Vous êtes sur la page 1sur 84

WhatMakesYour

AppGoViral?

Introducingthe
ViralGrowthFormula

MScThesis
JeroenSakkers(0516147)
jeroen@victormundi.com
(+31)623946416

Supervisors:
dr.G.T.Vinig(first)
drs.ing.A.C.J.Meulemans(second)

June24th,2012

Abstract

Thisresearchprovidesinsightsinthedriversofviralgrowthformobileandweb
applications,teststhesethrougharealworldexperiment,andultimatelycombines
thefindingsintoanewlyintroducedviralgrowthformula.Theincrediblesuccess
of companies like ICQ, Hotmail, PayPal, Facebook, Dropbox, and Instagram can
almost always be traced back to viral growth. To understand the drivers, a
theoretical framework is proposed that combines the fields of viral marketing,
network dynamics, and incentive systems. Three types of virality arise: inherent
virality,networkedvirality,andincentivizedvirality.Arealworldexperimentwith
a sample size of 313 customers is conducted via Victor Mundi an online
application targeting entrepreneurs to show how networked virality and
incentivized virality can be created or enhanced. Seeding a product or service to
network hubs a small fraction of (potential) customers that are many times
better connected than average and incentivizing wordofmouth via both
materialandpurposiveincentivesisfoundtobeveryimportantinordertoinduce
viralgrowth.Theexperimentshowsthatthenumberofinvitessentpercustomer
andthecycletime,twoofthethreevariablesthattogetherformtheviralgrowth
equationtested,are(highly)correlatedwithseedingandincentivizingstrategies.
Managerialimplicationsarediscussedforfoundersandproductmanagersofstart
ups andotherbusinesses that want tomake theirapp(s) go viral.Ultimately, the
ViralGrowthFormula(VGF)isintroducedincludingaviralroadmapforweband
mobileapplications.

Keywords:viralgrowthformula,viralcoefficient,cycletime,incentives,networks,
networksize,networkhubs,wordofmouth

MastersThesis(BusinessStudies)CopyrightsJ.Sakkers(2012)Page2of84

Tableofcontents

1.Introduction

2.Theoreticaloverview

2.1.TheHolyGrailforbusinessesgoingviral

2.1.1.Driversofwordofmouth

2.1.2.Effectsofwordofmouth

2.1.3.Typesofvirality,strategiesandmeasures

2.2.Thestickandthecarrotincentivizingwordofmouth
2.3.Theidealcustomerimportanceofnetworks

2.4.Conceptualframework

3.Researchmethodanddatacollection

3.1.Methodologyanddesign

3.2.Variables

3.3.Datacollection

3.4.Reliability

4.Results

4.1.Kindofdata

4.2.Statisticalprocedures

5.Discussionandlimitations

5.1.Potentialimpact

5.2.IntroducingtheViralGrowthFormula

5.3.Limitationsandweaknesses

5.4.Futureresearch

Appendices:

AppendixI.References
AppendixII.Designofexperiment

12
14
16
19
20
28
33
39

40
41
42
43
45

47
47
47

54
55
56
59
60

MastersThesis(BusinessStudies)CopyrightsJ.Sakkers(2012)Page3of84


Thereisnothingmoredifficulttotakeinhand,moreperiloustoconduct,ormore
uncertaininitssuccessthantotaketheleadintheintroductionofaneworderof
things.

NiccolMachiavelli

1.Introduction

Bothlaunchedin1996,ICQ,afreeinstantmessagingservice,andHotmail,afree
email service, accumulated 12 million users in their first (ICQ) and second
(Hotmail)yearofexistence.In1998,twoyearsaftertheirlaunch,Microsoftbuys
Hotmail for $400 million and AOL buys ICQ for $408 million (De Bruyn & Lilien,
2008; Ignite Social Media Blog, 2012; Slashdot, 2012). Their secret recipe: ICQ
offered an option to invite ones friends automatically to join the communication
network,whileHotmailattachedthemessageGetyourfreeemailatHotmailat
thebottomofeveryemailsentbyaHotmailuser.
For Dropbox, a file hosting service founded in 2007, traditional Internet
marketing channels like Google AdWords did not work. Encouraging wordof
mouth, and therefore, virality, worked perfectly well though. By implementing a
referralprogramwithtwosidedincentivesinspiredbythe$5signupbonusthat
got PayPal to three million users within nine months of its launch Dropbox
permanently increased signups by 60%. Between September 2008 and January
2010 Dropbox got from 100,000 to 4 million users, stacking up to more than 50
million users in October 2011 relying mostly on incentivized wordofmouth
(StartupLessonsLearned,2012;Wikipedia,2012I).
WhatdocompanieslikeICQ,Hotmail,PayPal,Facebook,DropboxandInstagram
have in common? Indeed, they are all insanely successful within their respective
playingfield(s).Thankstotheirviralgrowththeybroughtthetermhockeystick
growthtoawholenewlevel.Butwhenwelookcloser,wecanfindmoredetailed
similarities. Namely, ICQ and Facebook are inherently viral products, meaning a
persongetsnovaluefromtheproductunlessotherpeopleuseitaswell

MastersThesis(BusinessStudies)CopyrightsJ.Sakkers(2012)Page4of84


Figure1Anoverviewofsuccessfulcompaniesanddifferenttypesofvirality

(TechCrunch, 2012). Both PayPal and Dropbox are highly successful thanks to
wellthoughtincentivesystems(DeBruyn&Lilien,2008;StartupLessonsLearned,
2012).AndHotmailaswellasInstagrambenefitedtotheutmostextentfromthe
networksoftheir(early)customers(Porter&Golan,2006;Skok,2012II).
Not coincidentally, the tremendous viral growth at all of these companies
happened after reaching a certain tipping point, that moment in an epidemic
whenavirusreachescriticalmass(Gladwell,2000:p.12).Thistermcomesfrom
theworldofepidemiology,andcanbefoundinvarioussocialepidemicstoo.Aids
tippedbackin1982,whenitwentfromanichediseaseaffectingafewgaymento
a worldwide epidemic; crime tipped in New York City in the mid 1990s when
murderrateplummeted;andteenagesuicidetippedintheSouthPacificislandsof
Micronesiainthe1970sand1980safterateenagesuicidewascoveredinoneof
the local papers (Gladwell, 2000: p. 135168; 216252). Hotmail was the first
example of an epidemic in business: people saw similarities between the rise of
thiscompanyandthewayaviruslikethefluspreadsfrommantoman(Porter
&Golan,2006).JeffreyRayportsubsequentlyintroducedthetermviralmarketing
in1996(Kaplan&Haenlein,2011).
Let us examine the impact of viral growth by a short calculative story: John
knowsasecreteverypersonontheplanetshouldknow.Butheistheonlyonewho

MastersThesis(BusinessStudies)CopyrightsJ.Sakkers(2012)Page5of84


knows, and he is only allowed to tell two others every half hour. Likewise,
everyonewhohearsaboutthesecretcantelltwoothersthathaventheardaboutit
yet every half hour as well. How long would it take before the whole human
population that surpassed 7 billion on March 12, 2012 (Wikipedia, 2012II)
wouldknowaboutthesecret?Only16.5hours(!).Theresultinggrowthpatternin
thisstoryisexponential(Kaplan&Haenlein,2012).Thispotentialimpactisoneof
thereasonsmarketersspentmorethan$1.54billiononwordofmouthmarketing
initiatives in 2008, and will spend more than $3 billion on it in 2013 (PQ Media,
2012). With the rise of platforms like Facebook, Google+, LinkedIn, and Twitter,
withhundredsofmillionsorevenclosetoabillionuserseach,viralitywillonlybe
facilitated more and more over the coming years. Thereby increasing
opportunitiesforstartupsandothercompaniesworldwide.
Apart from its potentially mindboggling growth figures, viral marketing can
alsobealotcheaperthantraditionalmarketingchannels;forexample,traditional
forms of advertising, such as the 30second television spot, are becoming less
effective everyday. Also, consumers get to know the advertising industry better
and better, and therefore are resistant against messages that are too obviously
trying to get them to buy stuff (Kaikati & Kaikati, 2004). Viral messages are not
perceived as intrusive, in opposite to traditional types of advertising (Golan &
Zaidner,2008).Nevertheless,traditionalmarketingeffortslikeadvertisingremain
effective in a relatively early stage of the growth cycle of the new product. After
that,theirefficacyquicklydiminishesandstrongandweaktiesbecomethemain
forces propelling growth though (Goldenberg et al., 2001). Moreover, customers
disperse over an increasing number of media channels and therefore it is
increasinglydifficultforcompaniestofocusontheirtargetgroups(Porter&Golan,
2006).
Ontheotherhand,therearealsocertainrisksattachedtoviralmarketing.Plaxo,
for example, launched its product in November 2002 with a viral marketing
campaign that attracted 5 million users in 3 years. The hockey stick growth
continued,withPlaxogrowingto15millionusersbyAugust2006(Kalyanametal.,
2007).However,acoupleofyearslaternumbershaveshrunktolessthan600,000
unique monthly visitors (Compete, 2012) due to worldwide complaints and a

MastersThesis(BusinessStudies)CopyrightsJ.Sakkers(2012)Page6of84


damagedbrandimagebecauseoftheirspamperceivedemailnotifications(ZDNet,
2012). Also, because companies are dependent on customers in their viral
marketingefforts,itisapossibleriskthatcustomerswillasksomethinginreturn
forperformingviralactions(Helm,2000).Andthewideadoptionofsocialmedia
has increased the competition among ideas for our finite attention. People share
messages via their networks but can only pay attention to a portion of the
informationtheyreceive.Intheemergingdynamicsofinformationdiffusion,afew
companiesgoviralwhilemostdonot(Wengetal.,2012).
Many research gaps and opportunities arisewhen lookingat the fieldsofviral
marketing, network dynamics, and incentive systems. First, let us look at the
common ground between viral marketing and network dynamics. Weng et al.
(2012: p. 6) speak of a virtually unbounded number of new epidemics that are
continuously injected into the system, and that are all fighting for our limited
attention. According to many researchers, it is unclear though how successful
viralmarketingcomesabout(Helm,2000;Dobeleetal.,2007;DeBruyn&Lilien,
2008; Berger & Schwartz, 2011). De Bruyn & Lilien (2008: p. 152) say that
consumers experience a high level of noise in their daytoday electronic
communications and for viral marketing campaigns to be designed more
effectively,thereisneedtobetterunderstandwhichonlinereferralsarelikelyto
cutthroughtheclutterandwhicharenot.
DonckelsandLambrecht(1997)posethatlittlehasbeendonetomakenetwork
theory operational in that effect, to help organizations position in networks or
develop network strategies. However, Watts & Strogatz (1998) and Kleinberg
(2000)didcomeupwiththeoriesthatstatethat(thecharacteristicsof)networks
are important for viral success. They conclude that the level of randomness in
networks has a striking effect on both the extend of an epidemic and the time
taken for it to spread. Only a small amount of randomness is required for large,
rapidlyspreadingepidemicstooccur.Further,thecorrelationlengthofnetworks
maybeimportant(Watts,2004:p.260).Boots&Sasaki(1999)suggestthatasthe
correlationlengthofanetworkshrinksonaccountoftheintroductionofrandom
shortcutstheworldvirtuallybecomessmalleronemightexpecttheevolution
of more virulent humantohuman diseases. If we bring this back to companies

MastersThesis(BusinessStudies)CopyrightsJ.Sakkers(2012)Page7of84


going viral, we see that electronic wordofmouth by definition spreads via long
range connections, due to their use of global mobile and social networks for
distributionpurposes.ThesenetworkscanbeseenastherandomshortcutsBoots
andSasaki(1999)talkabout,andthatmightexplaintheriseofmorerecentviral
successes like Instagram via Apples iOS platform, Draw Something via both
Googles Android platform and Apples iOS platform, and many of Zyngas
applicationsviatheFacebookPlatform.
Godes&Mayzlin(2009:p.738)alsosuggestthatidentifyingnewmeasuresthat
havetodowiththedisseminatorsnetworkwouldbeuseful.Bampoetal.(2008:p.
289) further stress the importance of executing more sophisticated and targeted
seedingexperimentssothattheroleofhubsasmallfractionofpeoplethatare
many times better connected than average (Watts, 2004: p. 250) in seeding
strategies would be better understood. The nature and arrangement of network
ties may play an important role in fostering the global spread of information as
well(Watts,2004;Brownetal.,2007;Sweeneyetal.,2007).Therefore,companies
should use social network information about mutual relationships to determine
theirviralmarketingstrategy,sayHinzetal.(2011:p.29).
Second,letusexaminetheresearchgapsandopportunitiesthatcomeforthout
of connecting the fields of viral marketing and incentive systems. According to
Godesetal.(2005:p.419)itisstillunclearwhetherincentivescaninfluencethe
sending of viral messages. They reckon there has been very little research to
investigate this important question, while companies need to understand better
theresponsefunctiontovarioustypesofincentivessuchasmonetaryincentives,
discountbasedincentivesandrecognitionbasedincentives.AccordingtoRyuand
Feick (2007) rewards increase referral likelihood. However, referral likelihood
doesnotalwaysleadtoactualreferrals(Mittal&Kamakura,2001).RyuandFeick
(2007) suggest that further research should focus on this area, amongst others
becausetherehasbeenalmostnoempiricalwork.Hinzetal.(2011:p.27)confirm
that experimental comparison of seeding strategies is unprecedented; previous
literature is based solely on mathematical models and computer simulations.
They indicate that from a marketing perspective, it would be very interesting to

MastersThesis(BusinessStudies)CopyrightsJ.Sakkers(2012)Page8of84


intensifyresearchregardingtheeffectsofincentives,astheyfoundtheirimpactto
bequitesignificantinoneoftheexperimentstheyconducted.
Porter&Golan(2006)statethatearlierresearchontheinfluenceofincentives
on recommendations focused on peertopeer communication of persuasive
messagesthatarebasedontheexperienceofthesenderwithacertainproductor
service. However, this research focuses on peertopeer communication of
persuasive messages independent of the positive or negative experience of the
sender(andrecipient)withacertainproductorservicepredominantlyreferred
toasanelementofintrinsicmotives(Schutz,1992;Ho&Dempsey,2010).
Making use of incentives can be both riskful and rewarding. On the one hand,
consumersspendrelativelymuchtimetowordofmouthbecausetheproviderof
the recommendation does not act out of selfinterest (Wirtz & Chew, 2002).
Therefore, introducing incentives to either senders or receivers might take away
theessenceofwordofmouthitslooseness.Anotherriskisthattherecipientof
theviralmessageperceivesitasspam(Dobeleetal.,2005).Ontheother,seeding
the right consumers with the right incentives yields up to eight times more
referrals than seeding the wrong ones. To combine this insight with networks
dynamics,seedinghubsandbridgescaneasilyincreasethenumberofsuccessful
referralsbymorethanhalf(Hinzetal.,2011).Finally,genderdifferencesarefound
in the effects of viral marketing and incentive systems (Garbarino & Strahilevitz,
2004; Awad & Ragowsky, 2008; The Next Web, 2012). Loscocco et al. (2009)
suggest that insights from gender construction perspectives should therefore be
integratedintonetworkandothergenderinequalitystudies.
Recently, many research papers sprout from the fields of viral marketing,
networkdynamics,andincentivesystems.Afewauthorsproposedtheintegration
ofthesefields,thereunto.Araletal.(2011)concludethattwoprincipalstrategies
have emerged to leverage the wordofmouth effect: seeding (or network
targeting)andincentives(referralstrategiesthatencouragepeeradoption).Buttle
(1998:p.246)stressestheimportanceofmoreresearchontheeffectivenessand
application of incentives, as he finds that both networks and incentives are
important drivers in the wordofmouth model. Godes and Mayzlin (2009) stress
the importance of future research on how extrinsic and intrinsic incentives

MastersThesis(BusinessStudies)CopyrightsJ.Sakkers(2012)Page9of84


compare in creating wordof
mouth, keeping in mind their

Viralmarke*ng

findings that one of the key


differences between loyal and
nonloyalcustomersistheissueof
intrinsic incentives while for a

Network
dynamics

Incen*ve
systems

product

with

low

initial

awareness level, wordofmouth


that is most effective at driving
sales is created by less loyal
customers and occurs between

Figure2Connectingthefields

acquaintances, not friends. In


additiontoalltheforegoing,afew

patents that are registered in the Unites States combine the concepts of viral
marketing, network dynamics, and incentive systems (Atazky & Barone, 2007;
MicrosoftCorporation,2011).
Thus,sciencematuredenoughtoconnectthefieldsofviralmarketing,network
dynamics,andincentivesystemsnow,andorganizationsarewaitingforanswers.
Therefore,theresearchquestionisofthisthesisis:

Whateffectdo(differenttypesof)incentivesandthesizeofcustomers
networkshaveontheviralsuccessofamobileorwebapplication?

Hence,thetitleofthisresearchpaperisWhatMakesYourAppGoViral,andthe
goals for this research are fourfold: (1) to introduce a model of incentivized
virality for products and services; (2) to test the above model with a real world
experiment; (3) to provide insights in the way different types of incentives and
customerprofilesinfluenceviralityofproductsandservices;and(4)tointroduce
theViralGrowthFormulaincludingaviralroadmapprovidingstartupsandother
entrepreneurswithaframeworkformakingtheirproductsandservicesgoviral.A
startupviewpointishandled,sotheresearchpaperisstrictlyfocused:forexample,
senders and recipients of wordofmouth being discussed are always (potential)

MastersThesis(BusinessStudies)CopyrightsJ.Sakkers(2012)Page10of84


customers, and the subject of wordofmouth being discussed is always the
organizationand/oritsproduct(s)orservice(s).
This research paper is organized as follows. First, an extensive theoretical
overviewthatbothexplainsandconnectsthemajorconceptswithinthefieldsof
viralmarketing,networkdynamicsandincentivesystems,isprovidedonthebasis
of a new theoretical framework. At the end of this theoretical overview a
conceptual framework is proposed, and hypotheses that will be tested are
provided. Second, the research and data collection methods are clarified. Third,
statistical procedures are explained and conducted. Fourth, the results, potential
impact of this research, the introduction of the Viral Growth Formula plus
accompanyinglimitationsandweaknessesandimplicationsforfutureresearchare
discussed. Last but not least, a list of references and the actual setup of the
experimentareadded.

MastersThesis(BusinessStudies)CopyrightsJ.Sakkers(2012)Page11of84


Thedifficultylies,notinthenewideas,butinescapingtheoldones,whichramify,
forthosebroughtupasmostofushavebeen,intoeverycornerofourminds.

JohnMaynardKeynes

2.Theoreticaloverview

Inthischapteracompleteoverviewofexistingacademicliteratureconcerningthe
theoreticalconceptsusedinthisresearchisprovided.Consecutively,maintopics
and concepts within the field of viral marketing, incentive systems, and network
dynamics will be discussed. But first, most important theoretical concepts within
thesefieldswillbeintegratedintoamodelofincentivizedviralityforproductsand
services(see:Figure3).
Beforedivingindeeperwiththetheoreticalconceptsofeachresearcharea,let
usfirstlookatthemainitemswithintheproposedmodel.Thekeyplayersinthese
researchfieldsaretheorganization,whoinitiatesordesignsviralmarketingand

Figure3Modelofincentivizedviralityforproductsandservices

MastersThesis(BusinessStudies)CopyrightsJ.Sakkers(2012)Page12of84


incentivesystems,thesender,whoinitiateswordofmouthbysendingpersuasive
messages peertopeer, and the recipient, who receives the persuasive message
and does or does not act upon it. We find these actors in numerous research
papers (Bansal & Voyer, 2000; Ryu & Feick, 2007; Berger & Schwartz, 2011;
Kaplan&Haenlein,2011).Sometimes,thewebsitethewordofmouthissentfrom
isalsoreferredtoasactor(Brownetal.,2007).
Bothsendersandrecipientsof(electronic)wordofmouth,whichisdefinedby
HennigThurauetal.(2004:p.39)ascommunicationthatconsistsofanypositive
or negative statement made by potential, actual, or former customers about a
product or company, which is made available to a multitude of people and
institutions (via the Internet), make use of their (personal) networks that affect
theirbehavior(Borgatti&Jones,1998;Goldenbergetal.,2001;Watts,2004;Ryu&
Han, 2009). Of course the organization also takes on different roles: observer,
moderator, mediator and/or participant (Godes et al., 2005). However, in this
researchwewillfocusonthesenderandtherecipientofwordofmouthandthe
wayorganizationscanbenefitofthenetworksofthesekeyplayersbyincentivizing
them.
Whereas the persuasive peertopeer communication from the sender to the
recipient,thewordofmouthcommunication,canbetriggeredbyvariousdrivers
(Sundaram et al., 1998; Dellarocas & Narayan, 2006; Mazzarol et al., 2007), the
interaction between the organization on the one hand, and the sender and the
recipient on the other hand, can be triggered by incentives (Biyalogorsky et al.,
2001;Wirtz&Chew,2002;HennigThurauetal.,2004;Bronner&DeHoog,2010).
Clark&Wilson(1961)defineincentivesastangibleorintangiblerewardsthatare
offeredbyorganizationstoinfluencecustomeroremployeebehavior.Wewillstick
to this definition; though use it only in reference with influencing customer
behavior.
Althoughitmayseemfarfetchedtocombinethreeoftheseemergingresearch
areas, Wilson (2005) already provided six steps to an effective viral marketing
strategy that combines the three key areas discussed in this paper: give away
products or services (incentives); provide for effortless transfer to others
(virality); scale easily from small to very large (virality); exploit common

MastersThesis(BusinessStudies)CopyrightsJ.Sakkers(2012)Page13of84


motivations and behaviors (incentives); utilize existing communication networks
(networks); and take advantage of others resources (networks). Moreover, De
BruynandLilian(2008)stretchthatthethreemoststrikingwaysofsuccessfully
implementing viral marketing occur when (a) customers willingly become
promoters of a product or service and start spreading the word (virality); (b)
customersstartspreadingthewordafterreceivingorbeingpromisedanexplicit
incentive(incentives);or(c)customerssimplydesiretosharetheproductbenefits
withtheirfriends(networks)preciselythesethreeresearchareashandledinthis
paper.Also,Hinzetal.suggestthatthestructureofnetworks,thecharacteristicsof
theactors,andtheincentivesforsharingmessagestogetherformthecriticalviral
marketingsuccessfactorsfororganizations.

2.1TheHolyGrailforbusinessesgoingviral:

First, the field of viral marketing is examined further. As explained, (electronic)


wordofmouthcanbedefinedascommunicationthatconsistsofanypositiveor
negativestatementmadebypotential,actual,orformercustomersaboutaproduct
orcompany,whichismadeavailabletoamultitudeofpeopleandinstitutions(via
theInternet)(HennigThurauetal,2004:p.39).Wordofmouthmarketingisthe
firmsintentionalinfluencingofconsumertoconsumercommunications(Kozinets
et al., 2010). Viral marketing can furthermore be described as the pattern in
which Internet companies spread by making use of customer referrals (Helm,
2000)oranystrategythatencouragesindividualstopassonamarketingmessage
toothers,creatingthepotentialforexponentialgrowthinthemessagesexposure
and influence (Wilson, 2005: p. 1). Kaplan and Haenlein (2011: p. 254) finally
describe viral marketing as follows: electronic wordofmouth whereby some
formofmarketingmessagerelatedtoacompany,brand,orproductistransmitted
in an exponentially growing way other through the use of social media
applications. Wordofmouth marketing, including traditional and electronic
wordofmouth,canthereforebeseenasoneofthemostimportantaspectsofviral
marketing.

MastersThesis(BusinessStudies)CopyrightsJ.Sakkers(2012)Page14of84


While wordofmouth existed for quite some time already (Buttle, 1998),
marketing practitioners and theorists really got the hang of it after the
introductionoftheInternet,anddigitalizedsocialnetworksthereafter;marketers
and authors are increasingly convinced that positive wordofmouth can be an
important part of marketing a product or service (Ryu & Feick, 2007). And the
Internet enabled customers to interact faster, independent of place and time
(Riegner,2007).Moreover,electronicwordofmouthismoreinfluentialcompared
totraditionalwordofmouththankstoitsspeed,convenience,onetomanyreach,
anditsabsenceoffacetofacehumanpressure(Sunetal,2006:p.1105).Wordof
mouthplaysanimportantroleincustomeracquisition(Trusovetal,2009)andis
more influential on behavior than other marketercontrolled sources (Buttle,
1998:p.242).Traditionalandelectronicwordofmouthdifferfromeachotherin
manyways;theeffectsofbothonconsumersareequalthough(HennigThurauet
al.,2004;Gruenetal.,2006).Theantecedentsoftraditionalandelectronicword
ofmouthexhibitimportantsimilaritiesaswell(Dellarocas&Narayan,2006).
More than traditional marketing, viral marketing relies on actors that send
persuasivemessagespeertopeer,whichcanbetrackedbacktoacertainsponsor
acompany,inalmostallcases(Porter&Golan,2006).Likeavirus,information
about the company and its brand message, products, or services is spread to
(potential)customers,whothenpassitontoother(potential)customerssuchthat
a huge network is rapidly created (Dobele et al., 2005). And while the effects of
traditional advertising are fading, often resulting in consumers ignoring the ads
(e.g., banner blindness: overlooking distinctive banners) (Benway, 1998),
consumers perceive information received from another consumer fundamentally
differentthanthatreceivedfromacompany;thefirstistypicallyfreeofbiasborn
on the desire for a sale (Godes et al, 2005: p. 418). Marketers also refer to this
phenomenon as flipping the sales funnel; instead of targeting a wide range of
potential customers and hoping for a few real customers, the process is turned
upside down: letting a few (potential) customers spread the word via their
communication channels to get more customers interested (see: Figure 4; Jaffe,
2010:p.4980).

MastersThesis(BusinessStudies)CopyrightsJ.Sakkers(2012)Page15of84


Figure4Traditionalmarketingfunnelversusviralmarketingfunnel

Inthisresearchpaperwewillfocusonstatusupdatesthataresharedthroughthe
social networks of (potential) customers and invite messages via email sent by
(potential)customers,bothreferredtoasinviteshereafter.

2.1.1Driversofwordofmouth:

There are three important factors that can be manipulated to create a viral
epidemic: environment, message, and messengers (Kaplan & Haenlein, 2011).
While the environment is driven mainly by external factors like politics,
economics, and technology, message can be altered by companies and induced
with incentives, which will be discussed later. Moreover, (the right) messengers
canbepickedoratleasttargetedbythecompanythatstartsaviralcampaign.This

MastersThesis(BusinessStudies)CopyrightsJ.Sakkers(2012)Page16of84


is all about networks of messenger, which too will be discussed later. Hereunder
you will find an overview of literature on the three factors that are important in
creatingaviralepidemic:

Table1Overviewofliteratureonfactorsincreatingviralepidemics
Author

Factor

Contribution

Watts,2004

Environment

Propertiesofnetworks(correlationlength,
longrangeconnections)haveeffectonvirality

Brownetal.,2007

Environment

Websitesareperceivedasprimaryactors

Petty&Cacioppo,1986

Message

Pheripheralcuesindicateinformations
validity,besidescontentitself

Sussman&Siegal,2003

Message

eWOM:pheripheralcuesnotalwayspresent

Kalyanametal.,2007

Message

ViralIndexhelpscreateviralmessages

Riegner,2007

Message

Emailremainsdominantforallagegroups,
under25group:IM,chat,fora,blogs,texting

Strapparavaetal.,2011

Message

Viralityisinnatureofcontentbeingspread

Engeletal.,1969

Messengers

Bestsalesmanissatisfiedcustomer,innovator
orfirstuserisvaluableassettofirms

Valente&Davis,1999

Messengers

Diffusionoccursmuchfasterwheninitiatedby
opinionleaders

Bansal&Voyer,2000

Messengers

Recipientsexpertise,perceivedrisk,senders
expertise,andtiestrengtheffectvirality

Goldenbergetal.,2001

Messengers

Influenceofweaktiesatleastasstrongas
influenceofstrongtiesinachievingvirality

Watts,2004

Messengers

Highlyconnectedhubs(manyconnections)
ensurewidespreadinscalefreenetwork

Sunetal.,2006

Messengers

Efluentials:opinionleaderswhospread
informationviaInternet

Mazzaroletal.,2007

Messengers

Tiestrengthinfluencespurchaseintentions

Sweeneyetal.,2007

Messengers

ImpactofWOMonrecipient,tiestrength
influencespurchaseintentions

MastersThesis(BusinessStudies)CopyrightsJ.Sakkers(2012)Page17of84


DeBruyn&Lilien,2008

Messengers

Messagedisseminationbyconsumerscanbe
intentionalorunintentional

Cruz&Fill,2008

Messengers

Perceivedobjectivityandcredibilityofviral
messages:high;receivedfromacquaintances

Godes&Mayzlin,2009

Messengers

Opinionleaders:effectivespreadingofWOM
whenloyalcustomers;ifnot,risful

Wengetal.,2012

Messengers

Structureoftheunderlyingnetwork,acivityof
users,sizeof(potential)audience,degreesof
influence,intrinsicqualityofinformation
spreadareofeffectonviralityofmemes

According to De Bruyn and Lilien (2008: p. 152), consumers experience a lot of


noise during their use of electronic communications. Therefore, viral marketing
campaigns need to be designed more effectively there is need to better
understandwhichmotivesdrivepeopletoparticipate.Berger&Schwartz(2011:p.
877)alsocometotheconclusionthatlessisknownaboutthecausesofwordof
mouth and why certain products are talked about more than others. The group
mapping of providers of wordofmouth on what motivates their behavior by
HennigThurauetal.(2004:p.49)isusefulforthispurpose;theydistinguishfour
groups:(1)selfinterestedhelpers;(2)multiplemotiveconsumers;(3)consumer
advocates;and(4)truealtruists.Wecananalyzeandcategorizemotivesthatcause
messengerstoparticipateinviralcampaigninthesameway.Thereto,youwillfind
an overview of literature on the drivers and motives behind spreading wordof
mouth:

Table2Overviewofliteratureondriversofwordofmouth
Contributors

Motive

Driver

Wirtz&Chew,2002;Sundarametal,

Selfenhancement

Selfinterest

2008;Berger&Schwartz,2011
Berger,2011;Berger&Milkman,2012 Psychologicalarousal

Selfinterest

Buttle,1998;Biyalogorskyetal.,2001; Incentives

Selfinterest

MastersThesis(BusinessStudies)CopyrightsJ.Sakkers(2012)Page18of84


HennigThurauetal.,2004;Ryu&
Feick,2007;Godes&Mayzlin,2009;
Bronner&DeHoog,2010;Berger&
Schwartz,2011
Tuketal.,2009

Financialrewards

Selfinterest

HennigThurauetal.,2004;Ho&

Socialinteraction

Selfinterest

Needforcontrolorself

Selfinterest

Sundarametal,2008

Helpingthecompany

Multiplemotive

Berger&Schwartz,2011

Interestingproduct

Multiplemotive

Sundarametal,2008

Productinvolvement

Advocate

Sundarametal,2008

Productperformance

Advocate

Sundarametal,2008

Employeeconsumercontact

Advocate

Biyalogorskyetal.,2001;Wirtz&

Stateofdelight

Advocate

Altruism

Altruist

Dempsey,2010
Schutz,1992;Whiting&DeJanasz,
2004;Ho&Dempsey,2010

Chew,2002
Sundarametal,2008

2.1.2Effectsofwordofmouth:

De Bruyn and Lilien (2008) stress the fact that little is known on how wordof
mouth effects and influences consumers actual behaviors. However, Hennig
Thurauet al. (2004), Gruenet al.(2006)and Mazzaroletal.(2007)didfind that
theeffectsoftraditionalandelectronicwordofmouthareequalandmanyother
authorsdoexplaintheeffectsofwordofmouth.The(potential)effectofwordof
mouth is strongly dependent on the drivers and motives as mentioned before
(Sweeney et al., 2007). Of course, the foremost effect of wordofmouth is
exponentialgrowthofthecompanythatinitiatedit.Keytothisgrowthisvirality,
thuswordofmouth(Helm,2000).Inadditiontothismaineffect,asummaryofall
literatureonothereffectsofwordofmouthisprovidedhereunder,categorizedby

MastersThesis(BusinessStudies)CopyrightsJ.Sakkers(2012)Page19of84


thekeyphasesofthemodelofpurchasingbusinessservicesbyFitzsimmonsetal.
(1998) need identification, information search, vendor selection, and
performanceevaluation:

Table3Overviewofliteratureoneffectsofwordofmouth
Contributors

Phase

Effect

Bansal&Voyer,2000;

Needidentification Influenceandeffectonpurchase

Mazzaroletal.,2007;

intentionsofrecipient(also,differences

Steffes&Burgee,2009

betweenWOMandeWOM)

Wengetal.,2012

Informationsearch Informationoverload

Engeletal.,1969;

Vendorselection

Kiel&Layton,1981;

Influenceonproductselectionand
decisionsbyconsumers

Trusovetal.,2009
Helm,2000;

Performance

Variousrisks(a.o.customersasking

Dobeleetal.,2005

evaluation

somethinginreturnforviralactions)

2.1.3Typesofvirality,strategiesandmeasures:

Now we know what the drivers, motives and effects of wordofmouth, and
therefore virality, are, we can come to types of virality and strategies for
companies to get their product(s) or service(s) to go viral. And finally, we can
assess the different types of measures of virality. Let us start by examining the
typesofviralitythatexist.
Unfortunately, no research is available on the different types of virality that
exist. The only source available is TechCrunch (2012). This blog comes up with
eighttypesofviralitythatareofparticularinterestinthisresearchandbywhich
wecancategorizeproductsandservicesthataregoingviral.First,thereisinherent
virality;thisisthetypeofproductapersongetsnovaluefromunlessotherpeople
use it as well (e.g., ICQ, Skype). Second, there is collaboration virality; this is the
typeofproductapersonwillbenefitfromindividually.However,hegetsadditional

MastersThesis(BusinessStudies)CopyrightsJ.Sakkers(2012)Page20of84


valuefromcollaboratingwithotherpeople,soheinvitesthemtousetheproduct
as well (e.g., Dropbox, BaseCamp). Third, there is communication virality; this is
the type of product a person uses to communicate with people, some of which
might be potential customers. By raising awareness via this communication
channel (usually email), the product is being spread (e.g., Hotmail, Blackberry,
iPhone,iPad).Fourth,thereisincentivizedvirality;hereanincentiveusuallycash
oranotherbenefitinreturnforinvitingotherpeopletousetheproductisoffered
tospreadtheword(e.g.,Paypal,GiltGroupe,Dropbox).Fifth,thereisembeddable
virality;thisisthetypeofproductthatfacilitatesitscustomerstotakeapieceof
content and embed it anywhere on the Internet, with a link back to the original
website(e.g.,YouTube,SlideShare).Sixth,thereissignaturevirality;thisisthetype
of product customers are using either by embedding one of the features in their
own website or by referring people to a hosted page to use a specific feature. In
bothcases,whenthisfeatureisdisplayedtootherpeople,itincludesasignatureat
the bottom, which becomes a lead generator machine (e.g., SurveyMonkey,
Uservoice). Seventh, there is social virality; this is the type of product that
leverages existing social networks to spread the word about it. Typically a
customersignsupviaoneofhissocialnetworkIDs,andbyusingtheproduct,heis
tellinghissocialgraphaboutitautomatically(e.g.,Zynga,Spotify,Instagram).And
last but not least: eighth, there is pure wordofmouth virality; this is the type of
product customers spread the word about just because they enjoy it, it is free
and/or they think it is cool (e.g., Evernote, MailChimp). Table 4 provides an
overviewofalltypesofviralityincludingtworeallifeexamples.
Asstatedintheintroduction,wewillnarrowitdowntothreetypesofvirality
however:(1)inherentvirality(inherent,pureWOMe.g.,ICQ,Skype,Facebook);
(2) incentivized virality (collaboration, incentivized e.g., Paypal, Dropbox,
Groupon); and (3) networked virality (communication, embeddable, signature,
sociale.g.,Hotmail,Spotify,Instagram)however.Thereasonforthisisthatsome
oftheproductexamplesmentionedbeforeandhereunderuseahybridoftwoor
three types of virality. Dropbox, for example, uses collaboration virality and
incentivizedvirality.Therefore,thiscategorizationisproposed.

MastersThesis(BusinessStudies)CopyrightsJ.Sakkers(2012)Page21of84


Table4Overviewoftypesofvirality
Typeofvirality

Examples

Inherentvirality

Collaborationvirality

Communicationvirality

Incentivizedvirality

Embeddablevirality

Signaturevirality

Socialvirality

PureWOMvirality

MastersThesis(BusinessStudies)CopyrightsJ.Sakkers(2012)Page22of84


Seedingandincentives
After summing up the types of virality that can let companies products and
services go viral, it is interesting to emphasize which strategies successful
companiesusetogettheseproductsandservicestogoviral.Helm(2000),Dobele
etal.(2007)statethatitisunclearhowviralitysuccessfullycomesabout.Sweeney
etal.(2007)andGodesandMayzlin(2009)findthatlittleisknownabouthowto
enhance virality and what tactics a company should use. Wilson (2005) thereto
proposed six principles to get to the implementation of a successful viral
marketing campaign and, recently, Aral et al. (2011) find that two principal
strategies have emerged to leverage virality: seeding (or network targeting) and
incentives(referralstrategiesthatencouragepeeradoption).Thisresearchtakes
these two principal strategies and finally puts these to the test via an online
experiment.
Helm (2000) also stresses the importance of picking the right consumers that
shouldfirstpassontheviralmessage,asthecreationofviralnetworksdependson
thesepeople.Hinzetal.(2011)findthatmarketersachievethehighestnumberof
newcustomersiftheyseedthemessagetotherightpeople.Viafourexperiments
theyfindthataseedingstrategybasedonhubs(highdegreeseeding)orbridges
(highbetweenness seeding) is most effective, whereas hubs a small fraction of
peoplethataremanytimesbetterconnectedthanaverage(Watts,2004:p.250)
are used as the initial seeding point, and bridges the weak ties within peoples
networks are seeded. Seeding strategies based on the fringes (lowdegree
seeding)issignificantlylesseffectiveincausingvirality.
Araletal.(2011:p.3)furtherresearchedtheeffectivenessofthetwoprincipal
strategiesthathaveemergedtoleveragevirality,orsocialcontagionastheyliketo
call it. Most important, they find that incentive strategies dominate seeding
strategies [largely because] in addition to generating fewer total adoptions,
seedingisconstrainedbythesmallfractionofinfluencersthatexistinthenetwork.
Seeding more than 0.2% of the network is wasteful, because the gain from their
adoption is lower than the gain from their natural adoption (without seeding).
Moreover Aral et al. (2011) also find that incentives and seeding are
complementary.Finally,Watts(2004:p.261)statesthattheprobabilityofacting

MastersThesis(BusinessStudies)CopyrightsJ.Sakkers(2012)Page23of84


upon wordofmouth may change dramatically after the second or third (or
subsequent) exposure where the sudden jump occurs when the individuals
thresholdofbeliefisexceeded.Itisthereforeextremelyimportantthatpotential
customersareapproachedasoftenaspossible.

Measuresandvariables
Measures of virality have only been proposed the last decade (Reichheld, 2003;
Kalyanam et al., 2007; Hinz et al., 2011; Lawler, 2012; Skok, 2012I; 2012II),
probablyduetotheriseofsocialnetworksandthereforetheincreasingmagnitude
of impact of virality in these years. Nevertheless, Kermack and McKendrick
(Kermack & McKendrick, 1927; in Watts, 2004) were way ahead of their time
whentheycameupwiththeSIRmodelin1927.Thismodelcombinesthreebasic
classes:thesusceptible(S),theinfected(I),andtheremoved(R),whichallpeople
withinacertainpopulationcanbedividedinto.Theformula,N(t)=S(t)+I(t)+R(t),
ultimatelysumsuptothepopulationsize(=N).Althoughtheequationisoriginally
designed to quantify the spread of diseases, it can easily be applied to viral
campaigns: people that are within the target group but which have not been in
contact with the viral message yet, form the susceptible (S); people that have
become customers after receiving the viral message, are the infected (I); and
everyone that has been in contact with the viral message but did not become a
customerispartoftheremoved(R).
Almost a century later, Reichheld (2003) invented the Net Promoter Score
(NPS):amanagementtoolthatmeasurestheloyaltyofacompanyscustomersby
assessingthelikelihoodofthesetospreadthewordaboutthecompanysproducts
and services in opposite to measuring customer satisfaction. This measure is
proposed as a very simple tool that can be implemented easily by companies,
because it only entails a periodic survey amongst customers asking the question
Howlikelyisitthatyouwouldrecommendourcompanytoafriendorcolleague?
and the more positive answers (or the less negative answers) a company
receives on this question, the bigger its growth according to Reichheld (2003). A
company needs to send this survey to at least a hundred customers at a time,
resultinginaNPSofatmaximum+100(everybodyisapromoter)oratminimum

MastersThesis(BusinessStudies)CopyrightsJ.Sakkers(2012)Page24of84

100 (everybody is a detractor). A NPS of +50 is perceived as excellent for most


companies.
The only social network company that has ever opened up its data to neutral
scientists (i.e., scientists that are not paid by the company itself) up until now is
Plaxo, an online address book and social networking service. Kalyanam et al.
(2007) assessed the growth of this company during its heydays and reported on
results of implementing the socalled Viral Index that Plaxos management came
upwith:V=N*Cr.Thefactorofthenumberofinvitessentoutbyeachuser(N),the
conversionrateofthoseinvitesintonewusers(Cr)togetherformtheViralIndex
(V), whereas V=1 is the magic threshold, with V>1 causing exponential growth.
Plaxocameupwiththisviralequationtofocusonthemostimportantpartsofits
productandgettheseasviralaspossible,becauseitfoundthatitsviralenginedid
not grow automatically. Management of Plaxo then took a very datadriven,
adaptiveexperimentation approach to select among a few versions of product
features (Kalyanam et al., 2007: p. 81). They have looked well enough into the
three important factors that can be manipulated to create a viral epidemic
environment, message, and messengers (Kaplan & Haenlein, 2011), that are
coupled with the factors influencing viral growth, the number of invites and the

MastersThesis(BusinessStudies)CopyrightsJ.Sakkers(2012)Page25of84


conversion rate in the schematic overview of the implementation of Plaxos Viral
IndexbyKalyanametal.(2007:p.77)infigure5.
More recently, Hinz et al. (2011) reuses the fourdeterminant model of social
contagionbyVanderLansetal.(VanderLansetal.,2010;inHinzetal.,2011):SRi
=Ii*Pi*ni*wi.ThismodelcanbeseenasacombinationoftheSIRmodelandPlaxos
ViralIndexthatarediscussedabove,asitcombinestheinformationprobability(I),
theprobabilityofparticipating(P),theusedreach(n)andtheconversionrate(w).

Table5Theimpactofcycletimeonviralgrowth
U(0)initialcustomerset

10

10

10

Inumberofinvites

10

10

10

Cconversionrate

20%

20%

20%

pcycletime

5days

2days

1day

Kviralcoefficient

Customersafter20days

310

20,470

20,971,510

Ultimately, both Lawler (2012) and Skok (2012I; 2012II) propose a virality
formula that takes into account all facets of virality. They add the factor time to
Plaxos Viral Index (Kalyanam et al., 2007), which proves to be an extremely
important variable, and form an equation that lets us calculate the number of
peopleaproductorservicegoingviralwillreachintime.Duetothisflaw,Plaxos
ViralIndexassumesthateachcustomerwillcontinuetosendoutinvitesforevery
cycle.Whileinrealitythatisnotthecase(Skok,2012I).Cycletimehasdramatic
impactontheviralityofaproductorservice,though.Table5showstheimpactof
avaryingcycletimeviaacoupleofexamples.Thevariablesinitialsetofcustomers
(U(0)),theviralcoefficient(K,whichistheproductofthenumberofinviteseach
customer sends (I) and the rate at which invites convert new customers (C a
valuebetweenzeroandone):K=I*C)andthenumberofcyclestheinviteprocess
hasgonethrough(t/p,whichisthecurrent(orspecified)time(t)dividedbythe
amountoftimeittakesforanewcustomertosendhisbatchofinvitesortheviral
cycletime(p))togethermakeupforthisviralitymodel.

MastersThesis(BusinessStudies)CopyrightsJ.Sakkers(2012)Page26of84


NotethatsameaswithPlaxosViralIndex,theviralcoefficientmustbegreater
thanonetoachieveviralgrowth.Astheequationshows,themostimportantfactor
to increase growth is not the viral coefficient, however, but the viral cycle time
(c/t)whichshouldbemadeasshortaspossible.Thesecondmostimportantarea
to focus is the viral coefficient (K). Anything that you can do to increase the
number of invitations sent out, and the conversion rate, will have a significant
effectongrowth(Skok,2012I).Lawler(2012)alsostressestheimportanceof
loweringtheamountoftimeittakesforacustomertoinviteotherstotheproduct
or service over increasing the number of invites customers sen dor the rate at
whichinvitednoncustomersconvert.
To sum up, first we have looked into the foundations of the field of viral
marketingandthedifferencesandcommonalitiesbetweentraditionalandviral

Figure6ViralityformuladerivedfromLawler(2012)andSkok(2012I;2012II)

MastersThesis(BusinessStudies)CopyrightsJ.Sakkers(2012)Page27of84


marketingandwordofmouth.Second,wehavecategorizedthedriversofword
ofmouthintothefactorsenvironment,message,andmessengers.Third,wehave
categorized all motives behind spreading wordofmouth into the main drivers
selfinterest,multiplemotive,advocate,andaltruist.Fourth,wehaveexaminedthe
effects of wordofmouth on both companies and customers. Thereafter, we have
put together three sets of types of virality: inherent virality, incentivized virality,
and networked virality and found that there are two principal strategies for
companiestoimplementingaviralstrategy:seedingandprovidingincentives.Last
butnotleast,wewentthroughmeasuresandvariablesthataremostimportantin
thisfield,summedupinfigure6.

2.2Thestickandthecarrotincentivizingwordofmouth:

Second, the field of incentive systems is examined further. As mentioned before,


incentives give both senders and recipients of wordofmouth a motive for
spreadingtheword(see:Table2).Moreover,oneofthethreemaintypesofvirality
is incentivized virality, practiced by numerous companies (e.g., Dropbox and
Paypal). According to Aral et al. (2011), giving incentives defined as referral
strategiesthatencouragepeeradoptionisoneofthetwoprincipalstrategiesthat
haveemergedtoleveragevirality.Therefore,wewilltaketwotypesofincentives
astheindependentvariableinthesetupoftheexperiment.
Incentivizingwordofmouthisstillrelativelyunchartedterrain,whereasmost
contributionshavebeenmadeinthelasttwodecades.However,ClarkandWilson
already provided a definition: incentives are tangible or intangible rewards that
areofferedbyorganizationstoinfluencecustomeroremployeebenefits(Clark&
Wilson,1961).Wewillsticktothisdefinition;thoughuseitonlyinreferencewith
influencingcustomerbehavior.Buttle(1998:p.245)findsthatearlyon,companies
likeAmericanExpressandBritishTelecomwereattemptingtoincreasewordof
mouth by incentivizing customers to refer their friends and family. Lately, many
authors discussed the importance of providing incentives to customers for
successfullyrecommendingaproductorserviceinstimulatingviralmarketing

MastersThesis(BusinessStudies)CopyrightsJ.Sakkers(2012)Page28of84


Figure7Typesofincentivesandmaincontributors

(Biyalogorksyetal.,2001;HennigThurauetal.,2004;Dobeleetal.,2007;Godes&
Mayzlin,2009).
Incentives can influence intrinsic and extrinsic motives of consumers for
spreading the word (Buttle, 1998; Godes et al., 2005) and can lead to actual
increase of wordofmouth (Wirtz & Chew, 2002; HennigThurau et al., 2004; De
Bruyn&Lilien,2008;Godes&Mayzlin,2009;Bronner&DeHoog,2010;Berger&
Schwartz, 2011), and the intention of customers to recommend a product or
service(Dobeleetal.,2007;Ryu&Feick,2007).Companiescaninfluenceextrinsic
motivesasthesecomefromoutside,whileintrinsicmotivesaretheonesthatcome
from the consumer himself and therefore differ from person to person (Godeset
al., 2005). Intrinsic motives can be divided in: behavior, feelings, and the self
(Schutz, 1992; Ho & Dempsey, 2010). Incentives that influence extrinsic motives
can be divided in: material incentives, purposive incentives, and solidary
incentives. Material incentives are tangible rewards that are often monetary,
purposive incentives are intangible rewards related to the goals of the
organization, and solidary incentives are intangible rewards from the act of

MastersThesis(BusinessStudies)CopyrightsJ.Sakkers(2012)Page29of84


association (sociability, status, identification) (Clark & Wilson, 1961). Figure 7
provides an overview of types of incentives and main contributors in the field of
incentivesystems.
Incentive systems, or referral systems, have become a popular way to acquire
customersthelastdecade(Schmittetal.,2011).Thisispartiallyattributabletothe
digital nature of recommendations via electronic wordofmouth, which made it
possible to calculate the value that can be assigned to an eventual incentive
(Trusovetal.,2009).Akeyelementofcustomerrelationshipmanagement(CRM)
is that companies need to invest in retaining existing customers. Ryu and Feick
(2007: p. 84) therefore argue that an incentive system, or a referral reward
programastheycallit,canbeakeyCRMtoolbecauseinadditiontoitspotential
to attract new customers, it can improve retention by rewarding existing
customers.AccordingtoLubarsky(2005)thekeytoaworkingincentivesystemis
that:(1)thepotentialcustomerbelievesthatthedesiredbehaviorwillleadtothe
desired outcomes (i.e., the incentive will really be provider as promised by the
company); (2) the potential customer believes that these outcomes have positive
valueforhim;(3)thepotentialcustomerbelievesthatheisabletoperformatthe
desiredlevel.
Theeffectivenessandapplicationofincentivesneedsmoreresearchaccording
toButtle(1998:p.246).AfewyearsafterButtle(1998)statedthis,Wirtz&Chew
(2002) were the first who researched the effect of (financial) rewards on
stimulating wordofmouth behavior within different types of relationships. They
found that providing incentives indeed is an effective way to stimulare wordof
mouth that causes persuasive message that customers send peertopeer to be
morepositiveandpurchasetargeted.Theyalsofoundthatrewardsworkbestwith
wordofmouththatisdirectedtostrongties,whereasRyuandFeick(2007)found
the exact opposite during their experiments. They conclude that rewards only
work with stimulating wordofmouth that is directed to weak ties, and they
provideapossibleexplanationforthisintheexchangetheory:thesenderweighs
hispotentialbenefitsagainsttheriskofparticipatinginwordofmouth.Customers
perceive higher risk in recommending products or services to their strong
relations, and therefore incentives work only in increasing wordofmouth

MastersThesis(BusinessStudies)CopyrightsJ.Sakkers(2012)Page30of84


behavior that is directed to their weak ties. Tuk et al. (2009) subvene Ryu and
Feickinthismatter:iftherecipientknowsthatthesenderreceivesarewardfor
sending him a recommendation, he will guard himself against future persuasive
messagesofthisparticularsender,andthereforebelesslikelytorespondtofuture
advice.FriestadandWright(1994)alsofoundthatiftherecipientofthewordof
mouthknowsthatthesenderreceivesarewardforhisaction,hewouldtakethis
intoaccountandautomaticallyattachlessvaluetothepersuasivemessage.
However, the main effects of (financial) incentives are positive, and
implementingincentivesystemsisthereforehighlyrecommendedtocompaniesall
overtheworld.Schmittetal.(2011)arguethatreferredcustomershaveahigher
contribution margin, have a higher retention rate, and are more valuable in both
theshortandthelongrun.Theaveragevalueofareferredcustomeriseven16%
higherthanthatofanonreferredcustomer.
On the one hand we will test whether purposive and material incentives are
positively correlated with the number of invites a customer sends. On the other,
we will test whether purposive and material incentives are negatively correlated
withthecycletime,andthereforewouldincreasethemagnitudeoftheviraleffect.
Bytestingtheseincentives,wewouldliketofindouthowincentivesystemsfitinto
the virality formula (Lawler, 2012; Skok, 2012I; 2012II) mentioned above, thus
enhancingtheusefulnessandpossibleapplicationsoftheequation.Hereunderyou
willfindavisualoverviewofthefirstfourhypothesesthatwillbeputtothetestin
theexperiment:

MastersThesis(BusinessStudies)CopyrightsJ.Sakkers(2012)Page31of84

MastersThesis(BusinessStudies)CopyrightsJ.Sakkers(2012)Page32of84


2.3Theidealcustomerimportanceofnetworks:

Third, the field of network dynamics is examined further. As mentioned before,


messengers are one of three factors that can be used or manipulated to create a
viral epidemic (see: Table 1), and wordofmouth is of course in itself a facet of
networking.Moreover,oneofthethreemaintypesofviralityisnetworkedvirality,
practiced by numerous companies (e.g., Hotmail, Spotify, and Instagram).
AccordingtoAraletal.(2011),seedingdefinedasnetworktargetingisoneof
thetwoprincipalstrategiesthathaveemergedtoleveragevirality.Therefore,we
willtakethesizeofacustomersnetworkoneofthecharacteristicsofnetworks
andtestwhetherithaseffectonthevariableswithintheviralityformula.
Networks are of organized systems of relationships, and hence a network is
generally defined as a specific type of relation linking a defined set of persons,
objects,orevents(Nelson,1988;Szarka,1990;inDonckels&Lambrecht,1997).
Networkbased research is emerging as a new research area (Watts, 2004),
amongstothersbecauseofitsimportancetoentrepreneurialsuccess(Donckels&
Lambrecht,1997).Inaccordancewiththeresourcebasedview,entrepreneursuse
theirsocialrelationstheirnetworkinordertogettheresourcestheyneedto
launchandsucceedwiththeirbusiness(Jensen&Greve,2002:p.2).
According to Watts and Strogatz (1998), the selfproclaimed founders of the
research area of network dynamics, many realworld networks e.g., social
networks are smallworld networks. A smallworld network is a type of
mathematical graph in which most nodes are not neighbors of one another, but
mostnodescanbereachedfromeveryotherbyasmallnumberofhopsorsteps
(Wikipedia,2012III).Forexample,onaverageapersonisconnectedtoanyother
person in the world through six steps. The structures of different types of
networks are similar: roadmaps, electric power grids, and social influence
networks for example, are all sparse, decentralized and neither completely
orderednorcompletelyrandom(Watts,1999).
Moreover,thestructuresofnetworkshavehugeimplicationsforthecollective
dynamics of a system. For example, the level of randomness in a network has a
strongimpactonboththeextentofanepidemic,andthetimetheepidemictakes

MastersThesis(BusinessStudies)CopyrightsJ.Sakkers(2012)Page33of84


Figure8Enrichingtheviralityformulawithseedingofhubs,fringesandbridges

to spread (Watts & Strogatz, 1998). Kleinberg (2000) therefore claims that
network structure is important locally as well as globally, for each of us. And
within network structures, hubs, fringes, and bridges are extremely important
(Hinzetal.,2011).AccordingtoWatts(2004:p.250)inmanyrealworldnetworks
including social networks, the majority of nodes, normal people, have less than
averagedegree(i.e.,networksize),whileasmallfractionofhubs,connectors,are
many times better connected than average. These hubs and the nature and
arrangement of network ties may play a crucial role in the chance of successful
distributionofviralepidemics.
In that respect, the BarabsiAlbert model (Barabsi & Albert, 1999) on the
concept of scalefree networks is of particular interest. Barabsi and Albert
mappedapartoftheInternetbackinanexperimentbackin1998,andfoundthat
itsstructuredoesnotmatchthewidelyadoptedmodelofrandomconnectivity:(1)

MastersThesis(BusinessStudies)CopyrightsJ.Sakkers(2012)Page34of84


thenetworkgrowsbyatleastonenetworkhub(i.e.,node)perperiodoftime,and
(2) these new network hubs connect with existing hubs via two links, preferring
hubs that are more connected (i.e., have more links with other hubs) the so
called preferential attachment condition. In other words, a few highly connected
hubswitharelativelylargenumberofconnectionsbecomemoreimportantas
a network grows, eventually keeping together entire networks. Subsequently,
Bianconi and Barabsi (2001) proposed the BB model, further deepening the
conceptofscalefreenetworksbyexplaininghowfitnessofhubscausesomehubs
togrowfasterthanothers;anadaptationofthesurvivalofthefittestadage.These
fitterhubsareverylikelytobetheequivalentofconnectorsGladwell(2000:p.
3859)talksaboutinhisbookonthetippingpointofvirusesviasocialnetworks.A
visualizationofthedifferencebetweenrandomnetworksandscalefreenetworks
isaddedinFigure9.

Figure9Randomnetworksversusscalefreenetworks

Besidesconnectors,hubsaresometimesalsocalledopinionleadersorinfluencers
(Godes & Mayzlin, 2009). Fringes are the opposites of hubs, and bridges are the
people that connect different types of networks, or worlds (see: Figure 8). While
influencers are almost always referred to as highly important in achieving viral
epidemics(Watts,2004;Raz&Gloor,2007;Hinzetal.,2011),theyareonlyuseful

MastersThesis(BusinessStudies)CopyrightsJ.Sakkers(2012)Page35of84


whentheyareloyalcustomersofthecompanysproductorservicebecausethey
risk losing their status when they become too innovative, recommending
productsthatthoseintheirsocialcirclearenotreadytoadopt(Godes&Mayzlin,
2009:p.725).ValenteandDavis(1999)andNormanandRussell(2006)statethat
one of the defining aspects of the utility of influencers and followers both
customersandnoncustomersistheirnetworksizeandthetiestrengthbetween
themandtheirrelations.
In accordance with both Valente and Davis (1999) and Norman and Russell
(2006),Wengetal.(2012:p.6)emphasizethatthenetworksizeandtiestrength
are of particular importance to assess the probability of a message going viral. A
lotofresearchhasbeendoneontheeffectsthat(distributionvia)strongandweak
ties have on wordofmouth behavior. Tie strength is the strength of the relation
between to people (Brown et al., 2007), in this research the sender and the
recipient of wordofmouth. Huberman et al. (2008) strongly encourage research
in this area; they claim that every link between any two people is different and
that,withtheentranceofonlinesocialnetworks,alinkdoesnotsimplyimplyan
interaction between them. Thus we need to find the hidden social network: the
onethatmatterswhentryingtorelyonwordofmouthtospreadanidea,abelief,
oratrend(Hubermanetal.,2008:p.8).Table6theretoprovidesanoverviewof
all literature on tie strength and their effect on wordofmouth behavior,
categorizedbyresearcharea.
Manyauthorsarguethat,inadditiontotiestrength,networksize(alsoreferred
to as network degree) is (one of) the most important network property
(Granovetter, 1973; Borgatti & Jones, 1998; Jensen & Greve, 2002; Raz & Gloor,
2007). As mentioned before, Hinz et al. (2011) experimented by using hubs,
fringes,andbridgesindifferentseedingstrategies,andhefoundthathubsarethe
beststartingpointforasuccessfulseedingstrategy.Tosumup,alotresearchhas
been conducted on tie strength (see: Table 6) but despite of its importance
network size has not been the subject of many articles and research papers.
Therefore, the effectiveness of network size on virality will be assessed in the
experiment.

MastersThesis(BusinessStudies)CopyrightsJ.Sakkers(2012)Page36of84


Table6Overviewofliteratureonstrongandweaktiesinenhancingwordofmouth
Author

Researcharea

Contribution

Bansal&Voyer,2000

Virality

Tiestrengthsenderandrecipient
positivelycorrelatedwithpurchase
intentionsrecipient

Goldenbergetal.,2001

Virality

Effectofstrongtiesdiminishesas
networksizedecreases;whensmall
network,impactofweaktiesonWOM
strongerthanstrongties

Wirtz&Chew,2002

Virality,

RewardsworkbestwithWOMto

incentives

strongties,ifcustomersaresatisfied
withproduct/service

Ryu&Feick,2007

Virality,

RewardsforsenderworkbestonWOM

incentives

forweakties,rewardsforrecipientfor
strongties

Brownetal.,2007

Virality,

Tiestrengthinfluencespurchasing

networks

intentions,howevernotwitheWOM
recipientstestthewebsiteforits
credibility,notthesender

Sweeneyetal.,2007
Ryu&Han,2009

Virality,

Tiestrengthinfluencespurchase

networks

intentionsforservices

Virality,

Recommendationbystrongtie:more

networks

effectonpurchaseintentionthatthatof
weaktie

Steffes&Burgee,2009

Virality,

Weaktiesourcesaresometimesrated

networks

asmoreinfluentialthatstrongtie
referralsources

Walsteretal.,1973

Networks

Equitytheory:especiallyweakties
shouldhonorquidproquo

Clarketal.,1986

Networks

Strongtiesbasedoncaring,therefore
noreciprocationneeded

MastersThesis(BusinessStudies)CopyrightsJ.Sakkers(2012)Page37of84


We will test whether network size is positively correlated with the number of
invites a customer sends. Also, we will test whether network size is negatively
correlatedwiththecycletime,andthereforewouldincreasethemagnitudeofthe
viraleffect.Bytestingthesenetworkcharacteristics,wewouldliketofindouthow
networkdynamicsfitintotheviralityformula(Lawler,2012;Skok,2012I;2012
II) mentioned above, thus enhancing the usefulness and possible applications of
theequation.Hereunderyouwillfindavisualoverviewofthelasttwohypotheses
thatwillbeputtothetestintheexperiment:

MastersThesis(BusinessStudies)CopyrightsJ.Sakkers(2012)Page38of84


2.4Conceptualframework:

Theconceptualmodel(see:Figure10)summarizesthetheoreticalframeworkand
combinesthehypothesesthatwillbetestedintheexperiment.Thefoundationof
thisframeworkistheviralityformulaofLawler(2012)andSkok(2012I;2012II),
an equation combining the initial set of customers (U(0)), the number of invites
eachcustomersends(I),therateatwhichinvitesconvertnewcustomers(C),the
viral coefficient (K = I*C), and the number of cycles the invite process has gone
through(t/p).Byassessingtheinfluenceofpurposiveandmaterialincentives,and
network size on the number of invites and the cycle time, the purpose is to
enhance the usefulness and possible applications of the virality formula to
ultimately provide insights in the way different types of incentives and customer
profilesinfluenceviralityofproductsandservices,andcomeupwithmanagerial
implications providing startups and other entrepreneurs with a framework for
makingtheirproductsandservicesgoviral:

Figure10Conceptualframeworkfortestingincentivizedvirality

MastersThesis(BusinessStudies)CopyrightsJ.Sakkers(2012)Page39of84


Youmightaswellnotchangeitatallifyourenotgoingtomeasuretheimpactof
thatchange.

MarcPincus

3.Researchmethodanddatacollection

In this chapter a complete overview of the research methodology and data


collection is provided. Consecutively, variables, data collection and reliability are
discussed. But first, methodology and research design will be handled. Data is
collectedviathewebsiteofVictorMundifromApril30thuntilMay28thof2012.All
visitors that ultimately registered for Victor Mundi in this period of time
participatedintheexperiment.Duringthisperiod,10,151peoplevisitedthefirst
stepofthesignuppageofVictorMundiandtheplatformfinallywelcomed2,335
newcustomers.Thesecustomersweresubjectoftheexperimentofthisresearch
paper.
Intheexperiment,wemeasuretheviraleffect(boththenumberofinvitessent
per customer, which is part of the viral coefficient, and the cycle time) under
different scenarios of incentivizing and with participants with different network
sizes, within an actual web application that is available via any major Internet
browser Apple Safari, Google Chrome, Microsoft Internet Explorer, Mozilla
Firefox,orOperacombinedwithaninternetconnection.Thedifferentscenarios
are preset by altering the content of the share or invite message and the context
andtextexplanationoneachpageitself.Duringthetimetheexperimentruns,each
participant goes through all pages of the experiment and therefore will undergo
each scenario of incentivizing. We will track every action and event that occurs
(enter,share,invite,and/orexitpage).
Inordertocollectasmuchparticipants,andtherewithdata,aspossible,weset
upanexperimentwithinanexistingproduct:VictorMundi,thedigitalworkspace
forentrepreneurs(availableinDutchonlyatthetimeoftheexperiment).During
theactiveperiodoftheexperiment,eachnewcustomerofVictorMundimustgo

MastersThesis(BusinessStudies)CopyrightsJ.Sakkers(2012)Page40of84


through three wizards1 (See: Figure 11), each with three or four steps: a signup
wizard,animportwizard,andacreatewizard.

Figure11SequenceofwizardswithinVictorMundi

3.1Methodologyanddesign:

The variables within the conceptual framework that will be tested i.e., the
numberofinvitesandthecycletimearemeasuredbestbyobservingbehaviorof
participants, instead of interviewing people. Generally, the objective of
quantitativeresearchistodevelopandemploytheoriesand/orhypotheses;which
isthecaseforthisresearch.Therefore,aquantitativeapproachispreferredover
qualitativeresearch.
According to Hinz et al. (2011) it is necessary to conduct experiments to
compareseedingstrategiesinrealworldsituations(e.g.,theeffectofthenetwork
sizeofhubs).GodesandMayzlin(2009)stresstheimportanceofanexperimental
approachtoanswerquestionsontheeffectivenessofincentivesystemsincreating
virality. Moreover, an experiment is the most suitable way to verify, falsify, or
establish the validity of hypotheses in the conceptual framework for proving the
theoreticalframework(Ryu&Feick,2007).Therefore,anexperimentalapproach
ispreferredoversurveyresearch.

1See:AppendixIVDesignofexperimentfortheactualsetup,screenshots,and
theselectquerythatisperformedtoretrievedatafromtheVictorMundidatabase.

MastersThesis(BusinessStudies)CopyrightsJ.Sakkers(2012)Page41of84


3.2Variables:

Thedependentvariablesarethenumberofinvitessentpercustomer(I)andthe
cycle time (t/p). The product of the number of invites sent per customer (I) and
the conversion rate of these invites (C) forms the viral coefficient (K), which is
showninFigure12hereunder.Thisnumberneedstobehigherthan1toachieve
exponentialgrowth(Lawler,2012;Skok,2012I;2012II).Inordertoachieveviral
growth, only one number is more important than the viral coefficient: the cycle
time.Thecycletimeisthecurrent(orspecified)time(t)dividedbytheamountof
time it takes for a new customer to send his batch of invites (p), the latter being
alsoreferredtoastheviralcycletime.Thisneedstobeaslowaspossibleinorder
toachieveahigherpaceofviralgrowthforthespecificproductorservice(Skok,
2012I; 2012II). The conversion rate is not put to the test, because the research
environment the Victor Mundi platform relatively unknown, which can cause
anexorbitanteffectonpossibleoutcomes.

Figure12Equationfortheviralcoefficient

Figure13ViralformulabyLawler(2012)andSkok(2012)

The independent variables are purposive incentives, material incentives, and


network size. Purposive incentives are intangible rewards related to the goals of
an organization, while material incentives are tangible rewards that are often
monetary (Clark & Wilson, 1961). Watts (2004) and Hinz et al. (2011) find that
thesehubsareofhighimportanceinordertocreateviralepidemics.Therefore,we
willtestwhetherthesevariablesareofinfluenceonthenumberofinvitessentper

MastersThesis(BusinessStudies)CopyrightsJ.Sakkers(2012)Page42of84


customerandthecycletime.Lastbutnotleast,technology,productattractiveness,
andtheworkingofsharing/invitingmechanismsaregivenandcannotbechanged
inthisexperimentthesocalledceterisparibuscondition.

Hypotheses

In Table 7 an overview of the six hypotheses that will be tested in this research
paper, is provided including the independent variables, the dependent variables,
andtheassumedcorrelationbetweenthese.

Table7Overviewofhypotheses

Independent

Dependent

Correlation

H1

Purposiveincentives

Numberofinvitessentpercustomer

Positive

H2

Purposiveincentives

Cycletime

Negative

H3

Materialincentives

Numberofinvitessentpercustomer

Positive

H4

Materialincentives

Cycletime

Negative

H5

Networksize

Numberofinvitessentpercustomer

Positive

H6

Networksize

Cycletime

Negative

3.3Datacollection:

Westrivedtoletcollectanumberof1,000to10,000participants,andultimately
succeeded: 2,335 customers signed up for Victor Mundi between April 30th and
May 28th. However, only 313 completed all wizards so that is the actual sample
size.Thissamplewill(a)decreasetheimpactofchanceeventsanderror,and(b)
increasethereliabilityofresults.Byhavingthisnumberofcustomerswithinour
sample size, we will be able to generalize the experimental results to a certain
extent, which adds substantial value to this research paper. Table 8 provides an
overview of the dropouts from people that visited the first step of the signup

MastersThesis(BusinessStudies)CopyrightsJ.Sakkers(2012)Page43of84


wizard and people that actually signed up, and shows that 303 of the 2,335 new
customersofVictorMundiultimatelyfinishedallthreewizards.
Each participant first goes through part 4 of the signup page, which contains
thefirstcalltoactionwithoutanincentive.Thisisthebaselineorcontrolgroup.
Next all participants go chronically through pages with calltoactions #2 and #3
wheretheygetdifferentincentives.Allparticipantsareexposedtoallincentives,
as can be seen in the table hereunder so they are all part of both the
experimental and the control group. Three calltoactions are added to the three
wizards,eachonecontainingashareorinvitebuttonthatparticipantscaneasily
click on. Both a share and an invite action are counted as invites sent by a
customer.Thesenumbersaremultipliedbythenumberofpeopletheparticipant
sharethemessagewithorsendstheinviteto,sowecanthenumberofinvitessent
percustomer(I).

Table8Overviewofnumberofparticipantsinexperiment

Viawebsite

Viainvite

Signupstep1(inviteonly)

N/A

7,917

Signupstep2

2,234

3,219(40,7%)

Signupstep3

469(21,0%)

2,454(76,2%)

Signupstep4(CTA#1)

211(45,0%)

2,124(86,6%)

Wizardstep1

2,335

Wizardstep2(CTA#2)

1,761(75,4%)

Wizardstep3

1,835(78,6%)

Createstep1

1,785(76,4%)

Createstep2

901(38,6%)

Createstep3(CTA#3)

313(13,4%)

Createstep4

303(13,0%)

Moreover, date stamps for the time of registration and each share or invite are
registered as YYYYMMDD HH:MM:SS.SSS (including milliseconds) so we can
calculatethetimebetweentheregistrationofeachparticipant,andthemomentof

MastersThesis(BusinessStudies)CopyrightsJ.Sakkers(2012)Page44of84


sharing or inviting: the cycle time that is calculated is rounded to minutes. The
networksizeiscalculatedbycountingthenumberofcontactsimported(theseare
automatically deduplicated by Victor Mundi). If a customer does not import his
contacts to Victor Mundi, the database stores 0. In order to be able to take the
observationsoftheseparticipantsintoaccountintestinghypotheses5and6,we
presume these networks to exist of 150 contacts (the average network size,
accordingtoDunbar(1993)andGladwell(2000:p.3859)).Anoverviewoftheof
calltoactions and incentives, and the number of participants and actual actions
performedisprovidedinTable9hereunder:

Table9Overviewofincentives

Page2

Incentive

Participants

Actions

#1

Signupstep4

Noincentive(controlgroup)

2,335

20(0.9%)

#2

Wizardstep2

Materialincentive

1,761(75,4%)

794(34.0%)

#3

Createstep3

Purposiveincentive

313(13,4%)

187(8.0%)

3.4Reliability:

For this research we will use the database of Victor Mundi. Victor Mundi runs
several Microsoft SQL Servers with SQL Server 2008 that cannot be accessed by
unauthorizedpersonnel.TheserversarelocatedatthedatacentreofEvoSwitchin
Haarlem, which is one of the biggest and most reliable in The Netherlands. The
Victor Mundi team installed several protocols for accessing the database; only a
fewseniorprogrammersthathavehandedoveracertificateofgoodconductand
arehighlyexperiencedareauthorizedtodoso,mostarenot.
Moreover, another operation that is run under the same holding company as
Victor Mundi, serves many pension funds and insurers (internationally).
Governmental policies on data security and privacy issues make it necessary for

2See:AppendixIVDesignofexperimentfortheactualsetup,screenshots,and
theselectquerythatisperformedtoretrievedatafromtheVictorMundidatabase.

MastersThesis(BusinessStudies)CopyrightsJ.Sakkers(2012)Page45of84


this operation to treat user data with the utmost care. Victor Mundis users also
benefit of the implementation of these policies, because it is necessary (and cost
efficient)forallcompaniesunderthesameholdingtohavethesameprotocolsand
proceduresimplemented.
Also,VictorMundisprivacypolicythateverynewcustomeragreestobefore
actually finishing the registration procedure states that Victor Mundi can use
anonymoususerdatafor(academic)researchpurposeswithouthavingtoaskfor
permissioneverytimeitconductsresearch,toimproveitsproductsandservices.
Alluserdataisencryptedinthedatabaseandcommunicationaswellthankstothe
Hypertext Transfer Protocol Secure (HTTPS), which makes sure a visitor or
customerofVictorMundidoesnothavetoworryabouthissecurity.
The numbers are highly reliable, because we have run a double check on the
data retrieved from each step of the signup wizard, the import wizard, and the
createwizard.First,weimplementedourownscriptthatregisterseveryvisitand
click of each of Victor Mundis customers and analyze the data from this source.
Second, we run a Google Analytics script and analyze the same pages via the
GoogleAnalyticsdashboard.

MastersThesis(BusinessStudies)CopyrightsJ.Sakkers(2012)Page46of84


Realhonorablemenshowstaminaandalwaysfinishwhattheyhavestarted.Make
meproud,finishyourMastersDegree.

AlbertusW.A.Sakkers

4.Results

Inthischaptertheresultsoftheexperimentconductedforthisresearchpaperare
provided. Consecutively, the kind of data, statistical procedures, and testing of
hypothesesarediscussed.

4.1Kindofdata:

Duringtheexperimentwemeasureeachshareandinviteoftheparticipants,and
thenumberofcontactsinhisaddressbook,whichtheycanimportviaanimport
tool that is provided in the first steps of the import wizard on the Victor Mundi
platform (Microsoft Outlook, LinkedIn, Google Contacts, Facebook, Yahoo! and
*.csv/*.xls(x) files are supported). The contacts found will be deduplicated by
(unique) name and email address automatically, so we get a clean number for
measuringthesize/degreeofnetworkforallparticipants.Moreover,datestamps
ofbothtimeofregistrationandtimesofsharingandinvitingarecollected,sowe
canmeasurethetimebetweenregistrationdateandshare/invitedates.

4.2Statisticalprocedures:

Inordertotestthehypotheseswithintheconceptualframeworkofthisresearch,
twostatisticalproceduresareneeded:ontheonehand,thetwosidedttestneeds
tobeperformedforhypotheses1,2,3,and4,becausethesetsofobservationsthat
are compared for these are (partially) discrete and unequal. On the other,
Pearsonscorrelationcoefficientisusedtocalculatethecorrelationinordertotest
hypotheses5and6,becausebothvariablesareofcontinuousdatameasuredonan

MastersThesis(BusinessStudies)CopyrightsJ.Sakkers(2012)Page47of84


intervalscale.Figure14visualizesthefindingsoftheconductedexperiment,anda
summaryofallfindingsispresentedintable10:

Table10Resultsfortestedhypotheses

Method

Accepted

Meaning

H1

Twosidedttest

Yes

Statisticallysignificant(P=0.0276)

H2

Twosidedttest

Yes

Statisticallysignificant(P=0.0198)

H3

Twosidedttest

No

Notquitestatisticallysignificant(P=0.0783)

H4

Twosidedttest

Yes

Extremelystatisticallysignificant(P<0.001)

H5

Pearsons

Yes

Verystatisticallysignificant(P=0.0071)

H6

Pearsons

No

Extremelystatisticallysignificant(P<0.001)

H1:Purposiveincentivesarepositivelycorrelatedwiththenumberofinvitessentper
customer.

ThetwotailedPvalueequals0.0276,whichmeansthathypothesis1isaccepted:
purposiveincentivesarepositivelycorrelatedwiththenumberofinvitessentper
customer.Byconventionalcriteria,thisdifferenceisconsideredtobestatistically
significant.ThemeanofthetwosettingsNoincentiveandPurposiveincentive
equals145.20invites,andthe95%confidenceintervalofthetwosettingsranges
from16.20to274.19invites.Theintermediatevaluesusedincalculationsaret=
2.2193,df=205,andsed=65.424.Themeans,standarddeviation,standarderror
ofthemean,andsamplesizeareshowninTable11.
The accepting of this hypothesis means that online businesses providing
purposiveincentivestotheircustomersintangiblerewardsrelatedtothegoalsof
anorganization(Clark&Wilson,1961),thustheparticularcustomerthatusesthe
web or mobile application will induce a higher number of invites sent per
customer, an important part of the viral coefficient (number of invites sent per
customer times the conversion rate of these invites) described by Lawler (2012)
andSkok(2012I;2012II).

MastersThesis(BusinessStudies)CopyrightsJ.Sakkers(2012)Page48of84


Table11Datafortwosidedttesthypothesis1

Noincentive

Purposiveincentive

Mean

1.50

146.70

SD

1.00

291.95

SEM

0.22

21.35

20

187

H2:Purposiveincentivesarenegativelycorrelatedwiththecycletime.

ThetwotailedPvalueequals0.0198,whichmeansthathypothesis2isaccepted:
purposiveincentivesarenegativelycorrelatedwiththecycletime.Byconventional
criteria,thisdifferenceisconsideredtobestatisticallysignificant.Themeanofthe
twosettingsNoincentiveandPurposiveincentiveequals7,753.44minutes(=
approximately 5.4 days), and the 95% confidence interval of the two settings
ranges from 1,245.59 to 14,261.30 minutes (= approximately 0.9 to 9.9 days).
The intermediate values used in calculations are t = 2.3490, df = 205, and sed =
3,300.792.Themeans,standarddeviation,standarderrorofthemean,andsample
sizeareshowninTable12.
The accepting of this hypothesis means that online businesses providing
purposive incentives to their customers will induce a significantly shorter cycle
time, which makes up an important part of the viral growth equation of Lawler
(2012)andSkok(2012I;2012II)thatdescribeshowviralityforwebandmobile
applicationscanbesetup.Theresultsoftestinghypotheses1and2togethertell
usthatprovidingpurposiveincentivesisverybeneficialforanyorganizationthat
wants to create viral growth for their web or mobile application. Organizations
shouldthereforefindoutwhatreallydrivestheircustomers,andtrytousethese
insights by processing these drivers in the messages that are sent via share or
invite mechanisms within the particular web or mobile application. Ideally,
customers promote themselves in this message, instead of simply spreading the
wordaboutthewebormobileappitself.Onlinecompetitionsoftenusethistypeof

MastersThesis(BusinessStudies)CopyrightsJ.Sakkers(2012)Page49of84


incentivizing:peoplethatparticipateandwanttoapplyfortoprankingsareasked
toinviteotherstovoteforthem.

Table12Datafortwosidedttesthypothesis2

Noincentive

Purposiveincentive

Mean

21,368.45

13,615.01

SD

30,237.17

11,115.96

SEM

6,761.24

812.88

20

187

H3:Materialincentivesarepositivelycorrelatedwiththenumberofinvitessentper
customer.

The twotailed P value equals 0.0783, which means that hypothesis 3 cannot be
accepted: although we can see certain differences, we cannot conclude that
material incentives are positively correlated with the number of invites sent per
customer. By conventional criteria, this difference is considered to be not quite
statisticallysignificant.ThemeanofthetwosettingsNoincentiveandMaterial
incentive equals 107.84 invites, and the 95% confidence interval of the two
settings ranges from 12.37 to 228.05 invites. The intermediate values used in
calculations are t = 1.7626, df = 812, and sed = 61.185. The means, standard
deviation,standarderrorofthemean,andsamplesizeareshowninTable13.
Thedecliningofthishypothesismeansthat,incontrasttopurposiveincentives,
material incentives tangible rewards that are often monetary (Clark & Wilson,
1961)donotsignificantlyincreasethenumberofinvitessentpercustomer.But
note: This experiment does not test whether material incentives work in
increasing wordofmouth per se; the difference between the two means clearly
statethatprovidingmaterialincentiveshavesomeeffect.Theexperimentissetup
to test whether the average number of invites sent per customer is affected by
providing material incentives, because that is an important part of the viral
coefficientmentionedbefore.Unfortunately,assaid,thatcannotbeconcluded.

MastersThesis(BusinessStudies)CopyrightsJ.Sakkers(2012)Page50of84


Table13Datafortwosidedttesthypothesis3

Noincentive

Materialincentive

Mean

1.50

109.34

SD

1.00

273.46

SEM

0.22

9.70

20

794

H4:Materialincentivesarenegativelycorrelatedwiththecycletime.

The twotailed P value is less than 0.0001, which means that hypothesis 4 is
accepted: material incentives are negatively correlated with the cycle time. By
conventional criteria, this difference is considered to be extremely statistically
significant. The mean of the two settings No incentive and Material incentive
equals17,730.15minutes(=approximately12.3days),andthe95%confidence
interval of the two settings ranges from 13,424.97 to 22,035.33 minutes (=
approximately9.3to15.3days).Theintermediatevaluesusedincalculationsare
t=8.0914,df=812,andsed=2,191.246.Themeans,standarddeviation,standard
errorofthemean,andsamplesizeareshowninTable14.
The accepting of this hypothesis means that online businesses providing
materialincentivestotheircustomerswillinduceasignificantlyshortercycletime,
whichmakesupanimportantpartoftheviralgrowthequationofLawler(2012)
andSkok(2012I;2012II).Theresultsoftestinghypotheses3and4togethertell
us that providing material incentives is only affecting the cycle time: current (or
specified) time (t) divided by the amount of time it takes for a new customer to
sendhisbatchofinvites.Moreover,theeffectofmaterialincentivesondecreasing
cycletimeseemsmuchlargerthantheeffectofpurposiveincentiveswhichmight
be a reason for online businesses to opt for material incentives over purposive
onesintheirtradeoff,ascycletimeisofmuchlargereffectontheviralgrowthofa
webormobileapplicationthanisthenumberofinvitessentpercustomer.

MastersThesis(BusinessStudies)CopyrightsJ.Sakkers(2012)Page51of84


Table14Datafortwosidedttesthypothesis4

Noincentive

Materialincentive

Mean

21,368.45

3,638.30

SD

30,237.17

8,602.91

SEM

6,761.24

305.31

20

794

H5:Networksizeispositivelycorrelatedwiththenumberofinvitessentpercustomer.

Pearsons

correlation

coefficient equals 0.626 ( =


0.63) with N = 1,597. This
means that hypothesis 5 is
accepted: network size is
positively correlated with the
number of invites sent per
customer. The twotailed P
value equals 0.0071; by conventional criteria, this difference is considered to be
very statistically significant. The intermediate values used in calculations are t =
2.6916, df = 3,192, and sed = 15.243. The means, standard deviation, standard
errorofthemean,andsamplesizeareshowninTable15.
The accepting of this hypothesis means that online businesses indeed benefit
from deploying network hubs in their efforts of increasing the number of invites
sentpercustomers:scalefreenetworksinoptimaforma.

Table15Datafortwosidedttesthypothesis5

Networksize

Numberofinvitessentpercustomer

Mean

202.56

161.53

SD

233.66

562.53

SEM

5.85

14.08

1,597

1,597

MastersThesis(BusinessStudies)CopyrightsJ.Sakkers(2012)Page52of84


H6:Networksizeisnegativelycorrelatedwiththecycletime.

Pearsons

correlation

coefficientequals0.050(=
0.05) with N = 1,089. This
means that hypothesis 6
cannot be accepted: based on
the observations, only a very
mild negative correlation is
found between network size
and the cycle time. However, the twotailed P value is less than 0.0001; so by
conventional criteria, this difference is considered to be extremely statistically
significant. The intermediate values used in calculations are t = 13.a5000, df =
2,176, and sed = 285.029. The means, standard deviation, standard error of the
mean,andsamplesizeareshowninTable16.
The declining of this hypothesis means that online businesses should not
countonnetworkhubsintheireffortsofdecreasingtheirappscycletimeinorder
to get it to go viral. This result contradicts the preconception of network hubs
being innovators or first movers: customers with an average (or lower than
average) number of people within their network are as important in decreasing
thecycletimeofawebormobileapplication.

Table16Datafortwosidedttesthypothesis6

Networksize

Cycletime

Mean

201.55

4,049.45

SD

224.47

9,403.27

SEM

6.80

284.95

1,089

1,089

MastersThesis(BusinessStudies)CopyrightsJ.Sakkers(2012)Page53of84


Anyonewhothinkstheskyisthelimithaslimitedimagination.

Anonymous

5.Discussionandlimitations

By simply comparing the means of each data set, we can conclude that
incentivizingworksinordertomaximizethenumberofinvitessentpercustomer
andtominimizethecycletime.Whereasthefirstisanimportantpartoftheviral
coefficient,thelatterhasanexponentialimpactontheviralityofamobileorweb
application. Moreover, it does matter which kind of customer a mobile or web
applicationattractsinordertogoviral.Hubs,i.e.asmallfractionofpeoplethatare
many times better connected than average (Watts, 2004: p. 250), are very
important in order to increase the number of messages sent per customer. We
have not found evidence for a correlation between network size and cycle time
however. A combination of both seeding and incentivizing strategies are highly
recommendedtocompaniesthatwanttomaketheirappgoviral.
Thisresearchfocusesonpartsoftheviralequationandthereforedoesnottake
other findings into account. Nevertheless, one special mention goes out to the
striking effect of both purposive and material incentives on triggering sharing or
inviting behavior by customers. In testing the hypotheses, the number of invites
sent per customer are calculated and used in order to come to the conclusions
hereunder(See:Figure14).However,thedifferenceinpercentagesofparticipants
taking action in the control group and the experimental group vary widely: only
0.01%oftheparticipantsperformedashareorinviteactioninthesettingwithout
an incentive, while respectively 45.1% and 59.1% of the participants performed
such action in the settings including material and purposive incentives.
Therewithal, as mentioned before, the number of invites sent per customer is
positivelycorrelatedwithpurposiveaswellasmaterialincentives.

MastersThesis(BusinessStudies)CopyrightsJ.Sakkers(2012)Page54of84


Figure14Conceptualframeworkfortestingincentivizedvirality(outcomes)

5.1Potentialimpact:

The research question that is central to this research paper is What effect do
(differenttypesof)incentivesandthesizeofcustomersnetworkshaveontheviral
success of a mobile or web application?. By conducting extensive research and
testing hypotheses in an experimental setup we have come up with answers to
that main question. First of all, each Internet startup should peruse the model of
incentivized virality for products and services as proposed in the theoretical
framework,becausetheeffectsofincentivizingcustomerswiththerighttriggersat
the right time can have incredible impact. But, like Kaplan and Haenlein
emphasize,oneneedstoberealistic.Eventhemostsuccessfulbuzzcannotheala
bad product, inappropriate price, or insufficient distribution. To reveal its true
potential, viral marketing needs to be accompanied by changes in the rest of the
marketing mix (Kaplan & Haenlein, 2011: p. 260). The results of this research
must be seen in this broader picture; making an app go viral only works if
combinedwithanutmostbalancedmarketingmixwithaproductthatfulfillsreal
customerneedsorremediesrealcustomerpainsatitscore.
TheresultsoftheexperimenthaveansweredthequestionsraisedbyGodesand
Mayzlin (2009: p. 738) in their article: What should the firm do to encourage

MastersThesis(BusinessStudies)CopyrightsJ.Sakkers(2012)Page55of84


peopletogooutandtellothersaboutthefirm?,alsosuggestingthatanumberof
strategies, among which incentivizing, should be tested via an experimental
approach.Ourresultsprovethatmaterialincentivesworkreallywellindecreasing
cycle time, while providing purposive incentives works really well in both
increasing the number of invites sent per customer and decreasing cycle time,
which is of critical importance to creating viral growth for a certain product or
service just like Aral et al. (2011) found in their studies. Also, we found that a
seeding strategy that is focused on hubs or connectors can work really well in
ordertomaximizethenumberofinvitessentpercustomerjustlikeWatts(2004)
predicted,andHinzetal.(2011)foundintheirstudies.However,cycletimeisnot
affectedbyemployingnetworkhubs.
These results show thatfoundersand productmanagersofstartups andother
companies should put at least a portion of their time in finding out how this
incentivizing and seeding can be implemented for their respective products and
services; because without using this strategies, they will pass on a huge
opportunity that put companies like Hotmail, Spotify, Instagram (networked
viraliy),Paypal,Dropbox,andGroupon(incentivizedvirality)totheworldstopof
todays Internet businesses. However, it is important they find the right
combinationofusingincentivizingandseedingstrategies,inordertooptimizethe
balancebetweenthenumberofinvitessentpercustomer,theconversionrateof
theseinvites,andthecycletime.

5.2IntroducingtheViralGrowthFormula:

TheviralformulaaspresentedbyLawler(2012)andSkok(2012I;2012II)takes
intoaccounttheinitialsetofcustomers,theviralcoefficientbuiltbymultiplying
thenumberofinvitessentpercustomerwiththeconversionrateoftheseinvites
andthecycletime(see:Figure12and13).However,thissimpleformuladoesnot
take into account several factors, thereto it could be dramatically improved. The
most critical part of getting a web or mobile application to go viral is actually
generatingsomeinterestfortheparticularapp.Therefore,astartupshouldgetto
aproductmarketfitmeaningthestartupnotonlyanalyzesthatthereisagroup

MastersThesis(BusinessStudies)CopyrightsJ.Sakkers(2012)Page56of84


of(potential)customersthatarewillingtouseand/orbuytheproductorservice
offered, but actually finds this group as well (Diamantopoulos et al., 1995). If we
addtheproductmarketfit(F)totheexistingequationfortheviralcoefficientby
Lawler(2012)andSkok(2012I;2012II),wecometoamorerealisticequationas
visualizedinfigure15a.
Also,theviralformulamentionedabovepresumesthecustomerwillonlysend
one batch of invites, ignoring customer lifetime and multiple viral cycles. On the
other hand, some customers will actually stop using the product before or right
aftersendingtheirfirstbatchofinvites:thesocalledlostcustomers.Totakeinto
account the latter, the churn rate, the net viral coefficient is calculated by
subtractingthechurnrate(percentofcustomerthatstopusingtheproductbefore
or right after sending their first batch of invites) from the viral coefficient as
visualizedinfigure15b.

Figure15aImprovedequationfortheviralcoefficient

K = F I C

Figure15bEquationforthenetviralcoefficient

V = K L

Takingintoaccountcustomerlifetimeandmultipleviralcyclespercustomer,we

have added the average number of viral cycles per customer (c) to the viral
formulaofLawler(2012)andSkok(2012I;2012II),whichistheaveragenumber
oftimesthecustomeractuallysendsabatchofinvitestohisnetwork(s).Lastbut
notleast,thecurrentviralformulaassumesaninfinitelyaddressablemarket,while
the web or mobile app will eventually always come to a certain saturation point
(Umax) if it goes (and stays) viral, because the number of customers can never
exceed the actual size of all markets addressed by the product or service.
Therefore, an important condition of the Viral Growth Formula introduced in
figure15cisthatUtcanneverexceedUmax.Ultimately,itshouldbenotedthatthe
improvedformulaforviralgrowthisstillaverysimplereproductionofthehighly

MastersThesis(BusinessStudies)CopyrightsJ.Sakkers(2012)Page57of84


complicated world around us. But startups and other companies looking to get
their app to go viral should definitely know the Viral Growth Formula and try to
monitor,analyze,andimprovetheunderlyingfactorasmuchaspossible.

Figure15cViralGrowthFormula

U(t) = U(0) c

t
( +1)
p

1
V 1

Inordertoprovideapracticaltoolboxforfoundersandproductmanagersofthese

startups and other organizations, a simple viral roadmap for web and mobile
applicationsisconstructedhereunder.Assaid,thefirststepisthattheparticular
productorserviceshouldreachproductmarketfit(Diamantopoulosetal,1995).
The second step is the viral loop or viral cycle, consisting of (2a) optimizing the
conversion rate(s) of (a) landing page(s), (2b) minimizing cycle time, and (2c)
maximizing the number of invites sent per customer. These sub steps are
interchangeableandcantakeeachothersplace.Moreoverthedarkshadedblocks
(2band2c)canbedramaticallyimprovedbyofferingincentivesasprovedinthis
experiment. For example, by offering signup bonuses, referral bonuses, or
purposiveincentivesjustlikePayPal,Dropbox,andGroupondid(i.e.,incentivized
virality).Thethirdstepconsistsofmaximizingtheaveragenumberofviralcycles
per customer. This will create an exponential effect, as new customers will on
averagegetmoreandmoreusedtoinvitingtheirnetwork(s)overandoveragain
as c>1. This is the strategy companies like Hotmail, Spotify, FourSquare, and
Instagramuse(d)(i.e.,networkedvirality).
Ultimately,asourresearchhasshown,itisverybeneficialtostartupstryingto
get their app(s) to go viral to reach network hubs by using seeding strategies.
Startups should really focus on identifying network hubs by analyzing which
small number of (potential) customers is many times better connected than
averagevia(social)networkmetrics,andthereafteronprioritizingthesehubsin
order to exploit the dynamics of scalefree networks: by reaching out to a small

MastersThesis(BusinessStudies)CopyrightsJ.Sakkers(2012)Page58of84


numberofnetworkhubs,thesestartupscouldbeabletoreach(almost)thewhole
networkofpotentialcustomers.

Figure16Viralroadmapforwebandmobileapplications

Step1
Create}itbetween
productandmarket

Step2a
Optimizeconversion
rateoflanding
page(s)

Step2b
Minimizecycletime

Step2c
Maximizenumberof
invitessentper
customer

Step3
Createmultiple
cyclesforeach
customer

Step4
Identify,prioritize,
seedtonetwork
hubs

5.3Limitationsandweaknesses:

We conclude by mentioning some of our studys limitations and associated


directions for further research. As McGrath et al. (1981) state, all research is
flawed. We cannot escape this rule of thumb. First of all, sampling biases in our
datacouldhavestemmedfromVictorMundinotrepresentingallwebapplications
that use incentives to increase their virality, plus the fact that the 313
entrepreneurial participants that actually completed all parts of our experiment
only account for a small sample of all entrepreneurs worldwide. Moreover, the
participants are all from the Netherlands. While there are many similarities
between entrepreneurs worldwide, we cannot generalize for that aspect of our
conclusions.
Also, order effects may exist, because the sets of participants in the different
experimentalsettingsarenotdisjunctive.Wepresumethateachparticipantwould
reactthesameifthepagesandtherefore,thedifferentincentiveswouldbeina
differentorder.Intherealworld,itisplausiblethattheparticipantismorelikely

MastersThesis(BusinessStudies)CopyrightsJ.Sakkers(2012)Page59of84


tosendaninviteviathefirstpageratherthanviathesecondpage,viathesecond
pageratherthanviathethirdpageetcetera.
Moreover,thetitleofthisresearchpaperisWhatMakesYourAppGoViral?,
without stating which kind of web or mobile application is meant. The effects of
incentives, and the selection of the right people to seed a viral campaign to, can
differ for different types of apps. The results and discussion might also be
applicable to companies with freemium, 100% digital, or appbased business
modelsexclusively.Ontheotherhand,theoutcomesofthetestedincentivesand
typesofpeoplecouldhelpcompanieswithnondigitalproductsandservicestogo
viral.Furtherresearchheretoisstronglyencouraged.
Lastbutnotleast,VictorMundiisarelativelyunknownplatformthatonlyexists
forlittleoverayear.Resultsmaybebiasedbecausepotentialcustomersthatvisit
Victor Mundis website are not aware of the benefits of the product yet or
stimulated enough to ultimately become a customer. As mentioned before, the
probabilityofadoptinganewinnovationmaychangedramaticallyuponasecond
or third (or subsequent) exposure where the sudden jump occurs when the
individuals threshold of belief is exceeded (Watts, 2004: p. 261). Ultimately,
both working technology and sharing mechanisms, and product attractiveness in
caseofVictorMundiarepresumedandweretreatedasgivens.

5.4Futureresearch:

Additionalinsightscouldbegainedbyspecifyingmoredetailedtypesofincentives,
e.g.,monetaryandnonmonetary,andbyapplyingtheexperimentconductedand
described in this research paper to other contexts with participants of different
backgroundsfromaroundtheworld.Thesekindsofresearchthatfurtherenhance
our understanding of the effects and possible applications of incentivizing and
seedingstrategiesarestronglyencouraged.
AsproposedintheViralGrowthFormula,inordertogetanapptoviralitneeds
to first reach productmarket fit: the most important variable of the equation.
Unfortunately, very little research is conducted in this area. Therefore, other
researchersarestronglyurgedtodoso.

MastersThesis(BusinessStudies)CopyrightsJ.Sakkers(2012)Page60of84


TheviralequationsproposedbyLawler(2012)andSkok(2012I;2012II)take
into account three major variables: (1) the number of invites sent per customer,
(2) the conversion rate of these invites, and (3) the cycle time. The conducted
experiment was designed to only measure the first and the last variable. Future
experimentsshouldbesetuptomeasuretheconversionrateaswellasthechurn
rate, in order to really be able to measure an actual (net) viral coefficient of >1
(eachnewcustomerattractsatleastanewone).Likewise,everycompanytrying
togetitsapptogoviralisencouragedtoatleastmeasuretheimportantvariables
making up the Viral Growth Formula. In this way, they will be able to find the
perfectrecipethroughexperimentingwithincentivizingandseedingstrategiesby
themselves,justlikePlaxodid(Kalyanametal.,2007).
Finally,theoutcomesoftheexperimentshowthatmaterialincentivesdecrease
cycletimeandseedingtopeoplewithlarge(rthanaverage)networksinducethe
number of invites sent per customer, while providing purposive incentives both
decreases the apps cycle time and increases the number of invites sent per
customer. Future research should focus on discovering which (combination) of
thesestrategiesismosteffective,underwhichcircumstance(s).

MastersThesis(BusinessStudies)CopyrightsJ.Sakkers(2012)Page61of84

AppendixIReferences

Aral,S.,Muchnik,L.&Sundararajan,A.(2011).EngineeringSocialContagions:
OptimalNetworkSeedingandIncentiveStrategies.Workingpaper.Availableat
papers.ssrn.com/sol3/papers.cfm?abstract_id=1770982.

Atazky,R.&Barone,E.(2007).Advertisingandincentivesoverasocialnetwork.
UnitedStatesPatentApplication,publishedonMay31statUS2007/0121843.

Awad,N.F.&Ragowsky,A.(2008).EstablishingTrustinElectronicCommerce
ThroughOnlineWordofMouth:AnExaminationAcrossGenders.Journalof
ManagementInformationSystems,Vol.24(4),101121.

Bampo,M.,Ewing,M.T.,Mather,D.R.,Stewart,D.&Wallace,M.(2008).Theeffects
ofthesocialstructureofdigitalnetworksonviralmarketingperformance.
InformationSystemsResearch,Vol.19(3),273290.

Bansal,H.S.&Voyer,P.A.(2000).WordofMouthProcesseswithinaServices
PurchaseDecisionContext.JournalofServiceResearch,Vol.3(2),166177.

Barabsi,A.L.&Albert,R.(1999).EmergenceofScalinginRandomNetworks.
Science,Vol.286(5439),509512.

Benway,J.P.(1998).BannerBlindness:TheIronyofAttentionGrabbingonthe
WorldWideWeb.ProceedingsoftheHumanFactorsandErgonomicsSocietyAnnual
Meeting,Vol.42(5),463467.

Berger,J.(2011).ArousalIncreasesSocialTransmissionofInformation.
PsychologicalScience,Vol.22(7),891893.

Berger,J.&Schwartz,E.M.(2011).WhatDrivesImmediateandOngoingWordof
Mouth?.JournalofMarketingResearch,Vol.48(5),869880.

Berger,J.&Milkman,K.L.(2012).WhatMakesOnlineContentViral?.Journalof
MarketingResearch,Vol.49(2),192205.

Bianconi,G.&Barabsi,A.L.(2001).Competitionandmultiscalinginevolving
networks.EurophysicsLetters,Vol.54(4),436442.

Biyalogorsky,E.,Gerstner,E.&Libai,B.(2001).CustomerReferralManagement:
OptimalRewardPrograms.MarketingScience,Vol.20(1),8295.

Boots,M.&Sasaki,A.(1999).Smallworldsandtheevolutionofvirulence:
Infectionoccurslocallyandatadistance.ProceedingsoftheRoyalSocietyLondon
SeriesB,Vol.266(1432),19331938.

MastersThesis(BusinessStudies)CopyrightsJ.Sakkers(2012)Page62of84


Borgatti,S.P.&Jones,C.(1998).NetworkMeasuresofSocialCapital.Connections,
Vol.21(2),2736.

Bronner,F.&DeHoog,R.(2010).VacationersandeWOM:WhoPosts,andWhy,
Where,andWhat?.JournalofTravelResearch,Vol.50(1),1526.

Brown,J.,Broderick,A.J.&Lee,N.(2007).Wordofmouthcommunicationwithin
onlinecommunities:Conceptualizingtheonlinesocialnetwork.Journalof
InteractiveMarketing,Vol.21(3),220.

Buttle,F.A.(1998).Wordofmouth:understandingandmanagingreferral
marketing.JournalofStrategicMarketing,Vol.6,241254.

Clark,P.B.&Wilson,J.Q.(1961).IncentiveSystems:ATheoryofOrganizations.
AdministrativeScienceQuarterly,Vol.6(2),129166.

Clark,M.S.,Mills,J.&Powel,M.C.(1986).Keepingtrackofneedsincommunaland
exchangerelationships.JournalofPersonalityandSocialPsychology,Vol.51(2),
333338.

Compete(2012).SiteanalyticsforPlaxo.comfromApril2011toApril2012.
VisitedonMay27thatsiteanalytics.compete.com/plaxo.com/.

Cruz,D.&Fill,C.(2008).Evaluatingviralmarketing:isolatingthekeycriteria.
MarketingIntelligence&Planning,Vol.26(7),743758.

Dellarocas,C.&Narayan,R.(2006).AStatisticalMeasureofaPopulations
PropensitytoEngageinPostPurchaseOnlineWordofMouth.StatisticalScience,
Vol.22(2),277285.

DeBruyn,A.&Lilien,G.L.(2008).Amultistagemodelofwordofmouthinfluence
throughviralmarketing.InternationalJournalofResearchinMarketing,Vol.25,
151163.

Diamantopoulos,A.,Schlegelmilch,B.B.&DuPreez,J.P.(1995).Lessonsforpan
Europeanmarketing?Theroleofconsumerpreferencesinfinetuningtheproduct
marketfit.InternationalMarketingReview,Vol.12(2),3852.

Dobele,A.,Toleman,D.&Beverland,M.(2005).Controlledinfection!Spreadingthe
brandmessagethroughviralmarketing.BusinessHorizons,Vol.48(2),143149.

Dobele,A.,Lindgreen,A.,Beverland,M.,Vanhamme,J.&VanWijk,R.(2007).Why
passonviralmessages?Becausetheyconnectemotionally.BusinessHorizons,Vol.
50(4),291304.

Donckels,R.&Lambrecht,J.(1997).TheNetworkPositionofSmallBusinesses:An
ExplanatoryModel.JournalofSmallBusinessManagement,Vol.35(2),1326.

MastersThesis(BusinessStudies)CopyrightsJ.Sakkers(2012)Page63of84

Dunbar,R.I.M.(1993)Coevolutionofneocortexsize,groupsizeandlanguagein
humans.BehavioralandBrainSciences,Vol.16,681735.

Engel,J.F.,Kegerreis,R.J.&Blackwell,R.D.(1969).WordofmouthCommunication
bytheInnovator.JournalofMarketing,Vol.33(3),1519.

Fitzsimmons,J.A.,Noh,J.&Thies,E.(1998).Purchasingbusinessservices.Journal
ofBusiness&IndustrialMarketing,Vol.13(3/4),370380.

Friestad,M.&Wright,P.(1994).ThePersuasionKnowledgeModel:HowPeople
CopewithPersuasionAttempts.JournalofConsumerResearch,Vol.21(1),131.

Garbarino,E.&Strahilevitz,M.(2004).Genderdifferencesintheperceivedriskof
buyingonlineandtheeffectsofreceivingasiterecommendation.Journalof
BusinessResearch,Vol.57(7),768775.

Gladwell,M.(2000)TheTippingPoint.BackBayBooks/Little,Brownand
Company,NewYork(NY),USA.

Godes,D.,Mayzlin,D.,Chen,Y.,Das,S.,Dellarocas,C.,Pfeiffer,B.,Libai,B.,Sen,S.,
Shi,M.&Verlegh,P.(2005).TheFirmsManagementofSocialInteractions.
MarketingLetters,Vol.16(34),415428.

Godes,D.&Mayzlin,D.(2009).FirmCreatedWordofMouthCommunication:
EvidencefromaFieldTest.JournalofMarketingScience,Vol.28(4),721739.

Golan,G.J.&Zaidner,L.(2008).CreativeStrategiesinViralAdvertising:An
ApplicationofTaylorsSixSegmentMessageStrategyWheel.JournalofComputer
MediatedCommunication,Vol.13(4),959972.

Goldenberg,J.,Libai,B.&Muller,E.(2001).TalkoftheNetwork:AComplex
SystemsLookattheUnderlyingProcessofWordofMouth.MarketingLetters,Vol.
12(3),211223.

Granovetter,M.S.(1973).TheStrengthofWeakTies.AmericanJournalofSociology,
Vol.78(6),13601380.

Gruen,T.W.,Osmonbekov,T.&Czaplewski,A.(2006).eWOM:Theimpactof
customertocustomeronlineknowhowexchangeoncustomervalueandloyalty.
JournalofBusinessResearch,Vol.59(4),449456.

Helm,S.(2000).ViralMarketingEstablishingCustomerRelationshipsbyWord
ofmouse.ElectronicMarkets,Vol.10(3),158161.

HennigThurau,T.,Gwinner,K.P.,Walsh,G.&Gremier,D.D.(2004).Electronic
wordofmouthviaconsumeropinionplatforms:Whatmotivatesconsumersto

MastersThesis(BusinessStudies)CopyrightsJ.Sakkers(2012)Page64of84


articulatethemselvesontheInternet?.JournalofInteractiveMarketing,Vol.18(1),
3852.

Hinz,O.,Skiera,B.,Barrot,C.&Becker,J.U.(2011).SeedingStrategiesforViral
Marketing:AnEmpiricalComparison.JournalofMarketing,Vol.75(6),5571.

Ho,J.Y.C.&Dempsey,M.(2010).Viralmarketing:Motivationstoforwardonline
content.JournalofBusinessResearch,Vol.63(910),10001006.

Huberman,B.A.,Romero,D.M.&Wu,F.(2008).Socialnetworksthatmatter:
Twitterunderthemicroscope.HewlettPackardLaboratories,workingpaper.
Availableatpapers.ssrn.com/sol3/papers.cfm?abstract_id=1313405.

IgniteSocialMediaBlog(2012).47OutrageousViralMarketingExamplesoverthe
LastDecade.VisitedonJanuary25thatwww.ignitesocialmedia.com/socialmedia
examples/viralmarketingexamples/.

Jaffe,J.(2010).FliptheFunnel:HowtoUseExistingCustomerstoGainNewOnes.
JohnWiley&Sons,Inc.,Hoboken(NJ),USA.

Jensen,J.I.&Greve,A.(2002).DoestheDegreeofRedundancyinSocialNetworks
InfluencetheSuccessofBusinessStartups?.InternationalJournalof
EntrepreneurialBehaviour&Research,Vol.8(5),254267.

Kaikati,A.&Kaikati,J.(2004).Stealthmarketing:Howtoreachconsumers
surreptitiously.CaliforniaManagementReview,Vol.46(2),622.

Kalyanam,K.,McIntyre,S.&Masonis,T.(2007).Adaptiveexperimentationin
interactivemarketing:ThecaseofviralmarketingatPlaxo.JournalofInteractive
Marketing,Vol.21(3),7285.

Kaplan,A.M&Haenlein,M.(2011).Twoheartsinthreequartertime:Howtowaltz
thesocialmedia/viralmarketingdance.BusinessHorizons,Vol.54(3),253263.

Kiel,G.C.&Layton,R.A.(1981).DimensionsofConsumerInformationSeeking
Behavior.JournalofMarketingResearch,Vol.18(2),233239.

Kleinberg,J.(2000).Thesmallworldphenomenon:analgorithmperspective.
STOC00ProceedingsofthethirtysecondannualACMsymposiumonTheoryof
computing,163170.

Kozinets,R.V.,DeValck,K.,Wojnicki,A.C.&Wilner,S.J.S.(2010).Networked
Narratives:UnderstandingWordofMouthMarketinginOnlineCommunities.
JournalofMarketing,Vol.74,7189.

Lawler,K.(2012).AViralityFormula.VisitedonMay21stat
kevinlawler.com/viral.

MastersThesis(BusinessStudies)CopyrightsJ.Sakkers(2012)Page65of84

Loscocco,K.,Monnat,S.M.,Moore,G.&Lauber,K.B.(2009).EnterprisingWomen
AComparisonofWomensandMensSmallBusinessNetworks.Gender&Society,
Vol.23(3),388411.

Lubarsky,D.A.(2005).IncentivizeEverything,IncentivizeNothing.AnesthAnalg,
Vol.100,490492.

Mazzarol,T.,Sweeney,J.C.&Soutar,G.N.(2007).Conceptualizingwordofmouth
activity,triggersandconditions:anexploratorystudy.EuropeanJournalof
Marketing,Vol.41(11/12),14751494.

McGrath,J.E.,Martin,J.&Kulka,R.A.(1981).SomeQuasiRulesforMaking
JudgmentCallsinResearch.AmericanBehavioralScientist,Vol.25(2),211224.

MicrosoftCorporation(2011).Detectingviralitypathsandsupportingreferral
monetization.UnitedStatesPatentApplication,publishedonSeptember29thatUS
2011/0239103.

Mittal,V.&Kamakura,W.A.(2001).Satisfaction,RepurchaseIntent,and
RepurchaseBehavior:InvestigatingtheModeratingEffectofCustomer
Characteristics.JournalofMarketingResearch,Vol.38(1),131142.

Norman,A.T.&Russell,C.A.(2006).ThePassAlongEffect:InvestigatingWordof
MouthEffectsonOnlineSurveyProcedures.JournalofComputerMediated
Communication,Vol.11(4),10851103.

Petty,R.E.&Cacioppo,J.T.(1986).TheElaborationLikelihoodModelofPersuasion
inAdvancesinExperimentalSocialPsychology,Vol.19,123192.London:Academic
PressInc.

Porter,L.&Golan,G.J.(2006).Fromsubservientchickenstobrawnymen:A
comparisonofviraladvertisingtotelevisionadvertising.JournalofInteractive
Advertising,Vol.6(2),3038.

PQMedia(2012).ExclusivePQMediaResearch:DespiteWorstRecessionin
Decades,BrandsIncreasedSpendingonWordofMouthMarketing14.2%to$1.54
Billionin2008.VisitedonMay26thatwww.pqmedia.com/aboutpress
20090729wommf.html.

Raz,O.&Gloor,P.A.(2007).SizeReallyMattersNewInsightsforStartups
Survival.ManagementScience,Vol.53(2),169177.

Reichheld,F.F.(2003).TheOneNumberYouNeedtoGrow.HarvardBusiness
Review,Vol.81(12),4654.

MastersThesis(BusinessStudies)CopyrightsJ.Sakkers(2012)Page66of84


Riegner,C.(2007).WordofMouthontheWeb:TheImpactofWeb2.0on
ConsumerPurchaseDecisions.JournalofAdvertisingResearch,Vol.47(4),436447.

Ryu,G.&Feick,L.(2007).APennyforYourThoughts:ReferralRewardPrograms
andReferralLikelihood.JournalofMarketing,Vol.71(1),8494.

Ryu,G.&Han,J.K.(2009).Wordofmouthtransmissioninsettingswithmultiple
opinions:TheimpactofotheropinionsonWOMlikelihoodandvalence.Journalof
ConsumerPsychology,Vol.19(3),403415.

Schmitt,P.,Skiera,B.&VandenBulte,C.(2011).ReferralProgramsandCustomer
Value.JournalofMarketing,Vol.75(1),4659.

Schutz,W.C.(1992).BeyondfiroBThreenewtheoryderivedmeasures
elementB:behavior,elementF:feelings,elementS:self.PsychologicalReports,Vol.
70,915937.

Skok(2012I).LessonsLearnedViralMarketing(blogpost).VisitedonMay21nd
atwww.forentrepreneurs.com/lessonslearntviralmarketing/.

Skok(2012II).TheSciencebehindViralMarketing(presentation).VisitedonMay
21ndatwww.forentrepreneurs.com/lessonslearntviralmarketing/.

Slashdot(2012).AOLDumps$1.2BillionWorthofAcquisitions.VisitedonJune
24thathttp://news.slashdot.org/story/10/06/17/0038248/aoldumps12
billionworthofacquisitions.

StartupLessonsLearned(2012).DropboxStartupLessonsLearnedbyDrew
Houston(presentation).VisitedonMay25that
www.slideshare.net/gueste94e4c/dropboxstartuplessonslearned3836587.

Steffes,E.M.&Burgee,L.E.(2009).Socialtiesandonlinewordofmouth.Internet
Research,Vol.19(1),4259.

Strapparava,C.,Guerini,M.&zbal,G.(2011).PersuasiveLanguageandViralityin
SocialNetworks.LectureNotesinComputerScience,Vol.6974/2011,357366.

Sun,T.,Youn,S.,Wu,G.&Kuntaraporn,M.(2006).OnlineWordofMouth(or
Mouse):AnExplorationofItsAntecedentsandConsequences.JournalofComputer
MediatedCommunication,Vol.11(4),11041127.

Sundaram,D.S.,Mitra,K.&Webster,C.(1998).WordofMouthCommunications:A
MotivationalAnalysis.AdvancesinConsumerResearch,Vol.25,527531.

Sussman,S.W.&Siegal,W.S.(2003).InformationalInfluenceinOrganizations:An
IntegratedApproachtoKnowledgeAdoption.InformationSystemsResearch,Vol.14
(1),4765.

MastersThesis(BusinessStudies)CopyrightsJ.Sakkers(2012)Page67of84

Sweeney,J.C.,Soutar,G.N.&Mazzarol,T.(2007).Factorsinfluencingwordof
moutheffectiveness:receiverperspectives.EuropeanJournalofMarketing,Vol.42
(3/4),344364.

TechCrunch(2012).EightWaysToGoViral.VisitedonJanuary25that
techcrunch.com/2011/12/26/eightwaysgoviral/.

TheNextWeb(2012).MenarefromFoursquareandwomanarefromFacebook,
apparently.VisitedonMay22ndat
thenextweb.com/socialmedia/2012/02/15/menarefromfoursquareand
womenarefromfacebookapparently/.

Trusov,M.,Bucklin,R.E.&Pauwels,K.(2009).EffectsofWordofMouthVersus
TraditionalMarketing:FindingsfromanInternetSocialNetworkingSite.Journalof
Marketing,Vol.73(5),90102.

Tuk,M.A.,Verlegh,P.W.J.,Smidts,A.&Wigboldus,D.H.J.(2009).Salesandsincerity:
Theroleofrelationalframinginwordofmouthmarketing.JournalofConsumer
Psychology,Vol.19(1),3847.

Valente,T.W.&Davis,R.L.(1999).AcceleratingtheDiffusionofInnovationsUsing
OpinionLeaders.TheAnnalsoftheAmericanAcademyofPoliticalandSocial
Science,Vol.566(1),5567.

Walster,E.,Berscheid,E.&Walster,G.W.(1973).Newdirectionsinequity
research.JournalofPersonalityandSocialPsychology,Vol.25(2),151176.

Watts,D.J.&Strogatz,S.H.(1998).Collectivedynamicsofsmallworldnetworks.
Nature,393,440442.

Watts,D.J.(1999).Networks,Dynamics,andtheSmallWorldPhenomenon.
AmericanJournalofSociology,Vol.105(2),493527.

Watts,D.J.(2004).TheNewScienceofNetworks.AnnualReviewofSociology,Vol.
30,243270.

Weng,L.,Flammini,A.,Vespignani,A.&Menczer,F.(2012).Competitionamong
memesinaworldwithlimitedattention.ScientificReports.DownloadedonMay
23rdfrom
www.nature.com/srep/2012/120329/srep00335/full/srep00335.html.

Whiting,V.R.&DeJanasz,S.C.(2004).Mentoringinthe21stCentury:Usingthe
InternettoBuildSkillsandNetworks.JournalofManagementEducation,Vol.28
(3),275293.

MastersThesis(BusinessStudies)CopyrightsJ.Sakkers(2012)Page68of84


Wikipedia(2012I).Dropbox.VisitedonMay29that
en.wikipedia.org/wiki/Dropbox_(service).

Wikipedia(2012II).Population.VisitedonMay30that
en.wikipedia.org/wiki/Population.

Wikipedia(2012III).Smallworldnetwork.VisitedonJune1stat
en.wikipedia.org/wiki/Smallworld_network.

Wilson,R.F.(2005).TheSixSimplePrinciplesofViralMarketing.WebMarketing
Today,February1.

Wirtz,J.&Chew,P.(2002).Theeffectsofincentives,dealproneness,satisfaction
andtiestrengthonwordofmouthbehaviour.InternationalJournalofService
IndustryManagement,Vol.13(2),141162.

ZDNet(2012).Whymyaddressbookisspammingyou.VisitedonMay30thvia
WaybackMachineat
web.archive.org/web/20070105030148/http://review.zdnet.com/AnchorDesk/4
5207297_165111563.html.

MastersThesis(BusinessStudies)CopyrightsJ.Sakkers(2012)Page69of84

AppendixIIDesignofexperiment

In order to get the data and as much respondents as possible, we set up an


experiment within an existing product: Victor Mundi the digital workspace for
entrepreneurs.Duringtheperiodtheexperimentwasactive,eachnewcustomerof
Victor Mundi had to go through three wizards each with three or four steps: a
signupwizard,animportwizardandacreatewizard.

SequenceofwizardsstartingwithsignupforVictorMundibyeachparticipant:

Screenshotsincludingdatabasetags:

Signupwizard(noincentive)

MastersThesis(BusinessStudies)CopyrightsJ.Sakkers(2012)Page70of84

Importwizard(materialincentive)

Createwizard(purposiveincentive)

MastersThesis(BusinessStudies)CopyrightsJ.Sakkers(2012)Page71of84


The database tags are added to several pages to be able to store the information
thatisneededtoanalyzetheexperiment.Thevariablesusedare:
pageName:stringwiththenameofthepage
step:usedtostorethenumberofthepanelthatisvisibletotheuser
clientIP:databasetableusedtostorethedata
pagetrackingClicktbl:databasetableusedtostorethedata

Page_trackingpagetrackingtbl=newPage_tracking();
Page_tracking_clickedpagetrackingClicktbl=newPage_tracking_clicked();
stringpageName;
stringstep;
stringclientIp;

OnPage_LoadVictorMundirunsthefunctionsgetClientIP()togettheIPofthe
customer and the name of the page the customer is visiting. We need those first,
becausegetStepRegisterVisit()needsthosevaluestoregisterthenewentry:

protectedvoidPage_Load(objectsender,EventArgse)
{
getClientIP();
pageName=Path.GetFileName(Request.Path);
getStepRegisterVisit();
}

Another function checks which panel is visible for the customer and whether he
already visited it. If not, Victor Mundi registers a new entry with the
insertVisitEntrymethod.VictorMundistorestheuserID,theclientIP,pageName,
and a customer name for the step within the wizard(s) the customer is visiting.
Also,atimestampisautomaticallyaddedbythedatabase.TheinsertVisitEntry
method returns the number of the new entry within the database, and Victor
Mundiaddsthatnumberinthesession.Itisstoredinordertobeabletorecallthat
numberwhenregisteringtheclicks.
InthesignupwizardwhereauserIDisnotavailableyetavariationofthe
samemethodisused,whereasonlytheclientIPisstored.Theotherwizard,where
VictorMundiknowstheclientIPandtheuserID,atriggerstoresthelatestIPofthe
customertothedatabase.

MastersThesis(BusinessStudies)CopyrightsJ.Sakkers(2012)Page72of84


protectedstringgetStepRegisterVisit()
{

step="0";

if(panelStep1.Visible){
step="1";
if(Session["stepWiz1"]==null)
Session["stepWiz1"]=
pagetrackingtbl.insertVisitEntry(Globals.UserID,clientIp,pageName,"1_wizard");
}

if(panelStep2.Visible){
step="2";
if(Session["stepWiz2"]==null)
Session["stepWiz2"]=
pagetrackingtbl.insertVisitEntry(Globals.UserID,clientIp,pageName,"2_wizard");
}

if(panelStep3.Visible)
{
step="3";
if(Session["stepWiz3"]==null)
Session["stepWiz3"]=
pagetrackingtbl.insertVisitEntry(Globals.UserID,clientIp,pageName,"3_wizard");
}

returnstep;
}

Ifcustomersclickona(shareorinvite)button,otherfunctionsarecalled.Inthose
Victor Mundi adds the line insertClickEntry(sender, e);, which runs the above
function. This one stores the number of the visit and the ID of the button in the
table Page_tracking_clicked. Victor Mundi also stores the date stamp; that is
automaticallyaddedtothedatabase.

protectedvoidinsertClickEntry(objectsender,EventArgse){
if(sender!=null){
WebControlbtnClicked=(WebControl)sender;
stringbtnId=btnClicked.ID;
if(!String.IsNullOrEmpty(btnId)&&step!="0"){

pagetrackingClicktbl.insertClickEntry(Convert.ToInt32(Session["stepWiz"+
step].ToString()),btnId);
}
}
}

ThisisthefunctionthatgetstheclientIP:

MastersThesis(BusinessStudies)CopyrightsJ.Sakkers(2012)Page73of84


protectedvoidgetClientIP()
{
clientIp=Request.ServerVariables["HTTP_X_FORWARDED_FOR"];
if(!string.IsNullOrEmpty(clientIp))
{
string[]ipRange=clientIp.Split(',');
intle=ipRange.Length1;
stringtrueIP=ipRange[le];
}
else
{
clientIp=Request.ServerVariables["REMOTE_ADDR"];
}
}

In order to finally get the correct and complete data from the database of Victor
Mundithisextensiveselectqueryisperformed:

useVICTOR

go

IFOBJECT_ID('TEMPDB..#USER')ISNOTNULLDROPTABLE#USER

IFOBJECT_ID('TEMPDB..#IP')ISNOTNULLDROPTABLE#IP

selectpt.ip_address,

(selectcount(step_number)frompage_trackingpt1wherept1.ip_address=pt.ip_address

andpt1.page_name='signup.aspx'

and pt1.step_number = '1_signup_fnv') as [visited


1_signupFNV],

(selectmin(pt1.created_date)frompage_trackingpt1wherept1.ip_address=pt.ip_address

andpt1.page_name='signup.aspx'

and
pt1.step_number
=
'1_signup_fnv')
as
date_1_signupFNV,

(selectcount(step_number)frompage_trackingpt1wherept1.ip_address=pt.ip_address

andpt1.page_name='signup.aspx'

and pt1.step_number = '2_signup_fnv') as [visited


2_signupFNV],

(selectcount(step_number)frompage_trackingpt1wherept1.ip_address=pt.ip_address

andpt1.page_name='signup.aspx'

and pt1.step_number = '3_signup_fnv') as [visited


3_signupFNV],

(selectcount(step_number)frompage_trackingpt1wherept1.ip_address=pt.ip_address

andpt1.page_name='signup.aspx'

and pt1.step_number = '4_signup_fnv') as [visited


4_signupFNV],

MastersThesis(BusinessStudies)CopyrightsJ.Sakkers(2012)Page74of84

LinkedIn

(selectcount(*)frompage_tracking_clickedptc

join page_tracking pt1 on pt1.pk_id


ptc.FK_tracking_id

andptc.button_id='LinkedinShareFNV'

andpt1.step_number='4_signup_fnv'

wherept1.ip_address=pt.ip_address)asclicked_linkedInFNV,

(selectmin(ptc.created_date)frompage_tracking_clickedptc

join page_tracking pt1 on pt1.pk_id


ptc.FK_tracking_id

andptc.button_id='LinkedinShareFNV'

andpt1.step_number='4_signup_fnv'

wherept1.ip_address=pt.ip_address)asdate_linkedInFNV,

Tweet

(selectcount(*)frompage_tracking_clickedptc

join page_tracking pt1 on pt1.pk_id


ptc.FK_tracking_id

andptc.button_id='TweetShareFNV'

andpt1.step_number='4_signup_fnv'

wherept1.ip_address=pt.ip_address)asclicked_TweetFNV,

(selectmin(ptc.created_date)frompage_tracking_clickedptc

join page_tracking pt1 on pt1.pk_id


ptc.FK_tracking_id

andptc.button_id='TweetShareFNV'

andpt1.step_number='4_signup_fnv'

wherept1.ip_address=pt.ip_address)asdate_TweetFNV,

Facebook

(selectcount(*)frompage_tracking_clickedptc

join page_tracking pt1 on pt1.pk_id


ptc.FK_tracking_id

andptc.button_id='FacebookShareFNV'

andpt1.step_number='4_signup_fnv'

wherept1.ip_address=pt.ip_address)asclicked_FacebookFNV,

(selectmin(ptc.created_date)frompage_tracking_clickedptc

join page_tracking pt1 on pt1.pk_id


ptc.FK_tracking_id

andptc.button_id='FacebookShareFNV'

andpt1.step_number='4_signup_fnv'

wherept1.ip_address=pt.ip_address)asdate_FacebookFNV,

Google

(selectcount(*)frompage_tracking_clickedptc

join page_tracking pt1 on pt1.pk_id


ptc.FK_tracking_id

andptc.button_id='GoogleShareFNV'

MastersThesis(BusinessStudies)CopyrightsJ.Sakkers(2012)Page75of84

andpt1.step_number='4_signup_fnv'

wherept1.ip_address=pt.ip_address)asclicked_GoogleFNV,

(selectmin(ptc.created_date)frompage_tracking_clickedptc

join page_tracking pt1 on pt1.pk_id =


ptc.FK_tracking_id

andptc.button_id='GoogleShareFNV'

andpt1.step_number='4_signup_fnv'

wherept1.ip_address=pt.ip_address)asdate_GoogleFNV,

geenFNV

(selectcount(step_number)frompage_trackingpt1wherept1.ip_address=pt.ip_address

andpt1.page_name='signup.aspx'

and pt1.step_number = '1_signup') as [visited


1_signup],

(selectmin(pt1.created_date)frompage_trackingpt1wherept1.ip_address=pt.ip_address

andpt1.page_name='signup.aspx'

andpt1.step_number='1_signup')asdate_1_signup,

(selectcount(step_number)frompage_trackingpt1wherept1.ip_address=pt.ip_address

andpt1.page_name='signup.aspx'

and pt1.step_number = '2_signup') as [visited


2_signup],

(selectcount(step_number)frompage_trackingpt1wherept1.ip_address=pt.ip_address

andpt1.page_name='signup.aspx'

and pt1.step_number = '3_signup') as [visited


3_signup],

LinkedIn

(selectcount(*)frompage_tracking_clickedptc

join page_tracking pt1 on pt1.pk_id =


ptc.FK_tracking_id

andptc.button_id='LinkedinShare'

andpt1.step_number='3_signup'

wherept1.ip_address=pt.ip_address)asclicked_linkedIn,

(selectmin(ptc.created_date)frompage_tracking_clickedptc

join page_tracking pt1 on pt1.pk_id =


ptc.FK_tracking_id

andptc.button_id='LinkedinShare'

andpt1.step_number='3_signup'

wherept1.ip_address=pt.ip_address)asdate_linkedIn,

Tweet

(selectcount(*)frompage_tracking_clickedptc

join page_tracking pt1 on pt1.pk_id =


ptc.FK_tracking_id

andptc.button_id='TweetShare'

MastersThesis(BusinessStudies)CopyrightsJ.Sakkers(2012)Page76of84

andpt1.step_number='3_signup'

wherept1.ip_address=pt.ip_address)asclicked_Tweet,

(selectmin(ptc.created_date)frompage_tracking_clickedptc

join page_tracking pt1 on pt1.pk_id =


ptc.FK_tracking_id

andptc.button_id='TweetShare'

andpt1.step_number='3_signup'

wherept1.ip_address=pt.ip_address)asdate_Tweet,

Facebook

(selectcount(*)frompage_tracking_clickedptc

join page_tracking pt1 on pt1.pk_id =


ptc.FK_tracking_id

andptc.button_id='FacebookShare'

andpt1.step_number='3_signup'

wherept1.ip_address=pt.ip_address)asclicked_Facebook,

(selectmin(ptc.created_date)frompage_tracking_clickedptc

join page_tracking pt1 on pt1.pk_id =


ptc.FK_tracking_id

andptc.button_id='FacebookShare'

andpt1.step_number='3_signup'

wherept1.ip_address=pt.ip_address)asdate_Facebook,

Google

(selectcount(*)frompage_tracking_clickedptc

join page_tracking pt1 on pt1.pk_id =


ptc.FK_tracking_id

andptc.button_id='GoogleShare'

andpt1.step_number='3_signup'

wherept1.ip_address=pt.ip_address)asclicked_Google,

(selectmin(ptc.created_date)frompage_tracking_clickedptc

join page_tracking pt1 on pt1.pk_id =


ptc.FK_tracking_id

andptc.button_id='GoogleShare'

andpt1.step_number='3_signup'

wherept1.ip_address=pt.ip_address)asdate_Google

into#IP

frompage_trackingpt

wherept.fk_userisnull

groupby

pt.ip_address

SELECT[user_id],

MastersThesis(BusinessStudies)CopyrightsJ.Sakkers(2012)Page77of84

(selectmax(ip_address)frompage_trackingwherepage_tracking.FK_user=[user_id])asipaddress,

user_created_onasregistration_date,

isnull(companyname,'')ascompanyname,

isnull(firstname,'')asfirstname,

isnull(initials,'')asinitials,

isnull(middlename,'')asmiddlename,

isnull(lastname,'')aslastname,

isnull(gender,'')asgender,

(selectcount(step_number)frompage_trackingptwherept.fk_user=[user_id]

andpt.step_number='1_wizard')as[visited1_wizard],

isnull((selectsum(qty)fromLOG_user_imported_countimported

whereimported.fk_user=[user_id]),0)ascontacts_imported,

(selectcount(step_number)frompage_trackingptwherept.fk_user=[user_id]

andpt.step_number='2_wizard')as[visited2_wizard],

(selectcount(*)frompage_tracking_clickedptc

join page_tracking pt on pt.pk_id =


ptc.FK_tracking_id

andptc.button_id='btnUpdateStatus'

andpt.step_number='2_wizard'

wherept.fk_user=[user_id])asclicked_wizard_UpdateStatus,

(selectmin(ptc.created_date)frompage_tracking_clickedptc

join page_tracking pt on pt.pk_id


=ptc.FK_tracking_id

andptc.button_id='btnUpdateStatus'

andpt.step_number='2_wizard'

wherept.fk_user=[user_id])asdate_wizUpdateStatus,

(selectcount(step_number)frompage_trackingptwherept.fk_user=[user_id]

andpt.step_number='3_wizard')as[visited3_wizard],

(selectcount(*)frompage_tracking_clickedptc

join page_tracking pt on pt.pk_id =


ptc.FK_tracking_id

andptc.button_id='btnMailToAdress'

andpt.step_number='3_wizard'

where pt.fk_user = [user_id]) as clicked_wizMailToAdress,

PromoInvite

(selectcount(*)frompage_tracking_clickedptc

join page_tracking pt on pt.pk_id =


ptc.FK_tracking_id

MastersThesis(BusinessStudies)CopyrightsJ.Sakkers(2012)Page78of84

andptc.button_id='btnPromoInvite'

andpt.step_number='3_wizard'

wherept.fk_user=[user_id])asclicked_PromoInvite,

(selectmin(ptc.created_date)frompage_tracking_clickedptc

join page_tracking pt on pt.pk_id


=ptc.FK_tracking_id

andptc.button_id='btnPromoInvite'

andpt.step_number='3_wizard'

wherept.fk_user=[user_id])asdate_PromoInvite,

Marketing

(selectcount(step_number)frompage_trackingptwherept.fk_user=[user_id]

and pt.step_number = '1_marketplace_add') as [visited


1_marketplace],

(selectcount(step_number)frompage_trackingptwherept.fk_user=[user_id]

and pt.step_number = '2_assignment_add') as [visited


2_assignment],

(selectcount(step_number)frompage_trackingptwherept.fk_user=[user_id]

and pt.step_number = '2_product_add') as [visited


2_product],

(selectcount(step_number)frompage_trackingptwherept.fk_user=[user_id]

and pt.step_number = '2_service_add') as [visited


2_service],

(selectcount(step_number)frompage_trackingptwherept.fk_user=[user_id]

and pt.step_number = '3_assign_matching' ) as [visited


3_assign_matching],

Productstatusupdate

(selectcount(step_number)frompage_trackingptwherept.fk_user=[user_id]

and pt.step_number = '3_product_statusupdate') as


[visited3_product_statusupdate],

(selectcount(*)frompage_tracking_clickedptc

join page_tracking pt on pt.pk_id =


ptc.FK_tracking_id

andptc.button_id='btnUpdateStatus'

andpt.step_number='3_product_statusupdate'

wherept.fk_user=[user_id])

as
clicked_product_UpdateStatus,

(selectmin(ptc.created_date)frompage_tracking_clickedptc

join page_tracking pt on pt.pk_id =


ptc.FK_tracking_id

andptc.button_id='btnUpdateStatus'

andpt.step_number='3_product_statusupdate'

wherept.fk_user=[user_id])

as
date_productUpdateStatus,

Servicestatusupdate

(selectcount(step_number)frompage_trackingptwherept.fk_user=[user_id]

MastersThesis(BusinessStudies)CopyrightsJ.Sakkers(2012)Page79of84

and pt.step_number = '3_service_statusupdate') as


[visited3_service_statusupdate],

(selectcount(*)frompage_tracking_clickedptc

join page_tracking pt on pt.pk_id =


ptc.FK_tracking_id

andptc.button_id='btnUpdateStatus'

andpt.step_number='3_service_statusupdate'

wherept.fk_user=[user_id])

as
clicked_service_UpdateStatus,

(selectmin(ptc.created_date)frompage_tracking_clickedptc

join page_tracking pt on pt.pk_id =


ptc.FK_tracking_id

andptc.button_id='btnUpdateStatus'

andpt.step_number='3_service_statusupdate'

wherept.fk_user=[user_id])

as
date_serviceUpdateStatus,

Assignstatusupdate

(selectcount(step_number)frompage_trackingptwherept.fk_user=[user_id]

and pt.step_number = '4_assign_statusupdate') as


[visited4_assign_statusupdate],

(selectcount(*)frompage_tracking_clickedptc

join page_tracking pt on pt.pk_id =


ptc.FK_tracking_id

andptc.button_id='btnUpdateStatus'

andpt.step_number='4_assign_statusupdate'

wherept.fk_user=[user_id])

as
clicked_assign_UpdateStatus,

(selectmin(ptc.created_date)frompage_tracking_clickedptc

join page_tracking pt on pt.pk_id =


ptc.FK_tracking_id

andptc.button_id='btnUpdateStatus'

andpt.step_number='4_assign_statusupdate'

wherept.fk_user=[user_id])

as
date_assignUpdateStatus,

Productpromote

(selectcount(step_number)frompage_trackingptwherept.fk_user=[user_id]

and pt.step_number = '4_product_promote') as [visited


4_product_promote],

(selectcount(*)frompage_tracking_clickedptc

join page_tracking pt on pt.pk_id =


ptc.FK_tracking_id

MastersThesis(BusinessStudies)CopyrightsJ.Sakkers(2012)Page80of84

andptc.button_id='btnMailToAdress'

andpt.step_number='4_product_promote'

wherept.fk_user=[user_id])

as
clicked_product_MailToAdress,

(selectcount(*)frompage_tracking_clickedptc

join page_tracking pt on pt.pk_id =


ptc.FK_tracking_id

andptc.button_id='btnPromoInvite'

andpt.step_number='4_product_promote'

wherept.fk_user=[user_id])

as
clicked_product_promo,

(selectmin(ptc.created_date)frompage_tracking_clickedptc

join page_tracking pt on pt.pk_id =


ptc.FK_tracking_id

andptc.button_id='btnPromoInvite'

andpt.step_number='4_product_promote'

wherept.fk_user=[user_id])

as
date_product_promo,

Servicepromote

(selectcount(step_number)frompage_trackingptwherept.fk_user=[user_id]

and pt.step_number = '4_service_promote') as [visited


4_service_promote],

(selectcount(*)frompage_tracking_clickedptc

join page_tracking pt on pt.pk_id =


ptc.FK_tracking_id

andptc.button_id='btnMailToAdress'

andpt.step_number='4_service_promote'

wherept.fk_user=[user_id])

as
clicked_service_MailToAdress,

(selectcount(*)frompage_tracking_clickedptc

join page_tracking pt on pt.pk_id =


ptc.FK_tracking_id

andptc.button_id='btnPromoInvite'

andpt.step_number='4_service_promote'

wherept.fk_user=[user_id])

as
clicked_service_promo,

(selectmin(ptc.created_date)frompage_tracking_clickedptc

join page_tracking pt on pt.pk_id =


ptc.FK_tracking_id

andptc.button_id='btnPromoInvite'

andpt.step_number='4_service_promote'

MastersThesis(BusinessStudies)CopyrightsJ.Sakkers(2012)Page81of84

wherept.fk_user=[user_id])

as
date_service_promo

INTO#USER

FROMusers

select*from#IP

select*from#USER

VerwerkendeveldenmetaantallenzodatY/Nkomti.p.v.aantalclicksendejoinaanmaken

select

ip_address,

casewhen[visited1_signupFNV]>0then'Y'else'N'endas[visited1_signupFNV],

date_1_signupFNV,

casewhen[visited2_signupFNV]>0then'Y'else'N'endas[visited2_signupFNV],

casewhen[visited3_signupFNV]>0then'Y'else'N'endas[visited3_signupFNV],

casewhen[visited4_signupFNV]>0then'Y'else'N'endas[visited4_signupFNV],

casewhenclicked_linkedInFNV>0then'Y'else'N'endasclicked_linkedInFNV,

date_linkedInFNV,

casewhenclicked_TweetFNV>0then'Y'else'N'endasclicked_TweetFNV,

date_TweetFNV,

casewhenclicked_FacebookFNV>0then'Y'else'N'endasclicked_FacebookFNV,

date_FacebookFNV,

casewhenclicked_GoogleFNV>0then'Y'else'N'endasclicked_GoogleFNV,

date_GoogleFNV,

casewhen[visited1_signup]>0then'Y'else'N'endas[visited1_signup],

date_1_signup,

casewhen[visited2_signup]>0then'Y'else'N'endas[visited2_signup],

casewhen[visited3_signup]>0then'Y'else'N'endas[visited3_signup],

casewhenclicked_linkedIn>0then'Y'else'N'endasclicked_linkedIn,

date_linkedIn,

casewhenclicked_Tweet>0then'Y'else'N'endasclicked_Tweet,

date_Tweet,

casewhenclicked_Facebook>0then'Y'else'N'endasclicked_Facebook,

date_Facebook,

casewhenclicked_Google>0then'Y'else'N'endasclicked_Google,

date_Google,

[user_id],

registration_date,

companyname,

MastersThesis(BusinessStudies)CopyrightsJ.Sakkers(2012)Page82of84


firstname,

initials,

middlename,

lastname,

gender,

casewhen[visited1_wizard]>0then'Y'else'N'endas[visited1_wizard],

contacts_imported,

casewhen[visited2_wizard]>0then'Y'else'N'endas[visited2_wizard],

casewhenclicked_wizard_UpdateStatus>0then'Y'else'N'endasclicked_wizard_UpdateStatus,

date_wizUpdateStatus,

casewhen[visited3_wizard]>0then'Y'else'N'endas[visited3_wizard],

casewhenclicked_wizMailToAdress>0then'Y'else'N'endasclicked_wizMailToAdress,

casewhenclicked_PromoInvite>0then'Y'else'N'endasclicked_PromoInvite,

date_PromoInvite,

casewhen[visited1_marketplace]>0then'Y'else'N'endas[visited1_marketplace],

casewhen[visited2_assignment]>0then'Y'else'N'endas[visited2_assignment],

casewhen[visited2_product]>0then'Y'else'N'endas[visited2_product],

casewhen[visited2_service]>0then'Y'else'N'endas[visited2_service],

casewhen[visited3_assign_matching]>0then'Y'else'N'endas[visited3_assign_matching],

case when [visited 3_product_statusupdate] > 0 then 'Y' else 'N' end as [visited
3_product_statusupdate],

casewhenclicked_product_UpdateStatus>0then'Y'else'N'endasclicked_product_UpdateStatus,

date_productUpdateStatus,

case when [visited 3_service_statusupdate] > 0 then 'Y' else 'N' end as [visited
3_service_statusupdate],

casewhenclicked_service_UpdateStatus>0then'Y'else'N'endasclicked_service_UpdateStatus,

date_serviceUpdateStatus,

case when [visited 4_assign_statusupdate] > 0 then 'Y' else 'N' end as [visited
4_assign_statusupdate],

casewhenclicked_assign_UpdateStatus>0then'Y'else'N'endasclicked_assign_UpdateStatus,

date_assignUpdateStatus,

casewhen[visited4_product_promote]>0then'Y'else'N'endas[visited4_product_promote],

casewhenclicked_product_MailToAdress>0then'Y'else'N'endasclicked_product_MailToAdress,

casewhenclicked_product_promo>0then'Y'else'N'endasclicked_product_promo,

date_product_promo,

casewhen[visited4_service_promote]>0then'Y'else'N'endas[visited4_service_promote],

casewhenclicked_service_MailToAdress>0then'Y'else'N'endasclicked_service_MailToAdress,

casewhenclicked_service_promo>0then'Y'else'N'endasclicked_service_promo,

date_service_promo

from#IPip

leftjoin#USERusonip.ip_address=us.ipaddress

MastersThesis(BusinessStudies)CopyrightsJ.Sakkers(2012)Page83of84


DROPTABLE#USER

DROPTABLE#IP

MastersThesis(BusinessStudies)CopyrightsJ.Sakkers(2012)Page84of84

Vous aimerez peut-être aussi