Christian Schulze, Lisa Schöler, & Bernd Skiera
Not All Fun and Games: Viral
Marketing for Utilitarian Products
The success of products such as FarmVille has prompted many firms to engage in viral marketing on Facebook
and other social media websites. Yet is the viral marketing approach adopted for games suitable for other, more
utilitarian products? This study aims to answer questions that link product characteristics and contexts to viral
marketing success: Should primarily utilitarian products rely on the same sharing mechanisms for their viral
marketing campaigns as less utilitarian products? If not, why is this the case, and how should viral marketing for
primarily utilitarian products differ? This empirical study analyzes the Facebook viral marketing campaigns of 751
products and reveals that the same sharing mechanism that made FarmVille so successful is the worst possible
mechanism for promoting primarily utilitarian products. These findings are in line with theory from social
psychology: because consumers do not visit Facebook to learn about utilitarian products, they rely on simple cues
and heuristics to process viral marketing messages about these products. This study thus contributes to literature
on viral marketing in general and sharing mechanisms in particular; it also offers practical, hands-on
recommendations to marketing managers in charge of designing viral marketing campaigns for both more and less
utilitarian products.
Keywords: social media marketing, viral marketing, utilitarian, sharing mechanisms, Facebook, apps
that reached as many as 100 million customers in just a few
weeks (Schroeder 2011). Such success stories invariably
attract the interest of marketing practitioners eager to learn
the “recipe” for viral marketing success. If FarmVille and
similar products represent best-practice examples, we might
predict that the key ingredient is their sharing mechanism:
unsolicited broadcast messages from friends that contain
incentives to use the product. This view has found support
in initial research highlighting the importance of sharing
mechanisms for viral marketing success (Aral and Walker
2011).
However, for many firms aiming to follow FarmVille’s
lead to engage in viral marketing on Facebook, it is not at
all clear whether the findings from existing research and
best-practice examples actually translate to their own more
utilitarian products, for which consumer value is the result
of product usefulness, rather than fun. Consumers use Facebook and many other social network platforms primarily for
fun and entertainment; does this hedonic orientation influence their perception of viral marketing messages for more
utilitarian products?
Academic literature has remained largely silent about
viral marketing for utilitarian products, and yet several indications have suggested important distinctions. Chiu et al.
(2007) find that hedonic and utilitarian content in marketing
messages (not the product itself) differentially affects the
success of viral marketing campaigns. Pöyry, Parvinen, and
Malmivaara (2013) find that consumers claiming a primarily
utilitarian (rather than hedonic) motivation to visit a travel
agency’s Facebook presence are more likely to recommend
that company to their friends. In related word-of-mouth
research, Berger and Milkman (2012) show that high or low
emotional involvement—often linked to hedonic and utili-
iral marketing is not a new phenomenon; as early as
1996, the free e-mail provider Hotmail launched one
of the first viral marketing campaigns by attaching a
promotional message to every e-mail sent through its service. Today, viral marketing has become a mainstream marketing instrument used by multinational firms in various
industries, including Nike, Unilever, and Volkswagen.
Firms appreciate viral marketing as a means to benefit from
the additional trust and attention that messages from friends
and other trusted sources receive (Bampo et al. 2008;
Phelps et al. 2004) while remaining largely in control of the
message content (Van der Lans et al. 2010). The recent
trend toward viral marketing also has been fueled by the
growing popularity of social network platforms such as
Facebook, on which more than 1 billion connected consumers share marketing messages with a single click on
their computers or mobile devices. The resulting potential
for viral marketing campaigns on Facebook is evident from
popular examples such as FarmVille and CityVille—games
V
Christian Schulze is Assistant Professor of Marketing, Frankfurt School of
Finance & Management (e-mail: email@christian-schulze.de). Lisa
Schöler is a consultant and former doctoral student in Marketing, Goethe
University Frankfurt, and a member of the E-Finance Lab (e-mail:
lschoeler@wiwi.uni-frankfurt.de). Bernd Skiera is Professor of Marketing,
Faculty of Business and Economics, Goethe University Frankfurt, and a
member of the board of the E-Finance Lab (e-mail: skiera@skiera.de).
This research was financially supported by the E-Finance Lab at the
House of Finance in Frankfurt. The authors acknowledge three anonymous JM reviewers for their great feedback and suggestions. The authors
are also grateful for comments and suggestions provided by Myriam
Bechtoldt, Jan Becker, Jan Landwehr, Raji Srinivasan, Nils Stieglitz, and
Michael Trusov as well as from seminar participants at University of Texas
at Austin. Leigh McAlister served as area editor for this article.
© 2014, American Marketing Association
ISSN: 0022-2429 (print), 1547-7185 (electronic)
1
Journal of Marketing
Vol. 78 (January 2014), 1–19
tarian features—shapes consumers’ sharing behaviors.
Moreover, consumers’ reactions to viral marketing seemingly should differ for more versus less utilitarian products,1
according to theory from social psychology. Schema theory
(Aronson, Wilson, and Akert 2012; Bartlett 1932) suggests
that people unconsciously use situation-specific mental
structures to organize and filter information. Thus, consumers will pay little attention to viral marketing messages
for primarily utilitarian products in a fun- and entertainment-oriented environment such as Facebook. According to
the elaboration likelihood model (ELM; Petty and Briñol
2012; Petty and Cacioppo 1986), information processing
about these products will therefore be less deliberate and
instead rely more on heuristics, such as social cues. As a
result, sharing mechanism characteristics (e.g., unsolicited
broadcast messages from friends that contain incentives to
use the product) that work very well for products such as
games might be ineffective or even harmful for marketing
utilitarian products.
In this article, we investigate two main research questions: Should primarily utilitarian products rely on the same
sharing mechanisms that less utilitarian products use for
their viral marketing campaigns? If not, why is this the case,
and how should viral marketing for primarily utilitarian
products differ? To answer these questions, we use extensive information about viral marketing campaigns on Facebook for 751 products observed over the course of one year.
With this research, we contribute to literature on viral
marketing in general and sharing mechanisms in particular.
Viral marketing literature lacks insights into the impact of
products’ low- or high-utilitarian characteristics on viral
marketing success. We show that failing to consider these
product characteristics when designing viral marketing
campaigns can be devastating to their success. We also present a substantive contribution to literature on sharing
mechanisms in viral marketing by extending an existing
single-case study by Aral and Walker (2011), in terms of
size and dimensions covered. As a result, we can provide
detailed practical recommendations to marketing managers,
including which sharing mechanism characteristics (not) to
choose when designing viral marketing campaigns for both
low- and high-utilitarian products.
This article proceeds as follows: First, we link our
research to existing literature in viral marketing and detail
the motivation for this study. After we present the conceptual background, we derive the relevant characteristics of
sharing mechanisms in viral marketing campaigns and then
use information processing theory to develop hypotheses
about their effects on success for low- versus high-utilitarian
products. Next, we detail the setup and model for our
empirical study and present the results. We conclude with a
summary, a discussion of our research’s implications for
marketing theory and practice, and some limitations and
opportunities for further research.
1We
focus on products’ utilitarian dimension (high vs. low). We
leave it to the reader to regard the utilitarian and hedonic characteristics as either two opposite ends of the same scale (e.g., Drolet,
Williams, and Lau-Gesk 2007) or two separate dimensions (e.g.,
Voss, Spangenberg, and Grohmann 2003).
2 / Journal of Marketing, January 2014
Previous Empirical Literature on
Viral Marketing
Extant empirical research has primarily focused on two
determinants of viral marketing success: the characteristics
of the content (Chiu et al. 2007; Dobele et al. 2007; Teixeira, Wedel, and Pieters 2012; Tucker 2011) and the optimal
seeding strategies given the social network structure
(Bampo et al. 2008; De Bruyn and Lilien 2008; Hinz et al.
2011; Lee, Lee, and Lee 2009; Van der Lans et al. 2010).
Whereas the “what” (content) and “who” (senders and
receivers) of viral marketing have been researched extensively, the “how” (sharing mechanisms) has received little
attention, despite relevant case evidence that suggests the
mechanisms consumers use to share viral messages exert
strong influences on viral marketing success (Aral and
Walker 2011). Beyond the determinants of viral marketing
success, studies have investigated the manageability of viral
marketing campaigns (Kalyanam, McIntyre, and Masonis
2007; Leskovec, Adamic, and Huberman 2007; Van der
Lans et al. 2010) and how their effectiveness compares with
that of traditional media (Toubia, Stephen, and Freud 2009).
Whereas in these studies, the firms create and largely
remain in control of the viral marketing messages, Godes
and Mayzlin (2009) analyze a unique form of communication among consumers, in which the firm initiates the
process but does not prescribe the message content. Firms
also might adopt consumer-generated messages for commercial purposes (Joshi and Trusov 2012), manipulate the
word-of-mouth process (Dellarocas 2006), or use specific
brand characteristics to influence word of mouth (Lovett,
Peres, and Shachar 2013). We summarize prior empirical
studies on viral marketing in Table 1.
Our research contributes to this empirical literature in
two ways. In a narrow sense, we add to emergent research
on the impact of sharing mechanisms by (1) addressing all
four dimensions that distinguish and define them and (2)
demonstrating their relevance for viral marketing success.
We also evaluate their effect in viral marketing campaigns
for a range of products, extending previous research focusing on in-depth analyses of one specific product. In a
broader sense, we provide the first assessment of differences in viral marketing for low- versus high-utilitarian
products. That is, previous research has relied on products
with differing utilitarian value, but our study is the first to
employ information processing theory (i.e., ELM) to specify how utilitarian value should influence the design of the
corresponding viral marketing campaign (Table 1).
Conceptual Background
Sharing Mechanism Characteristics in Viral
Marketing Campaigns
Previous research has characterized viral marketing communication along six dimensions:
1. Social position of the sender and the receiver (i.e., consumers’ connectedness or bridging function across subnetworks; see Bampo et al. 2008; Hinz et al. 2011).
Viral Marketing for Utilitarian Products / 3
Research Focus
Teixeira, Wedel, and
Pieters (2012)
This article
Tucker (2011)
Aral and Walker (2011)
Influence of viral video characteristics on reach
versus persuasiveness
Influence of viral video characteristics on user
engagement
Viral marketing for low- versus high-utilitarian
products; differences in influence of sharing
mechanism characteristics
Effectiveness of two sharing mechanisms
Effect of emotions in viral messages on
consumers’ sharing behavior
Chiu et al. (2007)
Effect of message source, content, receiver
characteristics, and Internet connection speed
on e-mail forwarding
Kalyanam, McIntyre,
Influence of firm’s viral marketing intensity and
and Masonis (2007)
aggressiveness
Leskovec, Adamic, and
Identification of communities, products, and
Huberman (2007)
pricing categories suitable for viral marketing
Lee, Lee, and Lee
Influence of sender–receiver tie characteristics
(2007)
Bampo et al. (2008)
Influence of social structure on individual
message transmission behavior
De Bruyn and Lilien
Influence of viral messages on consumer
(2008)
awareness, interest, and decisions
Toubia, Stephen, and
Comparison of viral activity (as measured by
Freud (2009)
coupon redemption) for product samples
versus traditional offline media
Van der Lans et al.
Prediction and management of viral marketing
(2010)
performance
Hinz et al. (2011)
Effectiveness of seeding strategies
Dobele et al. (2007)
Publication
1
—
—
—
—
—
—
—
1
E-mail
E-mail over program interface
E-mail
E-mail over web interface
E-mail over web interface
E-mail
Offline (paper-based coupons)
E-mail over web interface
E-mail, e-mail over web
interface; direct messages on
social network site
Direct and broadcast messages on Facebook
—
—
4
—
Diverse sharing mechanisms
on Facebook
—
1
—
—
Sharing Mechanism(s) in
Empirical Study
Yes
No
No
No
No
No
No
No
No
No
No
No
No
No
Yes
Yes
Yes
No
Yes
No
No
No
No
No
Yes
No
No
Yes
Analysis of
Differences
Dimensions on Influence of Along UtilitarSharing
ian Dimension
Which Sharing
for Studied
Mechanism(s) Mechanism
Characteristics Product(s)
Differ
TABLE 1
Previous Empirical Studies of Viral Marketing
Yes
No
No
No
No
No
No
No
No
No
No
No
No
No
Analysis of
Influence of
Utilitarian
Product
Context
2. Sender–receiver relationship (i.e., are consumers close
friends or complete strangers? For research on this dimension in the viral marketing context, see Chiu et al. 2007; De
Bruyn and Lilien 2008; Lee, Lee, and Lee 2009; Leskovec,
Adamic, and Huberman 2007; for a more general discussion, see Granovetter 1973).
3. Communication exclusivity (i.e., does communication rely
on direct, one-to-one messages or on one-to-many broadcasts? See Aral and Walker 2011; Phelps et al. 2004).
4. Expressed interest (i.e., has the receiver requested or
searched for the message, or is it unsolicited spam? See De
Bruyn and Lilien 2008; Kalyanam, McIntyre, and Masonis
2007).
5. Message features (e.g., does the message come with an
incentive to act? See Bampo et al. 2008; De Bruyn and
Lilien 2008; Hinz et al. 2011; Lee, Lee, and Lee 2009;
Leskovec, Adamic, and Huberman 2007; Van der Lans et al.
2010).
6. Message content (e.g., is the message inspirational, funny,
or shocking? See Chiu et al. 2007; Dobele et al. 2007; Teixeira, Wedel, and Pieters 2012; Tucker 2011).
Viral marketing research has extensively covered two
dimensions: social position (the “who”) and message content (the “what”). In contrast, research on the other four
dimensions has been rather selective, despite being recognized as determinants of viral marketing success. We therefore focus on these dimensions (i.e., 2–5), which collectively characterize the sharing mechanisms (the “how”) by
which viral messages spread across consumers and which
have never been analyzed jointly (for details on these
dimensions, see Table 2).
Influence of Environment on Processing
Information About Utilitarian Products
Success stories about viral marketing campaigns on Facebook and similar platforms almost exclusively involve
products that fit well with the platforms’ fun-oriented environments. However, not all products are fun oriented, and
marketing literature abounds with examples illustrating that
what works for one product does not necessarily work for
all products. Thus, to shed more light on the role of product
contexts in viral marketing campaigns, it seems paramount
to investigate the differences between products that do and
do not fit well with this environment. A popular approach
captures these essential differences by classifying products
into utilitarian and hedonic categories (e.g., Batra and
Ahtola 1991; Chitturi, Raghunathan, and Mahajan 2008;
Hirschman and Holbrook 1982; Spangenberg, Voss, and
Crowley 1997). When purchasing primarily hedonic products, such as games, consumers seek items that are fun,
exciting, delightful, thrilling, and enjoyable. Primarily utilitarian products instead should be effective, helpful, functional, necessary, and practical to appeal to consumers
(Voss, Spangenberg, and Grohmann 2003).
When consumers choose to visit social network platforms such as Facebook, they also seek fun, enjoyment,
excitement, and, ultimately, a feeling of connectedness
through social interactions and content (Hoffman and
Novak 2013). Consequently, when Facebook users receive
viral marketing messages, they do not expect the messages
to promote primarily utilitarian products, because such
4 / Journal of Marketing, January 2014
products do not fit well with their fun-focused expectations
or schema.
According to schema theory, consumers build mental
structures to organize their knowledge and channel the processing of new information (e.g., Aronson, Wilson, and Akert
2012; Bartlett 1932). Marketing research has used schemas
to explain consumer biases in self-perceptions (Wheeler,
Petty, and Bizer 2005) or media (Roskos-Ewoldsen,
Roskos-Ewoldsen, and Dillman Carpentier 2009), for
example. In our study, we employ this parsimonious theory
to predict that consumers’ fun-oriented Facebook schema
primes them to focus on messages for hedonic products. In
other words, the Facebook environment biases consumers
such that they unconsciously devote fewer mental resources
to processing messages related to primarily utilitarian products (Dijksterhuis et al. 2005).
ELM
The ELM (for a general overview, see Petty and Briñol
2012; for a marketing-related perspective, see Rucker,
Petty, and Priester 2007) describes how the scarcity of mental resources affects consumers’ information processing.
This dual-process theory from social psychology suggests
that information processing, and thus attitude shifts, occurs
through two routes. In the “central route,” consumers diligently consider product-related information and invest
thought and effort into the process. The “peripheral route,”
in contrast, is characterized by consumers’ reliance on
heuristics, social cues, and simple inferences in their attitude formation. Therefore, consumers expend little thought
elaborating on a message; instead, they might reject or
accept and act on that message on the basis of simple cues,
such as whether the sender is a trusted expert.
Consumers unconsciously devote fewer mental
resources to processing messages about products outside
their situation-specific schema (Bitner and Obermiller
1985), so the ELM theorizes that consumers’ information
processing of such products follows the peripheral route. In
other words, Facebook users seeking fun and entertainment
should be fast and frugal in their information processing and
rely primarily on heuristics, simple inferences, and social
cues when evaluating (viral marketing) messages about primarily utilitarian products. In contrast, such heuristics and
cues may be less relevant when consumers encounter messages about less utilitarian products that fit the schema and
unconsciously devote more thought to the actual message.
Conceptual Model
Because consumers’ approaches to processing viral marketing messages differ depending on the product context, the
characteristics of sharing mechanisms for viral marketing
messages should differ in their effectiveness as well. The
empirical study tests the conceptual model depicted in Figure 1, which investigates the moderating effect of primarily
utilitarian products on the effectiveness of sharing mechanism characteristics for viral marketing campaigns in a primarily fun- and entertainment-oriented setting. Drawing on
the ELM, we derive four hypotheses about the effect of
receivers’ expressed interest, message features (particularly
Viral Marketing for Utilitarian Products / 5
Description
Product
success
(reach)
Primarily
utilitarian
products
Broadcast
messages
from friends
Broadcast
messages
from
strangers
Descriptive Details from Empirical Study
A Facebook user visits a friend’s “About”
page to see which apps the friend is using.
A viral message from a friend in the inbox
that promotes, for example, a new game.
(base case for
xdirectfriend,i and
xbroadcaststranger,i)
xut,i
Y
We characterize 233 of 751 apps (31%) as
primarily utilitarian.
In our study, the average app reaches
347,998 consumers (minimum reach of 12,
maximum reach of 10,132,646).
xbroadcaststranger,i
Products employ between 0 and 6 sharing
mechanisms (mean of 1.58) that rely on
broadcast messages from friends.
Products employ between 0 and 1 sharing
mechanisms (mean of .19) that rely on
broadcast messages from strangers.
(base case for
xincentivized,i)
Products employ between 0 and 8 sharing
mechanisms (mean of 2.50) that use viral
marketing messages without incentives.
Products employ between 0 and 3 sharing
mechanisms (mean of 1.02) that rely on
direct messages from friends.
xdirectfriend,i
xincentivized,i
(base case for
xunsolicited,i)
xunsolicited,i
Mi
Notation
Products employ between 0 and 2 sharing
mechanisms (mean of .29) that use viral
marketing messages with incentives.
Products employ between 0 and 5 sharing
mechanisms (mean of 1.79) that use
unsolicited messages.
Products employ between 0 and 5 sharing
mechanisms (mean of 1.00) that use
solicited messages.
Sharing mechanisms on Facebook include a Sixty-three products do not employ any
post on a friend’s timeline or a direct
sharing mechanism (i.e., do not use viral
message to friends with incentives.
marketing), and 688 products do. The maximum number of unique sharing mechanisms
per product is 8, with a mean of 2.79.
Example in Study Context
A viral marketing message such as “Try
Spotify now and get one month of premium
membership for free!” offers an incentive to
the receiver.
A viral marketing message such as “Try
Spotify now and listen to music today!” does
not offer a specific incentive to receiver.
Two befriended Facebook users can send
each other messages to their Facebook
inbox. Direct messages on Facebook require
consumers to be direct, first-degree contacts.
Broadcast messages from strangers
When a Facebook user posts a viral marketdescribe one-to-many communications
ing message to a friend’s timeline, that
between Facebook users who are not
friend’s contacts will see this message in
directly connected, such as second-degree
their news feeds and view it as a broadcast
contacts who share a common friend.
message from a stranger.
Broadcast messages from friends describe
When a Facebook user posts a viral
one-to-many communications between
marketing message to her own timeline, her
Facebook users who are direct (first-degree) direct friends see this message in their news
contacts.
feeds as a broadcast message from a friend.
Facebook apps that, for example, enable
A product is primarily utilitarian if its main
purpose is to be effective, helpful, functional, consumers to track their stock investments
and practical (rather than fun, exciting, and
or find a new house. In the context of our
enjoyable) (Voss, Spangenberg, and
study, we classify apps as primarily utilitarian
Grohmann 2003).
if they belong to the business, money, utility,
alerts, or politics categories.
Success of a viral marketing campaign is
We operationalized reach by the maximum
typically measured by its reach, or the
(estimated) number of installations. The estinumber of individual consumers that took
mated reach for the Facebook app “Cities
action after receiving the viral marketing
I’ve Visited” is approximately 10 million unique
message (e.g., Hinz et al. 2011).
consumers; the estimated reach for “City
Trip” is only approximately 1,000 consumers.
Number of
Sharing mechanisms should encourage
sharing
consumers to recommend product to their
mechanisms peers. Each sharing mechanism has four
dimensions. Viral marketing campaigns
can use several sharing mechanisms
simultaneously but must rely on at least one.
Unsolicited
Messages are unsolicited if receiver has not
messages
expressed interest in receiving the message
(e.g., De Bruyn and Lilien 2008).
Solicited
Messages are solicited if the message
messages
receiver has expressed an interest in
receiving the message or has searched for it
(e.g., Diamond and Noble 2001).
Messages with Incentives in viral marketing messages
incentives
provide greater benefits for message
receivers (compared with regular
consumers) if they use the product.
Messages
Without incentives, receivers of viral marketwithout
ing messages have no benefit over regular
incentives
consumers from using the product.
Direct
A direct message from a friend describes
messages
one-to-one communications (comparable to
from friends e-mail) between direct (first-degree) contacts.
Construct
TABLE 2
Overview of Constructs
FIGURE 1
Conceptual Model
Moderator
Primarily utilitarian
product
Sharing Mechanism Characteristics
Success Measure
H 1: –
Unsolicited messages
H 2: –
Messages with incentives
Direct messages
from friends
Broadcast messages
from strangers
incentives), communication exclusivity, and sender–receiver
relationships in viral marketing campaigns for low- versus
high-utilitarian products. Specifically, we hypothesize a
negative effect for unsolicited messages (vs. solicited messages) and messages with incentives (vs. messages without
incentives). We also expect a positive moderating effect for
direct messages from friends (vs. broadcast messages from
friends) and for broadcast messages from strangers (vs.
broadcast messages from friends). For more detailed
descriptions and examples of the focal constructs used in
this study, see Table 2.
Hypotheses
Receivers’ expressed interest. When information processing for high-utilitarian products follows the peripheral
route and consumers rely more on heuristics and simple
inferences (Petty and Cacioppo 1986), the ELM predicts
that they regard unsolicited push messages as a signal of the
product’s low quality and low personal relevance compared
with information they (actively) seek out (the thought
being, “If the product were truly useful and of high quality,
the company would not have to market it so aggressively”).
In contrast, if information processing follows a more central
route, we expect a much smaller negative effect or even
none at all from unsolicited messages. Consequently, we
hypothesize a moderating effect of primarily utilitarian
products:
H1: The effectiveness of sharing mechanisms that rely on
unsolicited viral marketing messages for which the
receiver has not expressed a prior interest (rather than
solicited messages) is lower for high-utilitarian products
than for low-utilitarian products.
Message features. According to the ELM, when information processing follows the peripheral route, consumers
6 / Journal of Marketing, January 2014
Product success
(reach)
H 3: +
H4: +
do not actively evaluate the usefulness of an incentive
included in a viral marketing message. Instead, they regard
the incentive as a signal of inferior product quality (“If the
product were good, the company would not have to offer
me an incentive”), which reduces the effectiveness of the
viral marketing message. At the same time, we expect a
positive effect from incentives for less utilitarian products
when consumers evaluate their usefulness more thoroughly.
H2: The effectiveness of sharing mechanisms that rely on viral
marketing messages containing an incentive for the
receiver to use the product (rather than no incentive) is
lower for high-utilitarian products than for low-utilitarian
products.
Communication exclusivity. When information processing follows the peripheral route, consumers do not expend
as much thought to elaborate on the merits of a broadcast
message. Instead, they believe that the lack of a direct, individual message signals low personal relevance (“Had my
Facebook friend thought this were truly relevant for me, she
would have sent me a direct message”). In contrast, we do
not anticipate a significant difference between direct and
broadcast messages from Facebook friends when consumers spend more thought evaluating the actual message,
as they should for low-utilitarian products.
H3: The effectiveness of sharing mechanisms that rely on
direct messages (rather than broadcast messages) from
Facebook friends2 is greater for high-utilitarian products
than for low-utilitarian products.
2In H , we compare the effect of broadcast and direct messages
3
from friends rather than strangers, because social networks such as
Facebook typically do not allow direct messages from strangers.
The same applies to H4, in which we only compare the effect of
messages from friends and strangers for broadcast (not direct)
messages.
Sender–receiver relationship. The moderating effect of
primarily utilitarian products on the effectiveness of messages from friends versus strangers should result from two
simultaneous effects. First, consumers seeking fun and
entertainment on Facebook rely on their friends for recommendations, because friends often have similar tastes and
preferences. Thus, recommendations from friends are useful to consumers, regardless of whether they evaluate these
recommendations on the basis of their objective, informational value (central route) or simply by following a heuristic (peripheral route) (Koroleva, Krasnova, and Günther
2011). We therefore expect viral marketing messages from
friends that promote hedonic products to be more effective
than messages from strangers.
Second, because consumers evaluate the quality of utilitarian products on the basis of their effectiveness, helpfulness, functionality, and usefulness (Voss, Spangenberg, and
Grohmann 2003), if they follow the peripheral route. They
should value information from a competent, neutral, external source at least as much as information from a friend
with similar views and experiences. Godes and Mayzlin
(2009, Table 3) support this notion, showing that word-ofmouth recommendations from actual customers are equally
effective whether they involve friends or strangers. Thus,
for utilitarian products, messages from strangers can be as
effective as messages from friends, but messages from
strangers should be significantly less effective for less utilitarian products.
H4: The effectiveness of sharing mechanisms that rely on broadcast messages from strangers (rather than friends) is greater
for high-utilitarian products than for low-utilitarian products.
Empirical Study
Setup
We test our proposed conceptual model with a large-scale
empirical study based on 751 Facebook apps (for details,
see Appendix A). Facebook apps are an ideal setting for this
study because they spread primarily through viral marketing, which minimizes potential confounding effects. Facebook currently lacks a convenient “app store” for users to
discover interesting new offerings, so producers must rely
on consumers’ recommendations. Typically, apps actively
attempt to initiate the viral marketing process by asking
users for permission to send a predefined marketing message on their behalf. For example, apps might request permission to post viral messages on users’ Facebook timeline
(Figure 2). Because all apps on Facebook are currently free,
there should not be any confounding price effects. To use an
app, a new consumer simply confirms the installation on his
or her Facebook page; after this initial installation, the consumer can return to use the app at any time.
For our study, we collected information about the number and characteristics of the sharing mechanisms employed
in each app. A team of coders, thoroughly trained by the
authors, evaluated the app characteristics. The average time
needed to rate an app was 30 minutes. Two coders independently coded approximately one-third of all apps. The interrater reliability achieved a correlation of .8; any differences
were discussed until agreement was reached. In our sample,
the average number of sharing mechanisms per app was
2.79, and 64% of all sharing mechanisms relied on unsolicited messages, whereas only 11% employed incentives.
In addition, 57% of sharing mechanisms used broadcast
messages from friends, 37% used direct messages from
friends, and only approximately 7% used broadcast messages from strangers.
For the apps in our study, we observed the product categories (22 categories in total, including, e.g., games, entertainment, business, money; for a complete list, see Table A2
in Appendix A) as selected by the app producer. Our sample
included the top 20 apps (in terms of number of installations) for each category as well as a random selection of
remaining apps. For the dichotomous classification of each
app as either more or less utilitarian, we relied on the category choice of the app developer. We considered an app
more utilitarian if it belonged to the business, money, utility, alerts, or politics categories (n = 233) and less utilitarian
otherwise (n = 518). To control for confounding effects, we
tracked the types of network effects that each app used and
included measures for app quality. We could not observe the
content of sent messages, which were prescribed by the app
and thus very similar across sharing mechanisms, to the
benefit of our study. Nor could we capture differences in
message characteristics between apps. Our data set thus
contains aggregate information on the app level; we did not
observe the behavior of individual consumers.
To analyze app success, we relied on daily information
about the number of installations for each app (for details,
see Table 2), which indicated product reach, a common success measure in practice that is widely used in viral marketing research (see, e.g., Bampo et al. 2008; De Bruyn and
Lilien 2008; Hinz et al. 2011; Van der Lans et al. 2010).
Although we observed the daily number of installations
over the course of an entire year, we faced censored data
about app success, because some apps existed before the
first day of our observation period and others were
launched near the end of the period such that they remained
in their growth phase. To achieve performance comparability across products, we estimated a diffusion curve for each
app using the technology substitution model (TSM; for
applications in marketing, see Gupta, Lehmann, and Stuart
2004; Schulze, Skiera, and Wiesel 2012), which is conceptually similar but more robust than the recursive Bass
model (Kim, Mahajan, and Srivastava 1995; for technical
details, see Appendix A). We then used the estimated maximum number of installations for each app as the primary
success measure in our analyses. We also replicated our
results with alternative success measures, such as the number of daily active users and the growth rate.
Statistical Model
We established our statistical model to test the notion that
utilitarian apps moderate the effectiveness of sharing mechanism characteristics on app success. As Equation 1 shows,
the model includes a parameter for the number of sharing
mechanisms Mi per app i (), because apps can use more
than one sharing mechanism. The four count variables
Viral Marketing for Utilitarian Products / 7
FIGURE 2
Screenshot: App Request to Post on User’s Timeline
(xunsolicited, i, xincentivized, i, xdirectfriend, i, and xbroadcaststranger, i)
capture how many of the Mi sharing mechanisms that an
app used exhibited a certain characteristic. We capture the
main effect of each sharing mechanism characteristic (e.g.,
unsolicited) and the main effect of high-utilitarian apps (ut).
To test H1–H4, we included the two-way interactions of
sharing mechanism characteristics and primarily utilitarian
apps (4 parameters, e.g., unsolicited_ut).
In addition, we integrated numerous control variables.
To control for category-specific effects, beyond the distinction into less and more utilitarian apps, we included dummy
variables for app categories (22 parameters, e.g.,
category_business). We also used control variables for five
types of positive network effects coded as dummy variables
8 / Journal of Marketing, January 2014
(5 parameters, e.g., networkexternalities) and their respective
interactions with high-utilitarian apps (5 parameters, e.g.,
networkexternalities_ut). The control variables for network
effects captured whether an app employed specific network
effects that might influence its diffusion. In particular, these
parameters captured the effect of apps without any base
utility, when used in isolation (noBaseUtility),3 such that the
app required connected peers to be of any value to the user
(similar to telephone or e-mail). We also captured the effect
for apps that encouraged interactions among users within
3Katz and Shapiro (1985, p. 426) describe base utility as the
“basic willingness to pay” (r) for the product, excluding consideration of network effects (see their Equation 1).
a forum (forum), allowed users to comment on others’
contributions (comment), or encouraged any other kind of
interaction (interaction). Finally, to capture other types of
positive network externalities, we employed a fifth parameter (networkexternalities) related to network externalities in
apps with a base utility (comparable to smartphones, which
offer functionality even without connected peers).
Finally, we included two measures to control for unobserved app quality: the number of fans and the number of
reviews an app had received. We observed both metrics at
the end of the observation period. To ensure they were comparable across apps of different ages (i.e., older apps have
had more time to accumulate fans and reviews), we divided
the recorded number of reviews and fans by the number of
days the app had been available. Thus, we captured the
main effects of fans per day and reviews per day (reviews,
fans), their two-way interactions with high-utilitarian apps
(reviews_ut, fans_ut) and with sharing mechanism characteristics (eight parameters; e.g., reviews_unsolicited), and their
three-way interactions with high-utilitarian apps and
sharing mechanism characteristics (eight parameters; e.g.,
reviews_ut_unsolicited). We did not include users’ star ratings as
a measure of app quality, because this measure was available for only a fraction of the observed apps. Correlations
between variables were generally low and should not affect
our analysis (Table A1 in Appendix A). As an overview, we
summarize the conceptual definition, empirical operationalization, and descriptive results for our focal constructs in
Table 2.
We used a standard multiplicative model with dummy
variables to explain the success of Facebook apps (Y),
because it is intuitive that sharing mechanism characteristics influence the relative success of Facebook apps rather
than add a fixed number of additional users. Equation 1 is
the transformed, log-linear version of the model, which we
used to analyze our data:
( )
(1) ln ( Yˆ i ) = + β incentivized × x incentivized, i + β directfriend × x directfriend, i
(1) ln ( Yˆ i ) = + β broadcaststranger × x broadcaststranger, i + β ut × x ut, i
ˆ )=+β
(1) ln ( Y
i
unsolicited _ ut × x unsolicited, i × x ut, i
ˆ i ) = + β incentivized _ ut × x incentivized, i × x ut, i
(1) ln ( Y
ˆ i ) = + β directfriend _ ut × xdirectfriend, i × x ut, i
(1) ln ( Y
ˆ i ) = + β broadcaststranger _ ut × x broadcaststranger, i × xut, i
(1) ln ( Y
ˆ )=+γ
(1) ln ( Y
i
category_business × x category_business, i + ...
ˆ i ) = + γ category_video × x category_video, i
(1) ln ( Y
ˆ i ) = + γ networkexternalities × x networkexternalities, i + ...
(1) ln ( Y
ˆ i ) = + γ interaction × x interaction, i
(1) ln ( Y
ˆ )=+γ
(1) ln ( Y
i
networkexternalities_ut × x networkexternalities, i × x ut, i
ˆ i ) = + ... + γ interaction_ut × x interaction, i × x ut, i
(1) ln ( Y
ˆ i = α ′ + θ × M i + β unsolicited × xunsolicited, i
(1) ln Y
( )
(1) ln ( Yˆ i ) = +γ reviews_ut × x reviews, i × x ut, i
(1) ln ( Yˆ i ) = + γ fans_ut × x fans, i × x ut, i
(1) ln ( Yˆ i ) = + γ reviews_unsolicited × x reviews, i × x unsolicited, i + ...
(1) ln ( Yˆ i ) = + γ reviews_broadcaststranger × xreviews, i × x broadcaststranger, i
(1) ln ( Yˆ i ) = + γ fans_unsolicited × x fans, i × xunsolicited, i + ...
(1) ln ( Yˆ i ) = + γ fans_broadcaststranger × x fans, i × xbroadcaststranger, i
(1) ln ( Yˆ i ) = + γ reviews_ut_unsolicited × xreviews, i × xut, i × x unsolicited, i
(1) ln ( Yˆ i ) = + ... + γ reviews_ut_broadcaststranger × xreviews, i × xut, i
(1) ln ( Yˆ i ) = × xbroadcaststranger, i + γ fans_ut_unsolicited × xfans, i
(1) ln ( Yˆ i ) = × xut, i × x unsolicited, i + ...
(1) ln ( Yˆ i ) = + γ fans_ut_broadcaststranger × x fans, i × x ut, i
(1) ln ( Yˆ i ) = × xbroadcaststranger, i
(1) ln Yi = +γ reviews × x reviews, i + γ fans × xfans, i
Endogeneity
We addressed potential endogeneity in our model in two
ways. First, we included control variables for app categories,
network effects, and the number of reviews and fans per
app as an indicator of unobserved app quality in the statistical model. The results for H1–H4 remained unchanged—
indeed, the parameters even increased in significance—
when we added these control variables, and the explained
variance increased substantially as well.
Second, we investigated potential endogeneity concerns
through in-depth interviews with 23 industry professionals
(marketing managers directly responsible for firms’ Facebook
activities, app programmers, and agency representatives),
which helped us better understand the data-generating process.
Among other things, we discussed three potential endogeneity concerns in our data set (for details on these questions and respondents’ answers, see Appendix B, Table B1).
First, if programmers and designers knew a priori which
sharing mechanism characteristics were superior and used
them more frequently, this would create endogeneity. However, average agreement with the item “I am fully aware of
which sharing mechanisms work and which do not” was
only 4.4 on a ten-point scale (1 = “do not agree,” and 10 =
“fully agree”). The respondents also disagreed widely about
their specific expectations. Therefore, we do not expect systematic prior expectations about the superiority of certain
characteristics to have an influence on our data; we find
support for this assumption in the relative frequency of the
various characteristics (i.e., some of the most effective characteristics were rarely used).
Second, endogeneity might result if app developers
used certain characteristics only as a “last resort” for apps
with otherwise very low chances of success. We cannot
entirely rule out this possibility, but the wide disagreement
with this notion among industry professionals (average
Viral Marketing for Utilitarian Products / 9
influence of primarily utilitarian products on the effectiveness of sharing mechanism characteristics. We find a significant negative effect of unsolicited messages in viral
marketing (−1.37, p < .01) for utilitarian products, which
confirms H1. Similarly, incentives for the receiver in viral
marketing messages that promote utilitarian products are
significantly less effective (−2.05, p < .05), which confirms
H2. Sharing mechanisms that use direct messages from
friends are significantly more effective for utilitarian products (1.84, p < .01), in support of H3. Finally, we find that
broadcast messages from strangers are significantly more
effective for utilitarian products (3.83, p < .01), in line with
H4. These effects are not only statistically significant but
also substantial. For example, unsolicited messages are
75% less effective (1 – exp(–1.37)) for primarily utilitarian
products. The results are consistent across the six regressions in Table 3, which differ in the number of focal and
control variables they use (for details on the control
variables’ parameters, see Appendix A).
agreement of 2.7 on a ten-point scale) suggests that such
behavior is not common.
Third, endogeneity could arise if app developers tailored their sharing mechanisms to certain types of apps
(e.g., utilitarian). In our survey, industry professionals concurred with the assertion that sharing mechanisms in less
utilitarian leisure apps should function differently than
those in primarily utilitarian business-related apps (7.7 on a
ten-point scale). However, when asked about their specific
expectations, most professionals answered that they had
“no idea.” Those who formulated specific expected differences varied greatly in their predictions. It is thus very
unlikely that endogeneity poses a problem for our analysis.
Results
Hypotheses Testing
In Table 3, we present the results for the full model (Equation 1) in the last column. The results confirm a moderating
TABLE 3
Effects of Sharing Mechanism Characteristics on Product Reach
1
Number of sharing mechanisms
.69***
(11.23)
Sharing Mechanism Characteristics
Unsolicited message
Message with incentive
Direct message from friend
Broadcast message from stranger
2
3
.82***
(7.17)
.73***
(6.27)
.63***
(5.53)
–.51
(–1.45)
1.14***
(3.49)
.59
(1.58)
–.54
(–.98)
–.28
(–.71)
1.49***
(4.18)
.31
(.68)
–1.18*
(–1.89)
–.09
(–.23)
1.09***
(2.77)
–.03
(–.07)
–1.46**
(–2.39)
–.02
(–.05)
1.38***
(3.66)
–.39
(–.94)
–1.34**
(–2.23)
.02
(.07)
1.36***
(3.58)
–.36
(–.85)
–1.44**
(–2.36)
–.90*
(–1.68)
–2.37***
(–2.68)
1.28
(1.60)
3.30***
(2.84)
–1.21**
(–2.35)
–1.84*
(–1.95)
1.33*
(1.78)
3.55***
(3.03)
–1.00*
(–1.88)
–2.00**
(–2.31)
1.23*
(1.75)
3.12***
(2.75)
–1.37***
(–2.58)
–2.05**
(–2.27)
1.84***
(2.62)
3.83***
(3.11)
.16
(.34)
1.03
(1.19)
Yes
.95
(1.10)
Yes
Yes
.37
(.48)
Yes
Yes
Yes
7.10***
(4.79)
751
.272
.240
6.97***
(4.71)
751
.318
.279
Direct message from friend utilitarian
Broadcast message from stranger utilitarian
Control for App Type
Utilitarian
N
R-square
Adjusted R-square
6.58***
(28.20)
751
.160
.155
6.43***
(21.62)
751
.175
.164
*p < .1.
**p < .05.
***p < .01.
Notes: t-values are in parentheses. We use robust standard errors in all regressions.
10 / Journal of Marketing, January 2014
6
.77***
(6.92)
Message with incentive utilitarian
6.64***
(34.47)
751
.129
.128
5
.75***
(6.61)
Moderator for Sharing Mechanism
Characteristics in Primarily Utilitarian Apps
Unsolicited message utilitarian
Controls for App Categories
Controls for Network Effects
Controls for App Quality (Reviews and Fans)
Constant
Constant
4
7.70***
(5.80)
751
.406
.353
Investigation of Moderating Effects
To better understand what drives the observed moderating
effects from utilitarian apps, we further investigate the
interaction effects. Figure 3 shows that the confirmation of
H1 stems from a significant negative effect of unsolicited
messages for high-utilitarian products. For low-utilitarian
products, we do not observe a significant difference for
solicited versus unsolicited viral marketing messages (as
the dotted line indicates). Both findings are in line with our
hypothesis. Regarding H3, we find a similar pattern, in that the
significant positive slope for high-utilitarian products drives
the observed effect. Primarily utilitarian apps benefit from the
increased personal relevance signaled by direct viral marketing messages from friends. We do not observe a significant
slope for low-utilitarian products, which indicates that broadcast and direct messages for these products work equally well.
Both findings—the superiority of direct messages for
high-utilitarian products and the insignificant difference
between broadcast and direct messages for low-utilitarian
products—challenge the generalizability of previous find-
ings (Aral and Walker 2011). Although we can only speculate that these differences might stem from the greater product variety we include in this study, our findings are fully in
line with the hypothesized effects.
Regarding H4, Figure 3 shows the expected significant
negative slope for low-utilitarian products. For high-utilitarian
products, however, broadcast messages from strangers are
not just equally but even more effective than broadcast
messages from friends. As we have noted, the significant
positive slope might reflect consumers’ desire for unbiased
outside information that complements their own perspective
regarding the functionality and usefulness of high-utilitarian
products, which seems to be more influential than recommendations from friends with often similar views.
Finally, we find the expected significant positive effect
from incentives in viral marketing messages for low-utilitarian
products (H2), but we do not observe a significant negative
effect from incentives for high-utilitarian products. In other
words, incentives are not detrimental to viral marketing
success for these products, just ineffective. A potential reason for this finding is that the value of the actual benefit of
FIGURE 3
Moderating Effects of Primarily Utilitarian Products
B: H2
H1
Product Success (Reach)
Product Success (Reach)
A: H1
Solicited Message
Unsolicited Message
H2
Message Without Incentive
Product Success (Reach)
Product Success (Reach)
Message with Incentive
D: H4
H3
Low utilitarian
High-ut
High-ut
Sl
ope n
Slope
Slope
Slope s
C: H3
Broadcast Message
from Friend
Low
Low uti
uti
Direct Message
from Friend
High utilitarian
H4
Broadcast Message
from Friend
Slope not significant
Broadcast Message
from Stranger
Slope significant (p < .05)
Viral Marketing for Utilitarian Products / 11
the incentive (e.g., one free month of premium membership) is comparable to the negative quality signal such that
we observe an insignificant overall effect.
Robustness Checks
To increase confidence in our results, we used the full
model (Equation 1) with a different sample as well as with
other dependent variables. We present the results for several
alternative success measures in Table 4. In the first column,
we present the full original model again for reference.
Estimated reach without the top 20 apps. In our original
analysis, we selected the top 20 apps and a random selection of remaining apps in each category. This selection
process might lead to biased results if sharing mechanism
characteristics in top 20 apps work differently than those
for other apps. We therefore excluded the top 20 apps from
the regression in the second column. Although the parame-
ter and significance levels changed slightly, the results
remain the same with this smaller sample.
Reach as observed in the data. To account for censored
data, we used the TSM to estimate the maximum number of
installations for each app in our original analysis. This estimation might lead to improper dependent variables if the
functional form of the TSM were unsuitable. For this
regression, we did not estimate the reach of each app;
instead, we used the highest observed number of installations in the observation period as an alternative dependent
variable. As we show in the third column of Table 4, the
results from this analysis were almost identical to those
from our original approach.
Usage as observed in the data. Reach is a popular, commonly used measure of viral marketing success, but other
potentially relevant success measures exist. Our interviews
with industry professionals indicated that the number of
active users and the growth rate are relevant, so we used the
TABLE 4
Robustness Checks: Effects of Sharing Mechanism Characteristics on Alternative Success Measures
Reach
(Estimated)
Number of sharing mechanisms
Sharing Mechanism Characteristics
Unsolicited message
Message with incentive
Direct message from friend
Broadcast message from stranger
Moderator for Sharing Mechanism
Characteristics in Primarily Utilitarian Apps
Unsolicited message utilitarian
Message with incentive utilitarian
Direct message from friend utilitarian
Broadcast message from stranger utilitarian
Control for App Type
Utilitarian
Controls for App Categories
Controls for Network Effects
Controls for App Quality (Reviews and Fans)
Constant
Constant
N
R-square
Adjusted R-square
.63***
(5.53)
Reach Without
Top 20
Reach
(Estimated) (Observed)
.41***
(4.08)
.61***
(5.14)
.57***
(5.48)
.50***
(4.76)
.02
(.07)
1.36***
(3.58)
–.36
(–.85)
–1.44**
(–2.36)
.20
(.61)
.76**
(2.10)
–.62*
(–1.69)
–.86
(–1.56)
.09
(.23)
1.24***
(3.24)
–.34
(–.81)
–1.51**
(–2.43)
–.01
(–.03)
1.36***
(4.06)
–.34
(–.89)
–1.25**
(–2.26)
.37
(1.05)
1.34***
(3.97)
–.64*
(–1.69)
–1.59***
(–2.85)
–1.37***
(–2.58)
–2.05**
(–2.27)
1.84***
(2.62)
3.83***
(3.11)
–.87*
(–1.93)
–3.08***
(–3.91)
1.47**
(2.49)
4.18***
(3.62)
–1.34**
(–2.51)
–1.90**
(–2.06)
1.83**
(2.58)
3.63***
(2.89)
–1.35***
(–2.91)
–1.54*
(–1.91)
1.92***
(3.06)
3.21***
(2.93)
–1.49***
(–3.16)
–1.38*
(–1.74)
1.96***
(3.20)
3.66***
(3.50)
.37
(.48)
Yes
Yes
Yes
.06
(.10)
Yes
Yes
Yes
.33
(.43)
Yes
Yes
Yes
–.02
(–.02)
Yes
Yes
Yes
–.02
(–.02)
Yes
Yes
Yes
7.70***
(5.80)
751
.406
.353
5.75***
(5.36)
597
.438
.374
7.76***
(5.90)
751
.401
.348
5.51***
(4.91)
751
.431
.380
3.37***
(2.89)
751
.408
.356
*p < .1.
**p < .05.
***p < .01.
Notes: t-values are in parentheses. We use robust standard errors in all regressions.
12 / Journal of Marketing, January 2014
Usage
Early Growth
(Observed) (Estimated)
observed maximum number of daily active users as a third
alternative dependent variable. Again, the results confirmed
our initial findings and were very similar to our original
analysis (see the fourth column of Table 4)—which is not
surprising, considering the high correlation between the
number of daily active users and the number of installations.
Estimated growth. For some practical applications, not
only potential reach but also the time frame in which a
product realizes its potential can be important. We therefore
constructed a measure for speed of growth that enables us
to compare apps that differ in their estimated reach. We calculated the number of days the app needed to reach 90% of
its potential installations (the asymptote in the TSM) and
then divided this number by the app’s total estimated reach
such that the measure was comparable across apps with different reach. Other growth measures exist, but this one
seemed least arbitrary for our purposes. The results in the
last column of Table 4 again resemble the results from our
original model.
Summary and Implications
Summary of Results
The viral marketing success of products such as FarmVille
has attracted many firms to Facebook—not only game producers but also makers of primarily utilitarian products,
whose value for consumers results more from their products’ usefulness and less from the fun they offer. Marketing
managers from all industries wonder whether they can
replicate the success of FarmVille and similar products by
simply imitating their approach—that is, encouraging consumers to broadcast unsolicited viral marketing messages to
their Facebook friends and offering small incentives to convince the receiver to try and use the product. Thus far, the
marketing literature has not answered this question.
In our study, we investigate two important questions
regarding viral marketing on primarily fun- and entertainmentoriented platforms such as Facebook: Should primarily
utilitarian products rely on the same sharing mechanisms
for their viral marketing campaign as products that are not
utilitarian? If not, why is this the case, and how should viral
marketing for utilitarian products differ? From our study of
751 products, we present new answers.
In particular, we note that primarily utilitarian products
should not rely on the same sharing mechanisms as FarmVille and other fun-oriented products. The very mechanism
that made FarmVille so successful is a recipe for failure
when used in a different product context. Unsolicited and
incentivized broadcast messages from friends are the least
effective sharing mechanisms for primarily utilitarian products (Table 3).
The reason for this somewhat surprising finding can be
found in dual-process theory from social psychology: the
ELM (Petty and Cacioppo 1986). Consumers use Facebook
with the intention of having fun and being entertained
rather than doing something useful. Thus, viral marketing
messages for utilitarian products on Facebook do not correspond with their situational expectations or schema (Bartlett
1932). Consequently, consumers unconsciously devote
fewer mental resources to evaluating the actual message
content and instead rely more on heuristics, simple inferences, and social cues. Put simply, consumers generally do
not visit Facebook to learn about utilitarian products, so
they process viral marketing messages about such products
differently than, say, messages about games.
Implications for Marketing Theory
The current research contributes to literature on viral marketing in general and work on sharing mechanisms in particular. This study is the first to examine differences in viral
marketing strategies for low- versus high-utilitarian products. We find stark differences in the effectiveness of sharing
mechanism characteristics for different types of products,
which should motivate further research along similar lines
for other known determinants of viral marketing success,
such as seeding strategies and message content. Moreover,
the theory we applied specifically focuses on viral marketing for utilitarian products in fun-oriented settings such as
Facebook. Our findings should translate to other social networks with similar settings, such as YouTube; further
research should investigate whether a reversal of our results
might occur in more utilitarian social networks (e.g.,
LinkedIn), as schema theory and the ELM would predict.
We also contribute to the sparse literature stream on the
effectiveness of sharing mechanism characteristics for viral
marketing. Aral and Walker (2011) focus on the in-depth
(customer-level) analysis of one particular product and analyze two sharing mechanisms that differ on one dimension
(communication exclusivity). In contrast, we compare 751
products representing 22 categories that include highly successful blockbusters with millions of customers as well as
average and entirely unsuccessful products. With this variety,
we can test the generalizability of previous results and—
through our comparison of low- versus high-utilitarian
products—extend and refine previous findings (Figure 3).
We also analyze all four dimensions (sender–receiver relationship, communication exclusivity, expressed interest,
and message features) of viral marketing sharing mechanisms. To our knowledge, ours is the first study to combine
all four dimensions, which collectively characterize the
sharing mechanisms through which viral messages spread
between consumers.
Implications for Marketing Practice
Marketing managers in charge of viral marketing activities
face difficult decisions when designing campaigns for their
products. In line with previous research (Aral and Walker
2011), we offer strong empirical evidence that the success
of such campaigns hinges on choosing the right sharing
mechanism. A suitable sharing mechanism should encourage consumers to recommend a product to their peers;
indeed, the results of our study show that a well-chosen
sharing mechanism can increase app success (as measured
by the number of installations) by a factor of 19. Conversely,
a viral marketing campaign that uses an inappropriate sharing mechanism can severely limit a product’s success.
Viral Marketing for Utilitarian Products / 13
The results of our analysis of viral marketing campaigns
on Facebook for 751 products shows that for low-utilitarian
products (e.g., games, music services), the ideal sharing
mechanism uses incentives and suggests that customers recommend products to their friends rather than to strangers.
We also find that sharing mechanisms that rely on unsolicited viral marketing messages are as effective as sharing
mechanisms that use solicited messages (Figure 3). In other
words, viral marketing campaigns for low-utilitarian products on Facebook are well-advised to follow the best-practice
strategy of products such as FarmVille and encourage unsolicited and incentivized broadcast messages from friends for
spreading viral marketing messages (Figure 2).
Recommendations for viral marketing campaigns for
high-utilitarian products on Facebook, such as job search or
stock market applications, differ radically. Sharing mechanisms for such useful offerings should avoid using unsolicited messages or broadcast messages from friends. In
addition, although incentives associated with these products
are not harmful, they are ineffective in Facebook’s fun- and
entertainment-oriented setting (Figure 3). Incentives (e.g.,
one free month of premium membership) often come at a
cost to the firm, so marketing managers should probably
avoid them. Instead, viral marketing campaigns for primarily utilitarian products should rely on solicited viral marketing messages that customers can either direct at individual
friends or broadcast to strangers. On Facebook, for example, promoting a useful product in the “Likes” section on
consumers’ “About” pages4 will be far more effective than
simply copying FarmVille’s approach. In fact, not using any
viral marketing sharing mechanism is four times better for
these products than FarmVille-type, unsolicited, incentivized broadcast messages from friends (Table 3).
In summary, marketing managers must adapt their Facebook viral marketing strategy to their product. The same
sharing mechanism characteristics that helped products
such as FarmVille reach more than 100 million consumers
in less than 40 days can severely harm the chances of success of primarily useful products.
Limitations and Further Research
In our empirical study, we analyzed the viral marketing
campaigns of 751 products. To achieve this breadth, we had
to accept some limitations in our empirical setup. First, our
data did not come from a controlled experiment. We went to
great lengths to control for product differences along relevant dimensions, such as the category, quality, and network
effects; still, a small risk of unobserved heterogeneity persists for our study. Second, our observed app data did not
allow us to distinguish individual users and their interactions with the sharing mechanisms. Such data would be
4Users can limit the visibility of the “About” section to
strangers through more restrictive privacy settings.
14 / Journal of Marketing, January 2014
desirable and helpful for understanding individual (rather
than typical) consumers’ reactions.
In addition to addressing these data limitations, our
findings point to at least five worthwhile directions for further research. First, we focused on viral marketing on Facebook, a fun- and entertainment-oriented social media platform. Additional research could attempt to replicate our
findings on other fun-oriented platforms, such as YouTube,
as well as test our findings in more utilitarian social networks, such as LinkedIn. In the latter environment, schema
theory suggests that consumers will process viral marketing
messages about utilitarian products via the central route
rather than the peripheral route, which should reverse the
effects for H1–H3 but not for H4.
Second, we investigated the moderating effect of a primarily utilitarian product context on sharing mechanism
characteristics only. Understanding whether the effects of
other known determinants of viral marketing success, such
as content characteristics and seeding strategies, also differ
between low- and high-utilitarian settings could inform
both theoretical and managerial perspectives.
Third, in the Facebook environment, the basic versions
of all products are available free of charge to the consumer.
As a result, we could not investigate price effects (Fandrich,
Kübler, and Pauwels 2013; Lambrecht, Seim, and Skiera
2007), which might be an important area for future studies.
Fourth, it would be worthwhile to investigate whether
the findings from this study replicate in a word-of-mouth
context. The main difference between viral marketing and
word of mouth pertains to the origin of the marketing message. If consumers rather than firms create the content of a
marketing message, consumers’ reactions should be strongly
affected when they process information through the peripheral route, which implies that they rely strongly on social
cues (“I can trust another customer more than I can trust the
firm”) (Hinz, Schulze, and Takac 2014; Phelps et al. 2004).
Messages about primarily utilitarian products therefore
should benefit most from word-of-mouth messages on
Facebook. Yet we expect that the effectiveness of the
remaining sharing mechanism characteristics (solicited vs.
unsolicited messages, no incentives vs. incentives, and
direct messages from friends vs. broadcast messages from
friends or broadcast messages from strangers) is similar to
that within the viral marketing context.
Fifth, encouraging consumers to make recommendations to their contacts through sharing mechanisms might
raise ethical concerns. Viral marketing messages are not
always undesirable; for example, consumers tolerate unsolicited (or spam) messages if the promoted product fits well
with their expectations about the environment. However, we
cannot rule out the possibility that message receivers might
have difficulties distinguishing firm-generated viral marketing activities and original content from peers, at least in
some cases. Further research should continue to discuss the
ethical implications and limitations of (not clearly identifiable) viral marketing in more detail.
REFERENCES
Aral, Sinan and Dylan Walker (2011), “Creating Social Contagion
Through Viral Product Design: A Randomized Trial of Peer
Influence in Networks,” Management Science, 57 (9), 1623–39.
Aronson, Elliot, Timothy D. Wilson, and Robin M. Akert (2012),
Social Psychology, 8th ed. Upper Saddle River, NJ: Pearson.
Bampo, Mauro, Michael T. Ewing, Dineli R. Mather, David Stewart, and Mark Wallace (2008), “The Effects of the Social Structure of Digital Networks on Viral Marketing Performance,”
Information Systems Research, 19 (3), 273–90.
Bartlett, Frederic C. (1932), Remembering: A Study in Experimental
and Social Psychology. London: Cambridge University Press.
Batra, Rajeev and Olli T. Ahtola (1991), “Measuring the Hedonic
and Utilitarian Sources of Consumer Attitudes,” Marketing
Letters, 2 (2), 159–70.
Berger, Jonah and Katherine L. Milkman (2012), “What Makes
Online Content Viral?” Journal of Marketing Research, 49
(April), 192–205.
Bitner, Mary J. and Carl Obermiller (1985), “The Elaboration
Likelihood Model: Limitations and Extensions in Marketing,”
in Advances in Consumer Research, Vol. 12, Elizabeth C.
Hirschman and Morris B. Holbrook, eds. Provo, UT: Association for Consumer Research, 420–25.
Chitturi, Ravindra, Rajagopal Raghunathan, and Vijay Mahajan
(2008), “Delight by Design: The Role of Hedonic Versus Utilitarian Benefits,” Journal of Marketing, 72 (May), 48–63.
Chiu, Hung-Chang, Yi-Ching Hsieh, Ya-Hui Kao, and Monle Lee
(2007), “The Determinants of Email Receivers’ Disseminating
Behaviors on the Internet,” Journal of Advertising Research,
47 (4), 524–34.
De Bruyn, Arnaud and Gary L. Lilien (2008), “A Multi-Stage
Model of Word-of-Mouth Influence Through Viral Marketing,”
International Journal of Research in Marketing, 25 (3), 151–
63.
Dellarocas, Chrysanthos (2006), “Strategic Manipulation of Internet Opinion Forums: Implications for Consumers and Firms,”
Management Science, 52 (10), 1577–93.
Diamond, William D. and Stephanie M. Noble (2001), “Defensive
Responses to Charitable Direct Mail Solicitations,” Journal of
Interactive Marketing, 15 (3), 2–12.
Dijksterhuis, Ap, Pamela K. Smith, Rick B. van Baaren, and
Daniël H.J. Wigboldus (2005), “The Unconscious Consumer:
Effects of Environment on Consumer Behavior,” Journal of
Consumer Psychology, 15 (3), 193–202.
Dobele, Angela, Adam Lindgreen, Michael Beverland, Joëlle
Vanhamme, and Robert van Wijk (2007), “Why Pass on Viral
Messages? Because They Connect Emotionally,” Business
Horizons, 50 (4), 291–304.
Drolet, Aimee, Patti Williams, and Loraine Lau-Gesk (2007),
“Age-Related Differences in Responses to Affective vs. Rational Ads for Hedonic vs. Utilitarian Products,” Marketing Letters, 18 (4), 211–21.
Fandrich, Thomas, Raoul V. Kübler, and Koen Pauwels (2013),
“Apps Appeal: Where in the World Are Consumers Sensitive to
Price, Ratings and Product Characteristics?” working paper,
Kühne Logistics University.
Godes, David and Dina Mayzlin (2009), “Firm-Created Word-ofMouth Communication: Evidence from a Field Test,” Marketing Science, 28 (4), 721–39.
Granovetter, Mark S. (1973), “The Strength of Weak Ties,” American Journal of Sociology, 78 (6), 1360–80.
Gupta, Sunil, Donald R. Lehmann, and Jennifer A. Stuart (2004),
“Valuing Customers,” Journal of Marketing Research, 41
(February), 7–18.
Hinz, Oliver, Christian Schulze, and Carsten Takac (2014), “New
Product Adoption in Social Networks: Why Direction Matters,”
Journal of Business Research, 67 (1), 2836–44.
———, Bernd Skiera, Christian Barrot, and Jan U. Becker (2011),
“Seeding Strategies for Viral Marketing: An Empirical Comparison,” Journal of Marketing, 75 (November), 55–71.
Hirschman, Elizabeth C. and Morris B. Holbrook (1982), “Hedonic Consumption: Emerging Concepts, Methods and Propositions,” Journal of Marketing, 46 (Summer), 92–101.
Hoffman, Donna L. and Thomas P. Novak (2013), “Online Experience in Social Media: Two Paths to Feeling Close and Connected,” working paper, (accessed October 10, 2013), [available at http://dx.doi.org/10.2139/ssrn.1990005].
Joshi, Amit M. and Michael Trusov (2012), “Are You a ‘Viral
Star’? Conceptualizing and Modeling Inter Media Virality,”
working paper, Robert H. Smith School of Business, University of Maryland.
Kalyanam, Kirthi, Shelby McIntyre, and J. Todd Masonis (2007),
“Adaptive Experimentation in Interactive Marketing: The Case
of Viral Marketing at Plaxo,” Journal of Interactive Marketing,
21 (3), 72–85.
Katz, Michael L. and Carl Shapiro (1985), “Network Externalities,
Competition, and Compatibility,” American Economic Review,
75 (3), 424– 40.
Kim, Namwoon, Vijay Mahajan, and Rajendra K. Srivastava
(1995), “Determining the Going Market Value of a Business in
an Emerging Information Technology Industry: The Case of
the Cellular Communication Industry,” Technological Forecasting and Social Change, 49 (3), 257–79.
Koroleva, Ksenia, Hanna Krasnova, and Oliver Günther (2011),
“Cognition or Affect? Exploring Information Processing on
Facebook,” in Social Informatics: Third International Conference SocInfo 2011, Anwitaman Datta, Stuart Shulman, Baihua
Zheng, Shou-De Lin, Aixin Sun, and Ee-Peng Lim, eds. Berlin:
Springer, 171–83.
Lambrecht, Anja, Katja Seim, and Bernd Skiera (2007), “Does
Uncertainty Matter? Consumer Behavior Under Three-Part
Tariffs,” Marketing Science, 26 (5), 698–710.
Lee, Janghyuk, Jong-Ho Lee, and Dongwon Lee (2009), “Impacts
of Tie Characteristics on Online Viral Diffusion,” Communications of the Association for Information Systems, 24 (32), 545–
56.
Leskovec, Jure, Lada A. Adamic, and Bernardo A. Huberman
(2007), “The Dynamics of Viral Marketing,” ACM Transactions on the Web, 1 (1), 1–39.
Lovett, Mitchell J., Renana Peres, and Ron Shachar (2013), “On
Brands and Word of Mouth,” Journal of Marketing Research,
50 (August), 427–44.
Petty, Richard E. and Pablo Briñol (2012), “The Elaboration Likelihood Model,” in Handbook of Theories of Social Psychology,
Paul A.M. Van Lange, Arie W. Kruglanski, and E. Tory Higgins, eds. London: Sage Publications, 224–45.
——— and John T. Cacioppo (1986), Communication and Persuasion: Central and Peripheral Routes to Attitude Change. New
York: Springer.
Phelps, Joseph E., Regina Lewis, Lynne Mobilio, David Perry, and
Niranjan Raman (2004), “Viral Marketing or Electronic Wordof-Mouth Advertising: Examining Consumer Responses and
Viral Marketing for Utilitarian Products / 15
Motivations to Pass Along Email,” Journal of Advertising
Research, 44 (4), 333–48.
Pöyry, Essi, Petri Parvinen, and Tuuli Malmivaara (2013), “Can
We Get from Liking to Buying? Behavioral Differences in
Hedonic and Utilitarian Facebook Usage,” Electronic Commerce Research and Applications, 12 (4), 224–35.
Roskos-Ewoldsen, David R., Beverly Roskos-Ewoldsen, and
Francesca R. Dillman Carpentier (2009), “Media Priming: An
Updated Synthesis,” in Media Effects: Advances in Theory and
Research, 3d ed., Jennings Bryant and Mary Beth Oliver, eds.
New York: Routledge, 74–93.
Rucker, Derek D., Richard E. Petty, and Joseph R. Priester (2007),
“Understanding Advertising Effectiveness from a Psychological Perspective: The Importance of Attitudes and Attitude
Strength,” in The Sage Handbook of Advertising, Gerald J. Tellis and Tim Ambler, eds. Thousand Oaks, CA: Sage Publications, 73–88.
Schroeder, Stan (2011), “‘CityVille’ Reaches 100 Million Monthly
Active Users,” Mashable.com, (January 13), [available at
http://mashable.com/2011/01/13/cityville-100-million-users/].
Schulze, Christian, Bernd Skiera, and Thorsten Wiesel (2012),
“Linking Customer and Financial Metrics to Shareholder
Value: The Leverage Effect in Customer-Based Valuation,”
Journal of Marketing, 76 (March), 17–32.
Appendix A: Details on the
Empirical Study of 751 Facebook
Apps
Estimating Product Reach
In our empirical study, we analyzed the effect of sharing
mechanism characteristics on 751 Facebook apps. Although
we observed the daily number of installations over the
course of an entire year, we faced censored data about app
success, because some apps existed before the first day of
our observation period and others were launched near the
end of the period such that they remained in their growth
phase. To achieve performance comparability across products, we estimated a diffusion curve for each app using the
TSM (Gupta, Lehmann, and Stuart 2004; Kim, Mahajan,
and Srivastava 1995; Schulze, Skiera, and Wiesel 2012), as
Equation A1 shows. We then used the estimated maximum
reach for each app (denoted by in Equation A1) as the primary success measure in our analyses.
(A1)
Reach i =
α
.
1 + exp ( −β − γ × t )
For our empirical study, we collected information on
863 apps in total. However, information on the daily
number of installations was insufficient for 112 apps. For
these apps, the nonlinear least squares estimation of the
TSM did not converge, a common occurrence when Sshaped diffusion curves have not yet reached the inflection
point. Consequently, we focus our analysis on the 751 apps
with sufficient information (87% of the original sample).
16 / Journal of Marketing, January 2014
Spangenberg, Eric R., Kevin E. Voss, and Ayn E. Crowley (1997),
“Measuring the Hedonic and Utilitarian Dimensions of Attitude: A Generally Applicable Scale,” in Advances in Consumer
Research, Vol. 24, Merrie Brucks and Deborah J. MacInnis,
eds. Provo, UT : Association for Consumer Research, 235–41.
Teixeira, Thales, Michel Wedel, and Rik Pieters (2012), “EmotionInduced Engagement in Internet Video Advertisements,” Journal of Marketing Research, 49 (April), 144–59.
Toubia, Olivier, Andrew T. Stephen, and Aliza Freud (2009),
“Viral Marketing: A Large-Scale Field Experiment,” INSEAD
Working Paper No. 2009/48/MKT.
Tucker, Catherine (2011), “Ad Virality and Ad Persuasiveness,”
NET Institute Working Paper No. 11-06.
Van der Lans, Ralf, Gerrit van Bruggen, Jehoshua Eliashberg, and
Berend Wierenga (2010), “A Viral Branching Model for Predicting the Spread of Electronic Word-of-Mouth,” Marketing
Science, 29 (2), 348–65.
Voss, Kevin E., Eric R. Spangenberg, and Bianca Grohmann
(2003), “Measuring the Hedonic and Utilitarian Dimensions of
Consumer Attitude,” Journal of Marketing Research, 40
(August), 310–20.
Wheeler, S. Christian, Richard E. Petty, and George Y. Bizer
(2005), “Self-Schema Matching and Attitude Change: Situational and Dispositional Determinants of Message Elaboration,” Journal of Consumer Research, 31 (4), 787–97.
Estimation Details
Table A1 provides correlations between all variables. Overall, the correlations are low; only unsolicited message and
direct message from a friend show few correlations with
other variables greater than .5, but never greater than .75.
Thus, we do not expect correlations to affect our results.
Space constraints precluded us from presenting coefficients
for all control variables in Table 3. For full estimation
results, see Table A2.
Appendix B: Details on In-Depth
Interviews with Industry
Professionals
For our study, we conducted in-depth interviews with 23
industry professionals about the effect of sharing mechanism
characteristics on Facebook apps. Among our interview
partners were marketing managers directly responsible for
firms’ Facebook activities, app programmers, and agency
representatives whose main expertise was designing and
programming Facebook apps. We conducted 17 interviews
individually and in person during two major gatherings of
industry professionals and 6 interviews over Skype or telephone. None of the respondents were aware of our study
before the interview. We informed respondents that their
answers would be used to “improve an ongoing research
study about Facebook apps,” guaranteed them anonymity,
and offered them a management summary of the empirical
findings as compensation for their efforts. We report their
answers as related to this research in Table B1. In Question
5, instead of differentiating more and less utilitarian apps,
we used the terms “business-related apps” and “leisurerelated apps,” which are conceptually similar.
Viral Marketing for Utilitarian Products / 17
1.00
.74
.34
.51
.31
–.15
.20
–.06
.31
.32
.28
.03
.03
.26
.22
.27
.04
–.08
.18
.10
.13
.16
.26
.25
.15
2
1.00
.36
1
Notes: All values greater than |.07| are significant at p < .05.
1. Ln(Reach)
2. Number of sharing
mechanisms
3. Unsolicited message
4. Message with incentive
5. Direct message from friend
6. Broadcast message from
stranger
7. Utilitarian
8. Network externalities
9. Network externalities
without base effect
10. Forum
11. Comments
12. Interaction
13. Reviews
14. Fans
Variable
.17
.18
.18
.04
.01
–.21
.12
–.08
1.00
.60
.66
.64
3
–.05
–.07
.00
.07
.03
–.18
–.08
–.09
1.00
.51
.45
4
.16
.20
.25
.08
.03
–.13
.18
.02
1.00
.08
5
.03
–.10
.02
–.06
–.01
–.16
–.08
–.04
1.00
6
–.07
.00
–.16
–.06
–.01
1.00
–.04
.00
7
.44
.47
.49
.03
.00
1.00
.07
8
TABLE A1
Correlation Table: Data on 751 Facebook Apps
.00
–.01
.11
.00
.09
1.00
9
1.00
.50
.45
–.03
.00
10
1.00
.46
.08
–.03
11
1.00
.12
.05
12
1.00
.14
13
1.00
14
TABLE A2
Full Results: Effects of Sharing Mechanism Characteristics on Product Reach
1
Number of sharing mechanisms
Sharing Mechanism Characteristics
Unsolicited message
Message with incentive
Direct message from friend
Broadcast message from stranger
Moderator for Sharing Mechanism Characteristics
in Primarily Utilitarian Apps
Unsolicited message utilitarian
Message with incentive utilitarian
Direct message from friend utilitarian
Broadcast message from stranger utilitarian
Control for App Type
Utilitarian
Controls for App Categories
Business
Chat
Classified
Dating
Education
Events
Fashion
File sharing
Food and drink
Gaming
Just for fun
Messaging
Mobile
Money
Music
Photo
Politics
Sports
Travel
Utility
Video
Controls for Network Effects
Network externalities
Network externalities without base effect
Forum
Comments
Interaction
Network externalities utilitarian
Network externalities without base effect utilitarian
Forum utilitarian
Comments utilitarian
Interaction utilitarian
Controls for App Quality
Reviews
Fans
Reviews utilitarian
Fans utilitarian
Reviews unsolicited message
Reviews message with incentive
Reviews direct message from friend
Reviews broadcast message from stranger
Fans unsolicited message
Fans message with incentive
Fans direct message from friend
Fans broadcast message from stranger
Reviews utilitarian unsolicited message
Reviews utilitarian message with incentive
18 / Journal of Marketing, January 2014
.69***
2
3
4
5
6
.75***
.77***
.82***
.73***
.63***
–.51
1.14***
.59
–.54
–.28
1.49***
.31
–1.18*
–.09
1.09***
–.03
–1.46**
–.02
1.38***
–.39
–1.34**
.02
1.36***
–.36
–1.44**
–.90*
–2.37***
1.28
3.30***
–1.21**
–1.84*
1.33*
3.55***
–1.00*
–2.00**
1.23*
3.12***
–1.37***
–2.05**
1.84***
3.83***
.95
.37
–1.28**
.40
–1.92**
–.12
–.12
–1.30
.35
–1.24
–.55
.62
.38
–.07
–1.55*
–1.36**
–.65
–.56
–1.30**
–2.10**
–.30
–.54
–2.28***
–1.17**
–.01
–2.01***
–.48
–.38
–1.57**
.19
–1.42*
–.78
–.50
–.09
–.42
–1.85**
–2.21***
–.95
–.76
–1.65***
–1.92**
–.68
–.52
–2.25***
.25
2.52***
.30
.66
.74
.62
–1.69
.20
–1.00
1.14
.29
2.50**
.29
.50
.57
.43
–.69
–.32
–.98*
.35
.16
1.03
–1.22**
.89
–1.97**
–.26
–.29
–1.45*
.27
–1.09
–.62
.92
.28
.11
–1.70**
–1.31*
–.92
–.55
–1.20**
–1.33
–.33
–.36
–2.17***
.32***
.00
3.02**
.04
.04
–.22**
–.07
2.53**
.00
.00
.01
.00
–1.13*
3.05*
TABLE A2
Continued
1
Controls for App Quality (Continued)
Reviews utilitarian direct message from friend
Reviews utilitarian broadcast message from stranger
Fans utilitarian unsolicited message
Fans utilitarian message with incentive
Fans utilitarian direct message from friend
Fans utilitarian broadcast message from stranger
Constant
Constant
6.64***
N
751
R-square
.129
Adjusted R-square
.128
2
3
4
5
6
–.26
11.70
.19***
–.12
–.24***
–.24***
6.58***
751
.160
.155
6.43***
751
.175
.164
7.10***
751
.272
.240
6.97***
751
.318
.279
7.70***
751
.406
.353
*p < .1.
**p < .05.
***p < .01.
Notes: We use robust standard errors in all regressions.
TABLE B1
Survey Results: 23 Industry Professionals on Sharing Mechanisms in Facebook Apps
Survey Questions
Average
Score
How much do you agree with the following statements?
1. Sharing mechanisms in Facebook apps (e.g., wall
8.2
posts, direct messages) are a major driver of app
performance.
2. I would expect the effect of different sharing
8.2
mechanisms on app success to vary strongly.
3. I am fully aware of which sharing mechanisms
4.4
work and which do not.
3a. In particular, I expect...
4. There are certain sharing mechanisms I would
only use for apps that have a very small chance of
success.
4a. In particular, that means...
5. I would expect sharing mechanisms for businessrelated apps to work very differently from leisurerelated apps.
5a. In particular, I expect...
SD
Respondent Comments
1.3
1.5
2.5
•Nothing specific.
•Posts on my own wall will be more effective than
direct invitations.
•Direct invitations will be more effective than
posts on my own wall.
•Wall posts from strangers are better than wall
posts from friends.
•Incentives do not work.
2.7
2.4
7.7
2.8
The use of (expensive) incentives is a “last resort.”
•Nothing specific.
•Wall posts are better for leisure, whereas direct
messages are better for business.
•Wall posts are better for business, whereas
direct messages are better for leisure.
Notes: All items measured on a ten-point scale (1 = “do not agree,” and 10 = “fully agree”).
Viral Marketing for Utilitarian Products / 19
Copyright of Journal of Marketing is the property of American Marketing Association and its
content may not be copied or emailed to multiple sites or posted to a listserv without the
copyright holder's express written permission. However, users may print, download, or email
articles for individual use.
Prompts and Questions:
What did this paper hypothesize? Did the authors find support for their predictions? Explain.
Describe the experimental study and how it was intended to answer the authors' research question.
Is this paper about Farmville? Why or why not?
What is the Farmville approach to viral marketing?
Thinking back to the other papers we have read this semester, how does(n't) this paper fit in to the idea of
entrepreneurial marketing?
What is viral marketing?
What are other confusions, questions, or comments you have about this paper?
Purchase answer to see full
attachment