give the supplier another chance& Not all customer
critical proAects that are su<contracte'/ ho>ever/ >ill lea' to maAor pro<lems 'uring the proAect&
Customers may <e >illing to overlook some issues as the relationship 'evelops&
=hile su<contracting a current customer critical proAect is likely to have a negative effect
on customer retention/ there is also likely to <e a cumulative effect& "t may <e that customers
'onFt imme'iately leave if the supplier uses a su<contractor for one of their critical proAects/ <ut
instea' customers leave if they see a pattern of this type of activity <y the supplier& This is also
consistent >ith the foun'ations of transaction cost economics *e&g&/ =illiamson/ 501:./ >hich
asserts that <a' governance 'ecisions >ill result in cumulative effects that >ill harm the firm&
=hen critical proAects are su<contracte' <y the supplier over time the customers <ecomes
increasingly likely to lose confi'ence in the supplier/ lea'ing to lo>er retention& Customers may
not react if it only occurs onceJthere is no knee Aerk reactionJ<ut >ait to see if this activity is
an anomaly or a pattern& (utsourcing one critical proAect may <e for a goo' reason/ <ut a history
of it lea's the customer to @uestion the supplierFs internal competence *the in'epen'ent value the
supplier a''s. an' the supplierFs priorities& The customer may <elieve that continual
su<contracting of mission critical proAects in'icates that the supplier 'oes not 'eem the
customerFs proAect of sufficient importance to use their employees *i&e&/ the supplier is not paying
enough attention to the customer. 'espite the importance of these proAects to the customer&
The mere fact that a proAect is critical to the supplierFs <usiness is not likely to influence
retention/ <ut a series of critical proAects that are outsource' over time has a negative impact on
the customerFs perception of the supplierFs performance& This is likely to <e especially true in
B4B markets >here the customers are also kno>le'gea<le <usinesses&
5,
Hypothesis 63 The more of a particular customerFs prior mission critical proAects that the
supplier has su<contracte'/ the less likely the customer is to return&
The Role of Kno(ledge and Learning
The kno>le'ge2<ase' vie> of the firm *B?. *e&g&/ ogut D Han'er/ 5004. has focuse'
attention on kno>le'ge consi'erations/ >hich can also affect customer retention& =hile the
B? has e8amine' the effect of kno>le'ge consi'erations on organi)ational form *e&g&/ Argyres/
500-B Nickerson D Silverman/ 4,,+. an' competitive a'vantage *e&g&/ 7oss/ 500-B ogut D
Han'er/ 5004./ its effect on customer retention has yet to <e e8plore'& Some proAects have little
kno>le'ge impact on the firmJthe supplier completes one proAect an' moves on to the ne8t&
Some proAects/ ho>ever/ can influence the cost of future proAects <ecause the kno>le'ge
generate' to complete the first proAect can <e reuse' for future proAects& EroAects that may create
potentially reusa<le kno>le'ge may increase customer retention for t>o reasons&
7irst/ >hen the proAect represents a potential key a''ition to the supplierFs kno>le'ge
<ase/ then the supplier >ill <e highly motivate' to complete the proAect >ith high @uality& The
supplier has little incentive to shirk on @uality or effort as a successful proAect <enefits not only
the customer/ <ut a''s to the supplierFs kno>le'ge <ase as >ell& Shirking on @uality >oul' hurt
not only the customer <ut also the supplier <ecause the kno>le'ge create' >oul' <e of less
value& Thus the potential for kno>le'ge reuse effectively aligns the incentives of the customer
an' supplier/ >hich shoul' increase customer satisfaction an' retention&
A secon' factor is that some customers are more likely than others to >ork >ith suppliers
on proAects that may generate reusa<le kno>le'ge& Some customers are closer to the
technological frontier than others& Many of the proAects that can pro'uce potentially reusa<le
kno>le'ge involve customers >ith cutting e'ge re@uirements& Such customers are likely to have
55
future proAects that >ill also pro'uce reusa<le kno>le'ge/ so the supplier has an a''e' incentive
to perform >ell in or'er to earn future <usiness from such valua<le customers& Both factors
provi'e the supplier >ith strong incentives to perform >ell on proAects that may pro'uce reusa<le
kno>le'ge&
Hypothesis :3 =hen the current proAect is likely to generate kno>le'ge that the supplier
can reuse/ customer retention shoul' increase&
EroAects >ith potentially reusa<le kno>le'ge shoul' have a lasting effect on customer
retention& Customer retention shoul' increase as the supplier 'oes more proAects >ith the
potential to create reusa<le kno>le'ge for a particular customer& Civen the supplierFs strong
incentives to perform >ell in such situations/ customer satisfaction shoul' <e high each time
such a proAect is complete'& As the supplier 'oes more proAects >ith such strong incentives for
the customer/ the customer is likely to reap the repeate' <enefits of very positive supplier
performance& The more times a customer finishes an interaction >ith a supplier feeling very
satisfie' >ith the supplierFs performance/ the more likely the <uyer >ill return to the supplier 'ue
to the continue' pattern of superior performance&
A secon'/ an' perhaps more important/ effect is that the more times a proAect involves
potentially reusa<le kno>le'ge/ the more e8perience the supplier gets in learning ho> to
incorporate such kno>le'ge an' use it to effectively meet the nee's of this an' other customers&
%8perience is a po>erful 'river of learning in organi)ations *e&g&/ Argote/ 5000.& =hile reusa<le
kno>le'ge can <e <roa'ly applie' across many of the supplierFs customers/ it is clearly
applica<le to future proAects from the same customer& Thus the supplier <enefits from learning
ho> <est to re'eploy the kno>le'ge more generally across the customer <ase/ an' from a higher
likelihoo' of <eing a<le to reuse the kno>le'ge again >ith the same customer&
54
Hypothesis -3 The more prior proAects the supplier has complete' for the customer
involving potentially reusa<le kno>le'ge/ the greater the likelihoo' of retaining the
customer&
=hile potentially reusa<le kno>le'ge is important/ a secon' aspect of kno>le'ge an'
learning is a supplierFs e8perience >ith a particular customer& The longer a supplier >orks >ith a
particular customer in a given line of <usiness/ the more the supplier learns ho> to >ork more
effectively >ith that customer an' the more the customer may come to trust the supplier& There
are four 'istinct reasons >hy e8perience >ith a customer is likely to increase customer retention&
7irst/ mistakes are more common on early proAects <ecause they parties may have
misaligne' e8pectations or they may make honest mistakes relate' to their ine8perience at
>orking together *Mayer D Argyres/ 4,,6.& (nce the relationship survives these uncertainties
associate' >ith learning a<out one another/ the likelihoo' of the customer terminating the
relationship shoul' 'ecrease& Secon'/ >hen a relationship <et>een t>o firms is 'eveloping/
mistakes may <e more 'ifficult to interpret& 7or e8ample/ suppose that the supplier <elieves the
customer >ants one thing/ <ut the customer really >ante' something else that is more comple8
an' costly& The customer may not <e sure if the supplier simply ma'e a mistake or is trying to
shirk&
Thir'/ e8perience >orking together may in'icate that trust is 'eveloping *Culati/ 500:.
an' trust <et>een firms shoul' <e positively relate' to customer retention *Cannon D Eerreault/
5000B Ceyskens et al/ 500-.& 7ourth/ from a transaction cost perspective/ the more fre@uently
that a supplier an' customer >ork together/ the more likely that they >ill invest in relationship2
specific routines& These routines may ena<le the firms to an' interact more effectively& Thus the
customer >oul' <e more likely to return to take a'vantage of such investments&
Hypothesis 93 Customer retention >ill improve as the firm gains more e8perience
>orking >ith a particular customer in the current line of <usiness&
5+
$)*#R#CAL A!AL+S#S
"ata and Conte,t
=e teste' the hypotheses >ith 'ata from Compustar/ a provi'er of a variety of
information technology *"T. services an' computer2relate' har'>are& The "T in'ustry is i'eal for
testing the hypotheses& "t is large an' important to customers in virtually all sectors of the >orl'
economy an' kno>le'ge is a key source of competitive a'vantage for "T firms& The "T in'ustry
involves the storage/ transfer/ an' management of information/ typically using mainframes/
servers/ or relate' 'evices& "t inclu'es three tiers3 customers *e&g&/ 7ortune :,, firms./ "T
suppliers *e&g&/ "BM/ 7uAitsu/ CSC./ an' smaller su<contractors& The suppliers perform a variety
of "T proAects for their customers inclu'ing 'esigning customi)e' soft>are systems/ up'ating an'
maintaining e8isting soft>are or har'>are systems/ an' assisting >ith net>ork 'esign an'
security& The technological areas in this in'ustry inclu'e "BM2compati<le mainframes/ (SK+0,
programming/ Sun servers/ 'ata<ases *e&g&/ (racle/ "nformi8./ customi)e' soft>are support/ an'
'evelopment in a variety of languages& Capa<ilities in this in'ustry are technology2specific
rather than firm2specific/ >hich means that su<contracting options are almost al>ays availa<le&
=ork is performe' pre'ominantly on a proAect <asis& Customers i'entify an "T proAect
an' then secure resources to complete it& %ach proAect is source' separately& A customer may
engage "BM for one proAect an' CSC for another/ an' these suppliers must 'eci'e ho> to fulfill
the proAect re@uirements& The sourcing 'ecision is particularly important <ecause mem<ers of
the proAect team >ill likely interact 'irectly >ith the customer as the >ork is typically performe'
at the customerFs site& "n many cases/ in'epen'ent su<contractors are utili)e' to fulfill proAects
for customers&
56
Compustar/ a pro'ucer of mainframes an' relate' har'>are since the 509,s/ entere' the
platform2in'epen'ent "T services <usiness
4
in the mi'2501,s/ an' <y 5009/ its "T services
'ivision accounte' for revenues of appro8imately L5,, million >orl'>i'e& This gro>th >as
accomplishe' through the 'evelopment of an internal 'elivery force an' the use of a variety of
su<contractors& Compustar >as an i'eal firm for testing our hypotheses <ecause of its e8tensive
presence in several sectors of the "T in'ustry/ the significant value it place' on its kno>le'ge
<ase as a source of competitive a'vantage an' the variety of its kno>le'ge <ase& These
characteristics are typical of large "T firms such as "BM/ CSC/ %DS/ etc&
Compustar provi'e' access to all "T service contracts in its corporate contracts li<rary&
The contracts 'ate from 501- to early 5001& "n this stu'y/ >e analy)e a sample of 6,: of
CompustarFs "T contracts >ith North American customers& The contracts >ere 'ra>n from a
ran'om sample of the contracts in CompustarFs contracts li<rary& The sample >as selecte' <ase'
on the first letter of the customerFs name to generate an un<iase' sample that >oul' reflect the
entire population of proAects& The proAects in this sample consist of all contracts <et>een
Compustar an' 564 customers& (ur sample contains appro8imately 4:G of the entire population
of Compustar "T services contracts& A revie> <y Compustar personnel in'icate' that our sample
>as representative of the full population of contracts in terms of customer in'ustries represente'/
si)e of the customer firm/ num<er of contracts <et>een Compustar an' the firm/ etc&
"n a''ition to rea'ing the contracts/ >e intervie>e' several Compustar managers/
engineers/ an' "T personnel from outsi'e Compustar& The 'ata >ere 'ra>n primarily from the
contracts <et>een Compustar an' its customers/ su<contractor invoices/ an' other recor's
inclu'e' in the contract file& Compustar personnel co'e' any varia<le that re@uire' su<Aective
4
Elatform2in'epen'ent means that the firm supplies services to firms using a variety of 'ifferent types of har'>are&
These services inclu'e' net>ork support/ programming/ 'ata migration/ etc&
5:
Au'gment& The contract contains a 'etaile' 'escription of the proAect inclu'ing the type of
service re@uire' an' the responsi<ilities of the parties& "t is typically a<out five pages long an' is
'esigne' to accomplish a specific task for the customer&
%ariables
The 'epen'ent varia<le/ T%RM"NAT"(N/ is co'e' as one if the customer 'oes not return
to Compustar *i&e&/ if it terminates the relationship.& The main 'ecision in co'ing this varia<le
>as ho> to co'e the last proAect in the sample <et>een Compustar in each customer& Some
o<servations >ill <e right censore'Ji&e&/ the customer is still active >hen the sample en'sJ
>hile others >ill have terminate' the relationship prior to 5001& =e use' a variety of techni@ues
*see the Ro<ustness Tests section <elo>. an' finally settle' on the follo>ing proce'ure& =e
create' an average time <et>een proAects for each customer that ha' multiple proAects& =hen >e
got to the last proAect for each customer/ >e a''e' t>o stan'ar' 'eviations a<ove that average to
'etermine >hen >e coul' confi'ently say the customer >as not coming <ack& "f that 'ate *the
last proAect plus the mean time <et>een proAects plus t>o stan'ar' 'eviations. >as <efore the
en' of the sample perio'/ then >e co'e' the customer as terminate'B if that 'ate >as after the
en' of the sample perio'/ then the customer >as co'e' as right censore'& 7or customers that ha'
only single proAect >ith Compustar/ >e use' t>o stan'ar' 'eviations a<ove the entire sample
average to make the same 'etermination& (ur results are ro<ust to 'ifferent >ays of measuring
termination/ inclu'ing fi8e' perio's such as a year *see Ro<ustness Tests <elo>.&
The a<ility to o<serve @uality is the su<Aect of Hypotheses 5 an' 4& The varia<le
CURR%NT M%ASUR%M%NT C(ST captures >hether the technology employe' in the current
proAect makes it 'ifficult to 'etermine the @uality of the output generate' <y the proAect team& "t
5-
is 'esigne' to capture the cost of measuring @uality e8 post <ase' solely on the technological
nature of the proAect& Compustar stipulate' that the measurement cost varia<le <e co'e' <y its
engineers as a 'ummy varia<leB one if @uality is 'ifficult to 'etermine an' )ero if it is rea'ily
apparent& The @uestion that 'etermine' the value of this varia<le >as >hether a <rief/
ine8pensive test or inspection coul' 'etermine the @uality of the >ork 'one on the proAect&
Hypothesis 4 e8amines the effect of prior proAects >ith high measurement costs& ER"(R
M%ASUR%M%NT C(ST is a count of the num<er of prior proAects/ not inclu'ing the current
one/ >hich Compustar has complete' >ith this customer that ha' high measurement costs&
Testing Hypotheses + an' 6 re@uires an interaction varia<le to e8amine the effect of
customer critical proAects that are su<contracte'& CURR%NT CR"T"CA$ SUBC(NTRACT is a
'ummy varia<le that is one if the current proAect involves a customer critical proAect that is
su<contracte' <y Compustar& Customer critical proAects are proAects that have the potential to
cause a Msignificant portionN of a customerFs 'ata center to shut 'o>n& Such 'isruptions are
e8tremely costly an' 'isruptive to the customer& Compustar engineers co'e' this varia<le <ase'
on their e8pertise an' the 'escription of the proAect in the contract& The recor's in the contract
file in'icate >hether a su<contractor >as use' on the proAect& Hypothesis 6 e8amines the effect
of prior critical proAects that >ere su<contracte'& ER"(R CR"T"CA$ SUBC(NTRACT is a
count of the num<er of prior customer critical proAects that Compustar has su<contracte' >ith
this customer& The current proAect is not inclu'e' in this count&
The potential to create reusa<le kno>le'ge *Hypotheses :2-. is capture' <y CURR%NT
R%US%/ >hich is a 'ichotomous varia<le that is one if the proAect is likely to pro'uce reusa<le
kno>le'ge an' )ero other>ise& Compustar provi'e' t>o engineers to help >ith co'ing varia<les
that coul' not <e 'irectly co'e' from the contracts& The engineers co'e' this varia<le an'
59
several others <ase' upon their e8pertise an' recor's in the contract file& R%US% >as co'e'
looking only at the 'escription of the proAect in the contract& The engineers 'i' not investigate to
see >hat happene' 'uring the e8ecution of the proAect& Hypothesis - e8amines the effect of prior
proAects >ith potentially reusa<le kno>le'ge& ER"(R R%US% is a count of the num<er of prior
proAects that Compustar has complete' >ith this customer that ha' potentially reusa<le
kno>le'ge& The current proAect is not inclu'e' in this count&
Hypothesis 9 e8amines customer2specific e8perience an' learning& "ntervie>s in'icate'
that the 'evelopment of a customer relationship can <e measure' <y the fre@uency of interaction
<et>een Compustar an' the customer& $ength of interaction >as reAecte' as a pro8y for this
effect <ecause a customer may utili)e Compustar once an' then not use them again for several
years& Thus several proAects un'ertaken over a three year span in'icate a much more active an'
>ell2'evelope' relationship than fe>er proAects over a longer perio'& ER"(R "T ER(!%CTS is
the num<er of previous "T proAects Compustar has performe' for each customer& =e co'e' this
varia<le <y e8amining the full contracting file for each customer&
To isolate the effect of customer critical proAects that are su<contracte'/ >e inclu'e
customer critical proAects/ regar'less of >hether they are su<contracte'/ in the ha)ar' mo'el to
ensure that the relevant effect is su<contracte' critical proAects an' not Aust critical proAects&
CURR%NT CR"T"CA$ is co'e' as one if the current proAect is critical to the customer an' )ero
other>ise& ER"(R CR"T"CA$ is a count of the num<er of prior "T proAects that Compustar has
'one for the customer that >ere critical to the customer& The current proAect is not inclu'e' in
this count&
There are other factors that may influence customer retention& 7or e8ample/ the <rea'th
of <usiness that the customer 'oes >ith the supplier may <e an in'icator of satisfaction >ith the
51
supplier an' it affor's the supplier the opportunity to learn more a<out the customer& The
supplier can share kno>le'ge among 'ivisions in or'er to provi'e <etter overall service to the
customer& BR%ADTH is the num<er of the supplierFs pro'uct lines/ outsi'e of information
technology services/ that the customer <uys from Compustar& =e e8pect a <roa'er relationship
<et>een the firms to increase customer retention&
Another set of factors that can affect customer retention are the supplierFs technical
capa<ilities *Argyres/ 500-B Eeteraf/ 500+.& =hen a proAect 'ra>s upon a technological area in
>hich the supplier has capa<ilities that are superior to competitors/ customer retention may
increase <ecause the pro<a<ility of a pro<lem on the proAect shoul' <e lo>er <ecause competence
pro<lems are less likely to occur& Compustar engineers are ackno>le'ge' e8perts at servicing
mainframes from other ven'ors 'ue to their e8perience an' training in all aspects of mainframe
technology& Compustar has 'esigne' an' manufacture' mainframes since the 509,s an' has
'evelope' very strong capa<ilities in this areaJmainframes are CompustarFs primary pro'uct&
MA"N7RAM% is co'e' as one if the contract involves >orking on a mainframe computer an'
)ero other>ise& %8ternal in'ustry e8perts verifie' CompustarFs capa<ilities in this area& "n
a''ition/ this measure is in line >ith e8perience2<ase' measures of capa<ilities/ <ecause
Compustar so many years of mainframe e8perience *an' less e8perience in areas such as
programming.&
=hen a proAect 'ra>s upon a technological area in >hich the supplier has capa<ilities
that are >eaker than competitors/ customer retention may 'ecrease <ecause of the increase' risk
that the supplier >ill make a mistake <ecause they lack superior capa<ilities& There are many
other firms that possess programming capa<ilities/ >hich have not <een part of CompustarFs
focus over the past t>enty years& Erogramming capa<ilities are highly speciali)e' an' comple8/
50
<ut are technology2specific *e&g&/ (racle 'ata<ases/ Uni8/ (SK+0,. rather than firm specific&
ER(CRAMM"NC is a 'ummy varia<le that is one if the proAect primarily involves
programming an' )ero other>ise& CompustarFs capa<ilities in >orking on these proAects are not
superior to those of several other firms in the in'ustry&
%8perience in the in'ustry may also affect customer retention& Compustar may get <etter
at customer retention over time& T"M%TR%ND is a linear time tren' that is co'e' )ero for 501-/
one for 5019/ an' so on up to t>elve for 5001 an' is inclu'e' in all mo'els&
The co'ing process >as as follo>s& %ach engineer co'e' the same eighty ran'omly
selecte' contracts& Then the t>o engineers an' one of the authors >ent through all eighty an' >e
foun' the follo>ing 'iscrepancies3 three for CR"T"CA$ SUBC(NTRACT/ three for
M%ASUR%M%NT C(ST/ t>o for ER(CRAMM"NC an' t>o for R%US%& After a <rief
'iscussion/ the engineers clarifie' the 'iscrepancies an' agree' on the i'entical classification
metho'ologies& The t>o engineers then co'e' the remaining contracts&
Ta<le 5 contains 'escriptive statistics for all varia<les& Correlations are generally lo> to
mo'erate/ >hich suggests that multicollinearity is not a pro<lem for this estimation&
2222222222222222222222222222222
"nsert Ta<le 5 a<out here
2222222222222222222222222222222
$stimation
As our 'epen'ent varia<le is e8it from a population/ in this case the population of
Compustar customers/ >e use a ha)ar' rate mo'el in our estimation& "n the ha)ar' function
frame>ork/ the 'ynamics of 'uration time can <e conveniently capture' <y the ha)ar' rate/ hi*t.
>hich is 'efine' as the customer iFs pro<a<ility of terminating the relationship 'uring a particular
interval/ given that the customer has continue' the relationship until time t3
4,
. 5 &&&*
t
tO until ip relationsh the continue' has customer the
given t. t *t/ interval time in the ip relationsh the s terminate i Pcustomer Er
$im . *
,

+
=
t
i
t h
The ha)ar' function can also <e 'efine' in terms of the cumulative function 7*t. an' the
pro<a<ility 'ensity function f*t.B
. 4 &&&*
. * 5
. *
. *
t F
t f
t h

=
Note that the ha)ar' is the con'itional likelihoo' that a customer >ill 'efect at time t/ given that
she has not 'efecte' in the 'uration interval *,/t.& Co8 *5094. propose the proportional ha)ar's
regression *henceforth/ EHR. techni@ue to estimate the ha)ar' mo'el&
+
The ha)ar' rate takes the
form
. + .&&&* 8 * . * . Q *
t ,
t h x t h
i
=
>here h,*t./ the <ase ha)ar' function/ represents the longitu'inal effect of time& *8t. represents
the effect of covariates an' changes h,*t. up or 'o>n proportionately to reflect the effect of
covariates& The Co8 proportional ha)ar' mo'el assumes that *8. is an e8ponential function of
the covariates/ i&e&/
. 6 .&&&* R e8p* . e8p* . *
5
x x x
p
j
j j
= =

=

an' the ha)ar' function is
>here S*5/ 5/T&& p. 'enotes the parameters associate' >ith the covariates& The a<ove mo'el is
terme' proportional ha)ar's regression *EHR.& Transformation of the pre'ictor varia<les to 5,,8
+
The use of the Co8 ha)ar' mo'el re@uires that the proportionality assumption hol' true/ >hich it 'oes for our 'ata&
. : .&&&* e8p* . * . Q *
5
,
=
=
p
j
j j i
x t h x t h
45
*e8p *A.25. helps us to interpret the coefficients associate' >ith the covariates& This represents
the percentage change in the ha)ar' rate for a unit change in the correspon'ing covariate/ 8A&
Co8 *5094. presente' the semi2parametric partial likelihoo' function that is in'epen'ent
of the <aseline ha)ar' function to estimate the parameters& Suppose that a customer i terminate'
the relationship at time t& Hence/ the customerFs uncensore' 'uration time is t& At this 'uration
time/ t/ a num<er of other customers >ere Mat riskN i&e& the other customers that ha' not
terminate' the relationship >ith the supplier& These customers <elong to the Mrisk setN& (f all
those customers at risk *of terminating the relationship. at time t/ it is the customer i >ho actually
terminate' the relationship& Thus/ if a customer i 'efecte' at time t/ the partial likelihoo'
function that the customer i 'efecte' is
. - &&&*
. *
. &&& / / Q *
. *
5
. *
. * 4 5
t h
h
j j j t i L
t n
k
j
t i
t n
k

=
=
/
>here n*t. is the num<er of customers at risk at t/ an' these customers are 'enote' A5T An*t.&
Su<stituting the proportional ha)ar' mo'el/ the longitu'inal effect of h,*t. cancels out leaving
. 9 &&&*
. *
5
R
R
.
. *
&&&
4
/
5
/ Q *

=
=
t n
k
x
e
it
x
e
t n
j j j t i L
t j
k

Ma8imi)ing the a<ove likelihoo' yiel's the partial estimate of & East >ork has sho>n that
ma8imi)ing the a<ove likelihoo' gives efficient *%fron 5099. an' consistent *Tsiatis 5015.
estimates of &
Note that EHR mo'els the likelihoo' that a customer terminates the relationship given
that the customer has not yet terminate'& As a result/ the signs of the ha)ar' mo'el coefficients
44
>ill <e the opposite to the signs of 'uration mo'el coefficients& "n other >or's/ if U,/ then it
increases the ha)ar' an' thus the likelihoo' of a customer 'efecting&
Results
=e <egin >ith Mo'el 5/ >hich inclu'es only our control varia<les in Ta<le 4& The
control varia<les yiel' some interesting insights& BR%ADTH is negative an' significant *p V
,&,,5./ >hich in'icates that the more pro'uct lines the customer <uys from Compustar/ the less
likely the customer is to terminate the relationship *i&e&/ customer retention increases.&
"nterestingly/ neither measure of CompustarFs technical capa<ilities has a significant impact on
customer retentionB nor 'oes customer retention vary <ase' on >hether the proAect >as critical
for the customer& CompustarFs overall e8perience in the "T services in'ustry/ T"M%TR%ND/ has
a negative an' significant *p V ,&,,5. relationship on the likelihoo' of termination&
2222222222222222222222222222222
"nsert Ta<le 4 a<out here
2222222222222222222222222222222
Mo'el 4 in Ta<le 4 a''s our hypothesi)e' varia<les to Mo'el 5& The likelihoo'
improvement <et>een the tra'itional mo'el an' the propose' mo'el is statistically significant at
the 5G level& The log likelihoo' for Mo'els 5 an' 4 are 25+69&1 an' 25++4&59 respectively& The
=al' test reAects Mo'el 5 in favor of Mo'el 4 *the calculate'
4
S +5&4- >hile the critical
4
*54/
,&,5. S 4-&44.& There are no significant changes to the control varia<les in Mo'el 4& There is
some support for Hypothesis 5 as proAects >here it is 'ifficult to 'etermine @uality *CURR%NT
M%ASUR%M%NT C(ST. increase the pro<a<ility of customer termination *i&e&/ 'ecreases
customer retention. *p V ,&5,.& Erior proAects >here it >as har' to 'etermine @uality of the
pro'uct provi'e' <y Compustar also 'ecrease customer retention *p V ,&,5./ provi'ing strong
support for Hypothesis 4& No support is foun' for Hypothesis +/ as governance misalignment
4+
*CURR%NT CR"T"CA$ SUBC(NTRACT. has no significant effect& Hypothesis 6/ ho>ever
receives strong support as the cumulative num<er of prior misalignments 'oes re'uce the
likelihoo' of termination *p V ,&,5.&
Both kno>le'ge2<ase' hypotheses *Hypotheses : an' -. are supporte' as <oth current
proAects that create reusa<le kno>le'ge *p V ,&,5. an' the cumulative prior num<er of such
proAects *p V ,&,:. re'uce the likelihoo' of customer termination& 7inally/ >e fin' that the
num<er of prior proAects <et>een Compustar an' the customer significantly *p V ,&,5. re'uces
the likelihoo' of termination/ provi'ing strong support for Hypothesis 9&
=hile the results >ere supportive of Hypothesis 9/ >e >ante' to con'uct a''itional
analysis to verify ho> a prior relationship influences customer retention& Mo'el 4 uses a linear
measure of prior relationship/ <ut >e thought that there might <e a non2linear relationship
<et>een prior ties an' customer retention an' >e inclu'e' *ER"(R "T ER(!%CTS.
4
in Mo'el +
to test for such a non2linear relationship& The log likelihoo' for Mo'els 4 an' + are 25++4&59 an'
25+49&41 respectively& Again/ the likelihoo' improvement <et>een Mo'el 4 an' Mo'el + is
statistically significant at the 5G level& The =al' test reAects Mo'el 4 in favor of Mo'el + *the
calculate'
4
S +5&4- >hile the critical
4
*54/ ,&,5. S 4-&44.&
The s@uare' num<er of prior ties is negative an' significant *p V ,&,5./ >hich provi'es
support for the i'ea that customer retention increases as a relationship 'evelops/ an' 'oes so at
an increasing rate& The only other significant change <et>een Mo'el 4 an' Mo'el + is that
CURR%NT M%ASUR%M%NT C(ST is no> significant at the ,&,: level *rather than the ,&5,
level in Mo'el 4./ >hich provi'es stronger support for Hypothesis 5&
Robustness Tests
46
"n or'er to test for the presence of multicollinearity an' ro<ustness of the results/ >e
con'ucte' various tests& (ne such test is the variance inflating factor *?"7. test& "n general/ ?"7
value of less than 5, >oul' in'icate that there is no pro<lem of multicollinearity& "n our conte8t/
the ?"7 factor >as less than + >hich clearly in'icates that there is no pro<lem of
multicollinearity in our 'ataset& =e also con'ucte' the tests <ase' on eigen values an' con'ition2
in'e8 >hich also in'icate that multicollinearity is not a pro<lem in this estimation&
7inally/ >e ran the analysis >ith a variety of 'ifferent >ays of 'etermining >hether the
last proAect <et>een Compustar an' each customer in the sample represente' a terminate'
relationship or a right censore' event& =e use' the follo>ing alternative measures to co'e
termination3 *5. t>o stan'ar' 'eviations a<ove the entire sample average rather than the
customer specific average time <et>een proAects an' *4. fi8e' perio's of 54 an' 51 months& The
results are very similar across all four mo'el specifications&
"#SCSS#O!
The results offer some interesting insights an' theoretical implications& T>o theories that
shoul' influence customer retention are transaction cost economics *TC%. an' the kno>le'ge2
<ase' vie> *B?.& The results offer interesting implications for TC%& =illiamson *501:/ 5000.
states that the governance of a transaction nee's to match the characteristics of the transaction&
Specifically/ transactions su<Aect to contractual ha)ar's arising from asset specificity/
measurement costs/ inter'epen'ence/ etc&/ shoul' <e internali)e'/ >hile transactions free from
such ha)ar's shoul' <e outsource'& "n terms of customer retention/ TC% >oul' imply that
Mmisaligne'N transactions >oul' 'ecrease customer retention& (ur results she' light on this
theory in t>o >ays& 7irst/ >e fin' that it is cumulative misalignment/ rather than Aust
misalignment of the current transaction/ that influences customer retention& Specifically/ >e fin'
4:
that su<contracting a customerFs current proAect that is mission critical 'oes not impact customer
retention/ <ut a history of su<contracting such mission critical proAects for a customer 'oes
re'uce customer retention& This result is not inconsistent >ith TC%/ <ut it represents an area that
nee's further theoretical attention& =illiamson argues that misalignment >ill lea' to >orse
performance/ <ut >hen 'oes misalignment of the current transaction matter more an' >hen 'oes
history of misalignment pose a cumulative pro<lemW "n particular/ >hen might customers
'isplay for<earance for a one2time mistake <ut punish a consistent pattern of such misalignment
mistakesW This highlights the importance of firms learning from their governance mistakes/ in
line >ith =illiamsonFs *5000. call for more attention to learning in TC% research&
Secon'/ >e fin' that proAects that are su<Aect to high measurement costs 'ecrease
retention& This fin'ing also she's light on TC%& 7or instance/ the TC% literature suggests that
>hen there is greater 'ifficulty in evaluating performance/ the firms *customers. coul' integrate
<ack>ar' to their supplier*s.& Ho>ever/ in many instances/ such as in this in'ustry/ <ack>ar'
integration is not feasi<le 'ue to lack of resources an'Kor e8pertise& (ur fin'ings suggest that in
these settings/ customers reveal their preference for re'uce' am<iguity <y not rene>ing their
relations >ith this specific supplier an' e8ploring other suppliers& This outcome is consistent
>ith TC% literatureJperformance am<iguity poses pro<lems for market transactions&
The B? also offers potential 'eterminants of customer retention& The first is the tacit
kno>2ho> that firms 'evelop through their e8perience in the market an' >ith in'ivi'ual
customers& This accumulation of market an' customer specific tacit kno>2ho> ena<les them to
<etter serve their customers an' enhance retention/ in sum enhance their capa<ilities in customer
retention& =e fin' strong evi'ence of this as Compustar gets <etter at customer retention over
the sample perio'& This is a particularly interesting 'ata set for looking at the effect as >e have
4-
'ata 'ating <ack to CompustarFs entry into the "T services in'ustry& %ven controlling for repeat
customers/ <y e8amining the num<er of prior "T proAects >ith the customer/ there is still a strong
increase in CompustarFs a<ility to retain customers over the sample perio'& This supports recent
research into the importance of capa<ilities in areas such as managing alliances *e&g&/ ale/ Dyer
D Singh/ 4,,4. an' governance *e&g&/ Mayer D Argyres/ 4,,6.& =hile customer retention is a
'ifferent capa<ility than kno>ing ho> to effectively colla<orate *Do)/ 500-./ the t>o are relate'&
A strong colla<orative capa<ility shoul' help lea' to greater retention& !ust as colla<orative
capa<ilities 'evelop as firms learn to >ork >ith other firms/ customer retention capa<ilities
shoul' also 'evelop over time as the firm has more e8perience 'ealing >ith managing customer
relationships& Managing customer relationships is a <it 'ifferent than managing an in'ivi'ual
colla<oration/ <ut <oth shoul' 'evelop in the same >ayJthrough e8perience&
Another interesting aspect of the results is that learning <enefits are not Aust relate' to
overall e8perience in the in'ustryJcustomer2specific learning matters& Customer termination is
more likely as the relationship is 'eveloping an' is less likely as Compustar 'oes more proAects
for the customer& This outcome is consistent >ith t>o theoretical motivations& 7irst/ it coul' <e
that relationships lea' to either termination or trust/ an' once trust is esta<lishe' it helps the
parties ri'e out the <umps in the roa' an' continue the relationship *Culati/ 500:B Ring an' ?an
'e ?en/ 5006.& Alternatively/ this result is also consistent >ith a learning an' governance
e8planation& The longer Compustar >orks >ith each customer/ the more they learn to >ork
together an' are likely to 'evelop customer2specific routines *e&g&/ kno>ing >hat to put in the
contract/ kno>ing >ho an' ho> to ask for changes. to make their interaction more efficient
*Mayer D Argyres/ 4,,6.& These customi)e' routines represent a specific investment that >oul'
<e lost if the relationship >ere terminate'/ so the parties >ill re@uire a larger 'istur<ance in or'er
49
to resort to termination& Customer2specific learning can represent a specific investment that
influences customer retention&
=hile learning an' governance play important roles in customer retention/ so 'o
kno>le'ge consi'erations& =hile customer2specific e8perience resulte' in increase' retention/
>e also foun' that proAects >here Compustar create' kno>le'ge that it coul' reuse also
increase' customer retention& =e attri<ute part of this to an incentive effectB >hen Compustar is
<eing pai' to create kno>le'ge it can reuse/ then Compustar has a strong incentive to perform
>ell <ecause they are 'eveloping kno>le'ge that >ill make them more competitive in the future
<y a''ing to their stock of kno>le'ge& "n a''ition/ >e fin' a cumulative effect as the more
proAects >ith the potential to create reusa<le kno>le'ge that Compustar 'oes >ith a particular
customer lea' to increase' customer retention& This result is consistent >ith an economic story
that increase' retention is a result of a supplier >ith strong incentives to perform >ell& "t also
consistent/ ho>ever/ >ith a more kno>le'ge2<ase' learning story& !ust as Compustar learns ho>
to >ork more effectively >ith each customer/ Compustar also learns ho> to <etter incorporate
reusa<le kno>le'ge to serve the customer in the future& =hile the kno>le'ge has the potential to
<e reusa<le for other customers/ it is certainly reusa<le for future proAects >ith the same
customer& Thus Compustar appears to <e 'eveloping a capa<ility in harnessing an' re'eploying
kno>le'ge the more they interact >ith each customer& This e8planation is consistent >ith the
fact that our sample 'ates <ack to the <eginning of CompustarFs participation in the "T services
in'ustry an' the fact that learning >as a focus of their efforts over this perio' as they sought to
gro> the <usiness an' improve their performance&
=hile kno>le'ge playe' an important role in customer retention <y helping Compustar
perform <etter in future proAects >ith that customer/ so 'i' another factorJthe <rea'th of
41
CompustarFs e8perience >ith that customer& The more pro'uct lines that the customer procure'
from Compustar/ outsi'e of "T services/ the greater CompustarFs a<ility to retain the customer in
its "T services 'ivision& There are t>o possi<le reasons for this& 7irst/ customers <uy a greater
variety of goo's an' services >hen they are happy >ith a supplierFs performance/ so the effect
coul' simply <e that Compustar has <uilt up goo'>ill >ith these loyal customers that has le' to
increase' retention& Secon'/ the a<ility of firms to sustain their a'vantage in customer retention
'epen's on ho> goo' they are in leveraging their kno>2ho> an' learning across proAects an'
ho> goo' they are at leveraging an'Kor learning from their other e8periences >ith their
customers& 7irms vary in their a<ility to utili)e customer kno>2ho> across 'ifferent
pro'uctsKservice lines/ as they are <ought from 'ifferent 'ivisionsB hence capturing this customer
kno>2ho> across 'ivisions <ecomes important as it ena<les suppliers like Compustar to offer
more targete' services that are <etter suite' to the customer&
The kno>le'ge <ase' vie> *B?. >oul' also suggest that a strong technical capa<ility
shoul' also influence customer retention through higher customer satisfaction& "nterestingly/ >e
fin' that strong technological capa<ilities have no impact on customer retention& There coul' <e
multiple reasons for this& 7irst/ our measures of strong an' >eak capa<ilities coul' <e fla>e'&
=e <elieve that this is unlikely <ecause of the consistency of responses >e receive'/ <oth from
>ithin Compustar an' from others in the in'ustry/ regar'ing CompustarFs capa<ilities >orking
>ith mainframes an' their fle'gling capa<ilities in programming& Secon'/ it coul' <e that firms
may <e >illing to pay more to access a supplierFs superior capa<ilities an' thus e8pect more
from the supplier in such situations& Alternatively/ >hen a firm engages a supplier >ith >eaker
capa<ilities/ lo>er e8pectations that are more likely to <e met may <e involve'& Eerceptions of
the supplierFs performance may thus <e mo'erate' <y the customerFs e8 ante perception of the
40
supplierFs capa<ilitiesJi&e&/ customers may <e more sensitive to performance relative to
e8pectations than to a<solute performance& "t coul' also <e the case that these technical
capa<ilities in themselves are not a source of competitive a'vantage& =e are ho>ever/ not a<le to
'isentangle these issues in this current research&
Limitations- .uture Research- and Conclusion
(ne of the strengths of this stu'y/ the microanalytic 'ata from >ithin a single firm/ is also
a limitation& (ur 'etaile' an' panel/ transaction2level 'ata ena<les us to offer insights on ho>
<uyer2supplier transaction characteristics over time impact customer retention/ an issue that is
har' to stu'y across firms an' in'ustries given lack of panel 'ata availa<ility& Civen that the
maAority of the Compustar customers are large companies that have many alternatives >hen
selecting "T service suppliers >e are confi'ent that the negotiate' contract 'oes not solely reflect
Compustar policy <ut also significantly integrates specific customer concerns& The fin'ings from
this stu'y are likely to generali)e to other B4B settings as kno>le'ge/ learning an' governance
consi'erations are important in virtually all B4B settings an' are not limite' to the "T services
in'ustries& 7or e8ample/ e8changes in many high technology in'ustries *e&g&/ aerospace/
telecommunications/ soft>are/ pharmaceuticals an' semicon'uctors. involve firms >ith 'iverse
kno>le'ge un'ertaking a series of comple8 proAects that leverage firm2specific intellectual
property& 7uture research to confirm or 'iscount the generali)a<iliy of our fin'ings >oul' <e
valua<le&
=hile this stu'y takes an important first step/ a''itional research is also nee'e' to more
fully unpack the comple8 'eterminants of customer retention in B4B settings& There has <een
little analysis of customer retention in strategy/ even though that is an important varia<le >hen
stu'ying customer2supplier relationships& =e <elieve that more stu'ies of inter2firm
+,
relationships shoul' incorporate a measure of performance an' one such measure that 'oes not
re@uire 'etaile' financial 'ata is customer retention& =hile this measure is not perfect/ no
measure is/ >e <elieve that it is highly correlate' >ith customer satisfaction/ >hich >oul' <e
valua<le to incorporate into more stu'ies of customer2supplier relationships& As long as there are
other suppliers availa<le/ >hich is the case in most in'ustries/ customers shoul' only return if
they are satisfie' >ith the supplierFs performance&
"n a''ition/ >e <elieve that more research is also necessary to un'erstan' the cumulative
effects of prior interaction& %8change 'oes take place >ithin a comple8 >e< of past/ present an'
anticipate' future e8changes *Cranovetter/ 501:./ an' >e nee' to account for the past in or'er to
un'erstan' >hat is happening >ith the current e8change& Erior interactions can result in trust
*Culati/ 500:./ learning *Mayer D Argyres/ 4,,6./ relationship2specific investment *=illiamson/
5005. an'Kor other effects that nee' to <e e8plore'& 7ailure to account for ho> the parties
reache' the current e8change may result in missing important effects an' in misattri<uting a
knee2Aerk reaction to a current pro<lem >hen >hat is really happening is that the customer is
respon'ing to a series of pro<lems that have <uilt up over time&
+5
References
Anan'/ B& D T& hanna & 4,,,& Do 7irms $earn to Create ?alueW The Case of Alliances&
Strategic Management Journal/ /03 40:2+5:&
An'erson/ %& 501:& The Salesperson as (utsi'e Agent or %mployee3 A Transaction Cost
Analysis& Marketing Science/ 1*+.3 4+624:6&
Argote/ $& 5000& Organizational Learning: Creating, Retaining and ran!ferring "no#ledge&
Boston3 lu>er Aca'emic Eu<lishers&
Argyres/ N& 500-& %vi'ence on the Role of 7irm Capa<ilities in ?ertical "ntegration Decisions&
Strategic Management Journal/ 023 54025:,&
Argyres/ N& D !& $ie<eskin'& 5000& Contractual Commitments/ Bargaining Eo>er/ an'
Covernance "nsepara<ility3 "ncorporating History into Transaction Cost Theory&
$cadem% of Management Re&ie#/ /13 602-+&
Argyres/ N& D &!& Mayer& 4,,6& Contract Design Capa<ility as a Source of Competitive
A'vantage3 "mplications for the Roles of Managers/ %ngineers an' $a>yers/ manuscript/
Boston University School of Management&
Bolton/ R&N& 5001& A Dynamic Mo'el of the Duration of the CustomerFs Relationship >ith a
Continuous Service Erovi'er3 the Role of Satisfaction& Marketing Science, 023 6:2-:&
Cannon/ !&E& D =&D& Eerreault !r& 5000& Buyer2Seller Relationships in Business Markets&
Journal of Marketing Re!earch/ 343 6+026-,&
Co8/ D&R& 5094& Regression Mo'els an' $ife Ta<les& Journal of the Ro%al Stati!tical Societ%/
313 519244,&
Demset)/ H& 5005& The Theory of the 7irm Revisite'& "n =illiamson an' =inter *%'s&. he
'ature of the Firm& Ne> #ork3 (8for' University Eress3 5:02591&
Do)/ #&$& 500-& The %volution of Cooperation in Strategic Alliances3 "nitial Con'itions or
$earning ErocessW Strategic Management Journal/ 023 ::21+&
Dyer/ !&% D H& Singh& 5001& The Relational ?ie>3 Cooperative Strategy an' Sources of
"nterorgani)ational Competitive A'vantage& $cadem% of Management Re&ie#/ /33 --,2
-90&
%fron/ B& 5099& %fficiency of Co8Fs $ikelihoo' 7unction for Censore' Data( Journal of the
$merican Stati!tical $!!ociation/ 2/ 5367.3 ::92:-:&
7oss/ N& 500-& no>le'ge2<ase' Approaches to the Theory of the 7irm3 Some Critical
Comments& Organization Science/ 23 69,269-&
+4
Cet)/ C/ R& Blatt<erg/ D !& Thomas& 4,,5& Cu!tomer )*uit%& Harvar' Business Eress3
Cam<ri'ge/ Mass&
Ceyskens/ "&/ !&B&%&M& Steenkamp/ $&& Scheer& D N& umar& 500-& %ffects of Trust an'
"nter'epen'ence on Relationship Commitment3 A Trans2Atlantic Stu'y& +nternational
Journal of Re!earch in Marketing/ 033 +,+2+59&
Cranovetter/ M& 501:& %conomic Action an' Social Structure3 The Ero<lem of %m<e''e'ness&
$merican Journal of Sociolog%/ 703 6152:5,&
Culati/ R& 500:& Does 7amiliarity Bree' TrustW The "mplications of Repeate' Ties for
Contractual Choice in Alliances& $cadem% of Management Journal/ 383 1:2554&
Henis)/ =&!& D !&T& Macher& 4,,6& 7irm2 an' Country2$evel Tra'eoffs an' Contingencies in the
%valuation of 7oreign "nvestment3 The Semicon'uctor "n'ustry/ 500624,,4&
Organization Science 063 :+92::6
Hoetker/ C& 4,,:& Ho> much you kno> versus ho> >ell " kno> you3 Selecting a supplier for a
technically innovative component& Strategic Management Journal/ 4-3 9:20-&
Holmstrom/ B& 5090& Moral Ha)ar' an' (<serva<ility& ,ell Journal of )conomic!/ 095Spring.3
96205&
Holmstrom/ B& D E& Milgrom& 5005& Multi2task Erincipal2Agent Analysis& Journal of La#,
)conomic!, and Organization/ 23 462:4&
ale/ E&/ !& Dyer D H& Singh& 4,,4& Alliance Capa<ility/ Stock Market Response/ an' $ong2
Term Alliance Success3 The Role of the Alliance 7unction& Strategic Management
Journal/ /33 96929-9&
ale/ E&/ H& Singh H&D H& Eerlmutter& 4,,,& $earning an' Erotection of Eroprietary Assets in
Alliances& Strategic Management Journal/ /03 45924+9&
ogut/ B& D U& Han'er& 5004& no>le'ge of the 7irm/ Com<inative Capa<ilities/ an' the
Replication of Technology& Organization Science/ 23 :,42:51&
Macher/ !&T& 4,,:& Technological Development an' the Boun'aries of the 7irm3 A no>le'ge2
Base' %8amination in Semicon'uctor Manufacturing& 7orthcoming in Management
Science&
Masten/ S&%&/ !&=& Meehan D %&A Sny'er& 5005& The Costs of (rgani)ation& Journal of La#,
)conomic!, - Organization/ 23 524:&
Mayer/ &!& D N& Argyres& 4,,6& $earning to Contract3 %vi'ence from the Eersonal Computer
"n'ustry& Organization Science/ 063 +06265,&
++
Mayer/ &!& D !&A& Nickerson& 4,,:& Antece'ents an' Eerformance Conse@uences of
Contracting for no>le'ge =orkers3 %vi'ence from "nformation Technology Services&
Organization Science/ 5-3 44:2464&
Mayer/ &!&/ !&A& Nickerson H& (>an& 4,,6& Are Supply an' Elant "nspections Complements or
Su<stitutesW A Strategic an' (perational Assessment of "nspection Eractices in
Biotechnology& Management Science/ 693 5,-625,15&
Nelson/ R&R& D S&C& =inter& 5014& $n )&olutionar% heor% of )conomic ,eha&ior and
Capa.ilitie!& Harvar' University Eress3 Cam<ri'ge/ Mass&
Nickerson/ !&A& D B& Silverman& 4,,+& =hy 7irms =ant to (rgani)e %fficiently an' =hat eeps
Them from Doing So3 %vi'ence from the 7or2Hire Trucking "n'ustry& $dmini!trati&e
Science /uarterl%/ 183 6++26-:&
Eeteraf/ M&A&& 500+& The Cornerstones of Competitive A'vantage3 A Resource2Base' ?ie>&
Strategic Management Journal/ 013 :502:41&
Eoppo/ $ D T& Henger& 4,,4& Do 7ormal Contracts an' Relational Covernance 7unction as
Su<stitutes or ComplementsW Strategic Management Journal/ /33 9,9294-&
Reichhel'/ 7& 500-& $earning from Customer Defections& 0ar&ard ,u!ine!! Re&ie#/ 213 :-2:0&
Reinart)/ =& !& D ?& umar& 4,,,& The "mpact of Customer Relationship Characteristics on
Erofita<le $ifetime Duration& Journal of Marketing/ 413 592+:&
Ring/ E&/ D ?an 'e ?en/ A& 5006& Developmental Erocesses of Cooperative "nterorgani)ational
Relationships& $cadem% of Management Re&ie#/ 503 0,2551&
Shelanski/ H D E&C& lein& 500:& %mpirical Research in Transaction Cost %conomics3 A
Revie> an' Assessment& Journal of La#, )conomic!, and Organization/ 003 ++:2+-5&
Schmittlein/ D&C&/ D&C& Morrison D R& Colom<o& 5019& Counting #our Customers3 =ho Are
They an' =hat =ill They 'o Ne8tW Management Science, 33/ 5246
Schmittlein/ D& C& an' R& A& Eeterson& 5006& Customer Base Analysis3 An "n'ustrial Eurchase
Erocess Application/ Marketing Science, 03/ 652-9&
Silverman/ B&S&/ !&A& Nickerson D !& 7reeman& 5009& Erofita<ility/ transactional alignment/ an'
organi)ational mortality in the U&S& trucking in'ustry/ Strategic Management Journal/ 08/
*Special "ssue.3 +52:4&
Thomas/ !&S& 4,,5& A Metho'ology for $inking Customer Ac@uisition to Customer Retention&
Journal of Marketing Re!earch/ 383 4-424-1&
Tsiatis/ A& A& 5015&/ A $arge Sample Stu'y of Co8 Regression Mo'el/ $nnal! of Stati!tic! 750::
0+25,1&
+6
=alker/ C& D $& Eoppo& 5005& Erofit Centers/ Single2Source Suppliers/ an' Transaction Costs&
$dmini!trati&e Science /uarterl%/ 343 --219&
=illiamson/ (&%& 509:& Market! and 0ierarchie!; Ne> #ork3 The 7ree Eress&
=illiamson/ (&%& 501:& )conomic +n!titution! of Capitali!m& Ne> #ork/ N#3 The 7ree Eress&
=illiamson/ (&%& 5005& Comparative %conomic (rgani)ation3 The Analysis of Discrete
Structural Alternatives& $dmini!trati&e Science /uarterl%/ 343 4-0240-&
=illiamson/ (&%& 500-& he Mechani!m! of 1o&ernance& Ne> #ork3 (8for' University Eress&
=illiamson/ (&%& 5000& Strategy Research3 Covernance an' Competence Eerspectives& Strategic
Management Journal/ /93 5,19255,1&
+:
Table 0: "escriptive Statistics

?aria<le
Mean
Std
Dev Max Min 1 2 3 4 5 6 7 8 9 10 11 12
1 T"M%TR%ND 8.14 2.88 12 0 1.00
2
ER"(R "T
ER(!%CTS 4.12 7.44 41 0 0.26 1.00
3
CURR%NT
M%ASUR%M%NT
C(ST 0.44 0.50 1 0
-
0.13
-
0.10 1.00
4
ER"(R
M%ASUR%M%NT
C(ST 1.00 1.66 11 0 0.23 0.60 0.11 1.00
5
CURR%NT
R%US% 0.39 0.49 1 0
-
0.04
-
0.07 0.38 0.08 1.00
6 ER"(R R%US% 1.04 1.72 9 0 0.24 0.74

0.04 0.82* 0.09 1.00


7
CURR%NT
CR"T"CA$
SUBC(NTRACT 0.07 0.25 1 0
-
0.03
-
0.06
-
0.18 0.01 0.00 -0.01 1.00
8
ER"(R
CR"T"CA$
SUBC(NTRACT 0.15 0.39 2 0 0.20 0.43
-
0.06 0.28 0.10 0.53
-
0.03 1.00
9 BR%ADTH 4.25 2.77 9 0 0.13 0.55
-
0.21 0.21
-
0.05 0.30 0.02 0.32 1.00
10
CURR%NT
MA"N7RAM% 0.26 0.44 1 0 0.02
-
0.07 0.06 -0.02 0.05 0.00 0.10
-
0.02
-
0.07 1.00
11
CURR%NT
ER(CRAMM"NC 0.46 0.50 1 0 0.02
-
0.06 0.27 0.17 0.24 0.09 0.02
-
0.07
-
0.09 0.07 1.00
12
CURR%NT
CR"T"CA$ 0.47 0.50 1 0 0.23
-
0.06
-
0.05 0.02 0.10 0.04 0.29 0.09 0.01 0.28 0.08 1.00
X
Because of the relatively high correlation/ >e ran severeal tests *inclu'ing ?ariance "nfaltion 7actor *?"7. test. in or'er to test for multicollinearity& ?"7 >as less
than +/ >hich in'icate that there is no pro<lem of multicollinearity&
+-
Ta<le 4
Regression Results
Depen'ent ?aria<le3 T%RM"NAT"(N
Model 1 Model 2 Model 3
Breadth -0.171 *** -0.117 *** -0.142 ***
(0.024 (0.031 (0.032
!i"e !rend -0.176 *** -0.179 *** -0.195 ***
(0.025 (0.026 (0.027
#$rrent %ro&ra""in& 0.001 0.067 0.029
(0.121 (0.131 (0.131
#$rrent Main'ra"e -0.044 -0.025 -0.037
(0.145 (0.147 (0.146
#$rrent #riti(al -0.015 -0.025 0.002
(0.131 (0.138 (0.138
#$rrent Mea)$re"ent #o)t 0.244 * 0.330 *
(0.145 (0.148
#$rrent +e$)e -0.388 ** -0.390
(0.148 (0.147
#$rrent #riti(al S$,(ontra(t -0.002 0.033
(0.274 (0.275
%rior Mea)$re"ent #o)t 0.216 ** 0.183 **
(0.071 (0.069
%rior +e$)e -0.201 * -0.177 *
(0.091 (0.087
%rior #riti(al S$,(ontra(t 0.606 ** 0.658 ***
(0.202 (0.194
%rior -! %ro.e(t) -0.041 ** 0.039
(0.016 (0.030
(%rior -! %ro.e(t)
2
-0.002 **
(0.001

/o&-li0elihood

25+69&1 25++4&59 25+49&41



Y p V &5, Z p V &,: ZZ p V ,&,5 ZZZ p V &,,5
As this is a ha)ar' rate mo'el e8amining termination/ a positive coefficient signifies a 'ecrease
in customer retention *i&e&/ increase in the likelihoo' of e8it.&
X
=al' test reAects Mo'el 5 in favor of Mo'el 4/ an' reAects Mo'el 4 in favor of Mo'el +&
+9

You might also like