33
Estimating the Dynamic Effects of Online Word-Of-Mouth
2
the Superbowl ads of Careerbuilder.com and GoDaddy.com), it is of broad interest to
understand the relative effectiveness of word-of-mouth versus other marketing
g
communication efforts. One of the fastest growing arenas of the World Wide Web is the
space of so-called social networking sites (e.g., Friendster, Facebook, Xanga). These sites
rely upon user-generated content to attract and retain visitors and obtain revenue
primarily from the sale of online advertising. They also accumulate user information that
may be valuable for targeted marketing purposes. The social network setting offers an
attractive context to study word-of-mouth, as the sites provide easy-to-use tools for
s for
current users to invite others to join the network. They are also capable to record these
activities.
Internet companies commonly employ several types of WOM marketing
activities. These include (1) viral marketing – creating entertaining or informative
messages designed to be passed on by each message receiver, analogous to the spread of
an epidemic, often electronically or by email; (2) referral programs – creating tools that
at
enable satisfied customers to refer their friends; and (3) community marketing – forming
g
or supporting niche communities that are likely to share interests about the brand (such as
user groups, fan clubs, and discussion forums) and providing tools, content, and
information to support those communities.
2
In this paper, we examine one specific form of WOM activity: electronic referrals.
Our objective is to estimate the elasticity, both short and long-run, of word-of-mouth
outh
referral activity at an Internet social networking site. We compare these elasticity
estimates with those obtained for media appearances (public relations) and event
2
A detailed overview of different forms of WOM marketing is available at the Word of Mouth Marketing
Association web site (www.womma.org).
.NET PDF SDK | Read & Processing PDF files by this .NET Imaging PDF Reader Add-on. Include extraction of text, hyperlinks, bookmarks and metadata; Annotate and redact in PDF documents; Fully support all
check links in pdf; add hyperlinks to pdf online PDF Image Viewer| What is PDF advanced capabilities, such as text extraction, hyperlinks, bookmarks and Note: PDF processing and conversion is excluded in NET Imaging SDK, you may add it on
add hyperlink in pdf; active links in pdf
25
Estimating the Dynamic Effects of Online Word-Of-Mouth
3
marketing – the main company-sponsored marketing activity. An important aspect of our
ur
approach is to recognize the potential endogeneity in customer acquisition, WOM
activity, and other marketing communication efforts. WOM may be endogenous because
it not only influences new customer acquisition but is itself affected by the number of
new customers. Likewise, traditional marketing activities may stimulate WOM; they
should be credited for this indirect effect as well as the direct effect they may have on
customer acquisition. We empirically test for this endogeneity using Granger causality
tests. We then develop a Vector Autoregression (VAR) model to handle endogeneity
problem. We link variation in the number of newly acquired customers (signups) with the
number of invitations (referrals) sent by existing members of the network to their friends
outside the network. The proposed model allows us to measure the short and long-run
effects of WOM and to compare the effects of WOM with those of other marketing
communications.
Our empirical results from the social networking site show that WOM referrals
strongly affect new customer acquisition. We estimate a long-run elasticity of 0.53. This
is approximately 2.5 times higher than the average advertising elasticity reported in the
literature (Hanssens et al 2001). For the company under study, WOM has a much
stronger impact on new customer acquisition than traditional forms of marketing. In
particular, WOM elasticity is about 20 times higher than the elasticity for marketing
events (0.53 vs. 0.026). We translate these findings into economic implications by
calculating how much the average acquired customer contributes to firm revenues. This
computation provides an upper limit to the financial incentives the firm could offer
27
Estimating the Dynamic Effects of Online Word-Of-Mouth
4
existing customers to stimulate word-of-mouth (a practice also growing rapidly in offline
ne
use).
Research Background
The earliest study on the effectiveness of WOM is survey-based (Katz and
Lazarsfeld 1955). The authors found that WOM was seven times more effective than
print advertising in influencing consumers to switch brands. Since the 1960s, word of
mouth has been the subject of more than 70 marketing studies (Money et al 1998).
Researchers have examined the conditions under which consumers are likely to rely on
others’ opinions to make a purchase decision, the motivations for different people to
spread-the-word about a product, and the variation in strength of influence people have
e
on their peers in WOM communications. Consumer influence over other consumers has
been demonstrated in scholarly research concerning social and communication networks,
opinion leadership, source credibility, uses and gratifications, and diffusion of
innovations (Phelps et al 2004).
Until recently WOM research relied on experimental methods versus studying
actual consumer actions in the marketplace. A major challenge in studying actual WOM
is obtaining accurate data on interpersonal communications. Studying WOM on the
Internet can help address this problem by offering an easy way to track online
interactions. The Internet, of course, gives only a partial view of interpersonal
communication; WOM exchange is not limited to the online world. Nevertheless, for
some products or product categories, Internet measures of WOM could be a good proxy
for overall WOM. We believe that for online communities, the electronic form of
f
26
Estimating the Dynamic Effects of Online Word-Of-Mouth
5
“spreading the word” is the most natural one. Thus, we suggest that online WOM should
be a good proxy for overall WOM in the Internet social network setting of our study.
Recent research has begun to study WOM in an Internet setting. De Bruyn and
Lilien (2004) observed the reactions of 1,100 recipients after they received an unsolicited
email invitation from one of their acquaintances to participate in a survey. They found
that the characteristics of the social tie influenced recipients’ behaviors but had varied
effects at different stages of the decision-making process: tie strength exclusively
facilitated awareness, perceptual affinity triggered recipients’ interest, and demographic
similarity had a negative influence on each stage of the decision-making process. Godes
and Mayzlin (2004) suggest that online conversations (e.g., Usenet posts) could offer an
easy and cost-effective opportunity to measure word of mouth. In an application to new
television shows, they linked the volume and dispersion of conversations across different
Usenet groups to offline show ratings. Chevalier and Mayzlin (2006) used book reviews
posted by customers at Amazon.com and BarnesandNoble.com online stores as a proxy
for WOM. The authors found that while most reviews were positive, an improvement in a
book’s reviews led to an increase in relative sales at that site and the impact of a negative
review was greater than the impact of a positive one. In contrast, Liu (2006) shows that
both negative and positive WOM increase performance (box office revenue). Finally,
Villanueva, Yoo and Hanssens (2006) compared customer lifetime value (CLV) for
customers acquired through WOM vs. traditional channels. In an application to a web
hosting company, the authors showed that marketing-induced customers add more short-
-
term value to the firm, but word-of-mouth customers added nearly twice as much long-
-
25
Estimating the Dynamic Effects of Online Word-Of-Mouth
6
term value. However, the authors do not observe the marketing inputs and thus can not
directly estimate the response of customer acquisition to WOM and to traditional efforts.
Our paper differs from above studies in research question and application. First,
we aim to directly compare the dynamic performance effects of word-of-mouth referrals
s
with that of traditional marketing efforts and quantify the economic value of each WOM
referral to the firm. Second, our empirical application is to an Internet social networking
site, a novel setting for a marketing study.
Internet social networking sites
While still a relatively new Internet phenomenon, online social networking has
already attracted attention from major industry payers. Microsoft, Google, Yahoo! and
AOL are among companies offering online community services. According to Wikipedia
(www.wikipedia.org), at present there are about 30 social networking web sites each with
more than one million registered users and several dozen significant, though smaller,
sites. In terms of web traffic, as of March 2006, ComScore MediaMetrix reports that the
largest online social networking site was MySpace.com with 42 million unique visitors
per month, followed by FaceBook.com with 13 million and Xanga.com with 7.4 million
unique visitors. ComScore MediaMetrix numbers suggest that every second Internet user
in the U.S. visits one of the top 15 social networking sites (Table 1).
[Table 1. Social Networking Sites Ranking]
A social networking site is typically initiated by a small group of founders who
send out invitations to join the site to the members of their own personal networks. In
turn, new members send invitations to their networks, and so on. Hence, invitations (i.e.
24
Estimating the Dynamic Effects of Online Word-Of-Mouth
7
WOM referrals) have been the foremost driving force for sites to acquire new members.
Typical social networking sites allow a user to build and maintain a network of friends
for social or professional interaction. In the core of a social networking site are
personalized user profiles. Individual profiles are usually a combination of users’ images
(or avatars), list of interests, music, books, movies preferences, and links to affiliated
profiles (“friends”). Different sites impose different levels of privacy in terms of what
information is revealed through profile pages to non-affiliated visitors and how far
“strangers” vs. “friends” can traverse through the network of a profile’s friends. Profile
holders acquire new “friends” by browsing and searching through the site and sending
requests to be added as a friend. Other forms of relation formation also exist.
In contrast to other Internet businesses, online communities rely upon user-
generated content to retain users. A community member has a direct benefit from
bringing in more “friends” (e.g., through participating in the referral program), as each
new member creates new content, which is likely to be of value to the inviting (referring)
party. Typically, sites facilitate referrals by offering users a convenient interface for
sending invitations to non-members to join the community. Figure 1 shows how two
popular social networking sites, Friendster.com and Tribe.com, implement the referral
process.
[Figure 1. Referrals Process at Friendster.com and Tribe.com]
Referrals made through the site’s provided interface are easily tracked. Some sites offer
incentives to make a referral. For example, Netflix.com recently offered its existing
customers to pass a “gift” of a month of free service to their non-member acquaintances.
61
Estimating the Dynamic Effects of Online Word-Of-Mouth
8
Many subscription-based services offer progressive discounts on monthly fees for each
referral made.
While the mechanics of social network formation through the WOM referrals
process may be straightforward, little is known about the dynamics and sustainability of
this process. Also, as social networking sites mature, they may begin to use traditional
marketing tools. Management therefore may start to question the relative effectiveness of
WOM at this stage. Our objective is to contribute a new set of empirical findings to this
topic.
Modeling Approach
A typical social networking site has several ways to attract new customers,
including event marketing (directly paid for by the company), media appearances
(induced by PR) and word-of-mouth (WOM) referrals. How should we model the
e
effectiveness of these communication mechanisms? As a base model, we may regress
signups on events, media and WOM, controlling for deterministic components such as a
base level (constant), a deterministic (time) trend, seasonality and lags of the dependent
variable (Box and Jenkins 1970). The time trend is intended to capture external factors,
including growth in Internet access, growth in people with high-speed bandwidth, general
increases in content and interest in social networking sites. Seasonal patterns may be high
(e.g. day-of-week) frequency, as most Internet use occurs during weekdays (Pauwels and
d
Dans 2001) and low frequency, e.g. yearly holiday periods. Equation (1) specifies our
base this model:
6
1
1
J
t
t
t
t
i
t
t j
t
i
j
Y
X
M
E
C T
d
H
Y
!
"
=
=
=
+
+
+ + +
+
+
+
#
#
(1)
Documents you may be interested
Documents you may be interested
