the communication network. Launched in 1996, ICQ had 12 million
users by 1997; its mother company Mirabilis was bought for
$287 million by AOL a year later.
The viral marketing concept and these examples suggest that
marketers can leverage the power of interpersonal networks to
promote a product or service. The concept assumes that electronic,
peer-to-peer communications are an effective means to transform
(electronic) communication networks into influence networks, capturing
recipients' attention, triggering interest, and eventually leading
to adoption or sales. Yet it is difficult to identify substantial evidence to
support these assumptions or to explain why and how viral marketing
works, which is perhaps why it is currently viewed as more of an art
than a science (Diorio, 2001).
E-mail seems here to stay, and there is no doubt that peer-to-peer,
e-mail-based communications will continue to play an informational
and influential role on recipients' behavior. The proliferation of spam
(i.e., unsolicited bulk e-mails) and e-mail-based electronic viruses
have made recipients suspicious of most unsolicited e-mails. Consumers
experience a high level of noise in their day-to-day electronic
communications and for viral marketing campaigns to be designed
more effectively, there is a need to better understand which online
referrals are likely to cut through the clutter and which are not.
To better understand why and how viral marketing can be
effective, we must understand its pass-along process and its underlying
mechanisms of influence. While the existing word-of-mouth
(hereafter WOM) literature can inform us, it is important to note that
electronic referrals differ from their “offline” counterparts in two
significant ways:
1. They are electronic by nature; there is no face-to-face communication.
2. Those referrals are usually unsolicited, that is, they are sent to
recipients who are not looking for information, and hence are not
necessarily willing to pay attention to them.
For reasons we review later, and despite an abundant literature,
little attention has been given to unsolicitedWOMcommunications. In
addition, researchers who have addressed WOM communications
have usually been limited in their ability to collect complete, detailed,
and accurate information. Largely as a consequence of the lack of such
data, the mechanisms by which WOM communications influence
behaviors are not well understood. We concur with Bansal and Voyer
(2000) that “…there is surprisingly little empirical research that
examines [WOM] ‘procedural’ aspects” (p.166). And the advent of the
viral marketing phenomenon underscores the importance of developing
both methods to study and generate substantive findings about
this phenomenon. Hence, our goals in this paper are three-fold:
1. To introduce multi-stage decision-making models as mechanisms
to study and refine our understanding of unsolicited, electronic
referrals.
2. To describe a research methodology we used to inform the model
using data collected unobtrusively and in real time.
3. To test the above model and methodology with a field study,1 and
to compare our results with those from traditional one-stage
models.
This paper is structured as follows. In the next section,we present a
brief overview of the WOM literature and examine why so little is
known about how unsolicited WOM communications influence
consumer decisions. We then cast the existing WOM literature into
a multi-stage framework to help decompose and predict the level,
antecedents, and moderating effects ofWOMinfluences on each stageof recipients' decision-making processes. Next we introduce a
research methodology to study the influence of WOM referrals at
each stage of the decision process, and present the results of a field
study in which we tracked the influence of one type of viral
marketing: e-mail-based, unsolicited peer-to-peer referrals. We find
that the antecedents of WOM influence (e.g., tie strength, demographic
similarity) vary significantly and predictably across stages,
thus enriching our understanding of the mechanisms of influence, and
demonstrating the value of this new methodology for future research.
We conclude with discussions of the theoretical, methodological, and
managerial implications of this work.