You are on page 1of 15

Electronic Commerce Research and Applications 13 (2014) 205–219

Contents lists available at ScienceDirect

Electronic Commerce Research and Applications


journal homepage: www.elsevier.com/locate/ecra

Gift or threat? An examination of voice of the customer: The case


of MyStarbucksIdea.com
Hanjun Lee a, JinYoung Han b,⇑, Yongmoo Suh a
a
Korea University Business School, 5 Ga Anamdong, Sungbookgu, Seoul, South Korea
b
Chung-Ang University, School of Computer Science and Engineering, 84 Heukseokro, Dongjakgu, Seoul, South Korea

a r t i c l e i n f o a b s t r a c t

Article history: In Web 2.0 environment, the influence of customers’ voices is increasing. Many companies have built
Received 11 March 2013 their online brand communities for listening to the diverse voices of the customer (VOC) and promoting
Received in revised form 28 January 2014 customer’s engagement. The information obtained from these brand communities is exploited for orga-
Accepted 11 February 2014
nizational innovation. However, the VOC with negative contents could possibly present threats to a firm
Available online 12 March 2014
in an online context. In this study, we develop a research model that includes the types of customer
expectations, opinion leader engagement, negative valence of VOC, interaction, and innovation. Subse-
Keywords:
quently, we empirically validate the effects of customer expectations on the interactions among commu-
Voice of the customer (VOC)
Expectancy Disconfirmation Theory
nity members and organizational innovation by using a dataset from MyStarbucksIdea.com. The results
Online brand community show that the effects of VOC on the interaction within a brand community and organizational innovation
Open innovation are different depending on the types of expectation in the VOC. Opinion leaders’ engagement positively
MyStarbucksIdea.com affects the interaction among community members. Moreover, the negative valence of VOC affects the
Zero-inflated negative binomial model relationships between customer expectations and interactions, and results in a possible threat within
an online brand community. The research results give some insight into managing the brand community
and analyzing the VOCs so as to achieve organizational innovation.
Ó 2014 Elsevier B.V. All rights reserved.

1. Introduction online communities with the purpose of listening to diverse VOC


and promoting customer engagement and loyalty (Gangi and
In 2005, a customer requested Dell to fix a product-related Wasko 2009).
problem. However, once Dell ignored the claim, the dissatisfied As VOC is considered as a resource for innovation, firms such as
customer posted his experience on a blog. His post spread exten- Proctor & Gamble (P&G), and Starbucks are exploiting their online
sively throughout the Internet, and consequently the reputation brand communities as conduits for collecting VOC (Gangi and
of Dell suffered to such an extent to give birth to the expression, Wasko 2009). The phenomenon of open innovation has recently
‘‘Dell sucks’’ (Bernoff and Li 2008). Dell became to learn how to enabled firms to achieve collective intelligence and to reduce the
listen to customer voices at a great cost. The Dell example shows effort of developing new products or services (Lichtenthaler
the importance of customer voice management and the influence 2011). Online brand communities encourage customers to share
of word-of-mouth (WOM) through the Internet. their ideas, discuss and evaluate other customer ideas, and develop
Traditionally, customer voices have been considered a critical ideas into better ones through interaction (Teece and Pisano 1994,
resource for firms. This is because voice of the customer (VOC) nat- Gangi and Wasko 2009). Thus, encouraging customers to interact
urally includes expectations related to customer needs and with other customers and extracting useful VOC, critical for brand
requirements, and these expectations can help firms improve their communities, can help a firm innovate.
products or services. As the habitual use of Web 2.0 emerges, more A generally acceptable view point is that VOC should be consid-
VOC is being generated and shared easily through various online ered as a gift to help firms develop new products or services that
communities. Accordingly, it is prevalent for firms to manage meets the needs of their customers (Barlow and Moller 2008,
Denove and Power 2007). On the other hand, VOC is also perceived
as a possible threat as we saw in the case of Dell. VOC with nega-
⇑ Corresponding author. Address: Bld. #208-526, 84 Heukseokro, Dongjakgu,
tive contents, whether it is true or not, can spread from customer
Seoul 156-756, South Korea. Tel.: +82 2 820 5935; fax: +82 02 823 5301.
E-mail addresses: hjlee1609@gmail.com (H. Lee), win1999@naver.com (J. Han),
to customer and give rise to negative WOM, which may result in
ymsuh@korea.ac.kr (Y. Suh). customers’ switching of brands and stores, and the loss of firm’s

http://dx.doi.org/10.1016/j.elerap.2014.02.001
1567-4223/Ó 2014 Elsevier B.V. All rights reserved.
206 H. Lee et al. / Electronic Commerce Research and Applications 13 (2014) 205–219

reputation and brand image (Richins 1983). As WOM effect may be outside an organization, and can also go to markets from inside
more powerful for an online community where members interact or outside an organization. An innovation process leads to a signif-
with each other easily and widely, VOC with negative contents icant amount of external knowledge exploration and exploitation
can be more risky in an online context. Therefore, deepening the (Chesbrough 2003b). Open innovation can be explained as the flow
understanding of the characteristics and influences of such VOC of ideas into and out of an organization. Prior research recom-
for online communities can be critical for firms. mends organizations to take an open strategy approach to comple-
Members within a community are affected by opinion leaders. menting internal innovation processes (Chesbrough 2003a,
Prior studies have investigated the roles and influences of opinion Laursen and Salter 2006).
leaders (Katz and Lazarsfeld 1955, Rogers 1962). Opinion leaders Enkel et al. (2009) divides the process of open innovation into
exist in online brand communities and affect community members three core processes: outside-in, inside-out and coupled. These three
as in other ordinary communities. Understanding the influence of processes are circulated. The coupled process, which is especially
opinion leadership on organizational innovation and the interac- crucial to combine the outside-in and inside-out processes (Enkel
tion of the community members can be meaningful when an on- et al. 2009) refers to co-creation with complementary partners.
line brand community is considered as an innovation platform. Ramaswamy (2009) also defines co-creation as the process by
We initiate our discussion on the influence of VOC in an online which products, services, and experiences are developed jointly
brand community. We suggest that the influence of each VOC on by firms, stakeholders, and customers. Derived from open-source
community promotion and organizational innovation can be dif- software development (von Hippel and von Krogh 2006), open
ferent depending on the type of VOC. The types are classified innovation has mainly emphasized the co-creation with customers
according to customer expectations which a VOC contains. We as- (Hienerth 2006, Lettl et al. 2006), lead users (Franke et al. 2006),
sume that VOC with negative contents can be a seed of threat that communities (Reichwald and Piller 2009), and partners from other
may result in negative WOM in an online community. Additionally, industries (Enkel and Gassmann 2010).
we expect that opinion leaders influence the interaction among on- In our study, we focus on the co-creation within a brand com-
line community members and organizational innovation. There- munity. Online communities as innovation platforms can be classi-
fore, three research questions are investigated in this study: fied by the purpose (e.g., understanding customers versus finding
solutions) and the operator (e.g., company versus third party)
 Does the type of the VOC expectation have different effect on (Stoetzel et al. 2011). This study selects an online brand commu-
the interaction among the community members and organiza- nity, which is organized to understand their customers and is oper-
tional innovation? ated by a company.
 How does VOC with negative contents affect the interactions
within an online brand community? 2.2. Online brand community
 What are opinion leader roles in an online brand community?
Brand community refers to a group of brand users with a struc-
These research questions can have practical importance for tured set of social relationships among the users (Muniz and
firms to achieve innovation and to manage online brand communi- O’guinn 2001). The users share enthusiasm for the brand and
ties. The data is collected from MyStarbucksIdea.com (MSI) which engage in group actions to accomplish collective goals (Bagozzi
was developed as an online brand community for customer-drive and Dholakia 2006). Customers within brand communities are
ideation where anyone can submit an opinion. This paper is orga- usually self-motivated and enthusiastic in creating what they truly
nized as follows. In Section 2, we review literature related to open want firms to achieve. They aggressively engage in group actions
innovation, VOC and online brand communities. Section 3 such as exchanging, sharing and debating ideas, and contribute
describes the research context. A research model and hypotheses to the improvement of products and services, which can help firms
are developed in Section 4. We describe the research methodology, innovate.
analysis, and results in Section 5. In Section 6, we discuss findings Moreover, the social relationships built within a brand commu-
of the study and conclude with study limitations and nity can create customer exit barriers and maintain the longevity
contributions. of customer loyalty (Algesheimer et al. 2004, McAlexander et al.
2002). Customers engaged within a brand community tend to be
loyal to the brand and self-motivated enough to transfer marketing
2. Literature review messages to other members (Fournier 1998, Muniz and O’guinn
2001). Customer stories shared in brand communities facilitate
2.1. Open Innovation member interactions and build a strong relationship between the
customers and the brand (Stokburger-Sauer 2010). Such interac-
Innovation can be defined as the creation of new ideas and tions can motivate the initial purchase of a product or service, or
knowledge to develop new business values and to meet customer instead, shared benefits can be accrued with time and experience
and market needs (Gloet and Terziovski 2004). Traditionally, firms in ways that profoundly influence customer loyalty (Schouten
innovated themselves using internal resources. Most firms have and McAlexander 1995). Accordingly, customer relationships and
practiced this closed approach as their innovation strategy. Thus, interactions are fundamental to the existence of a brand commu-
invention, development, and design are restricted to internal nity (McAlexander et al. 2006, Muniz and O’guinn 2001).
resources (Ahlstrom 2010, March 1991, Wyld 2010). Customer activities regarding innovation and interactions with-
However, in the past few decades, rapid change in the business in a community become more efficient and effective via online
environment did not allow the development of new ideas only social media. With the aid of computer-mediated communication,
from resources internal to the firm (Katz and Allen 2007). Firms online brand communities can provide venues where marketers
are acquiring external resources to complement their internal can obtain insight into customer needs, and where the relationship
knowledge since it is impossible for a single firm to obtain all between the brand and customers is established at negligible costs
related resources within that firm (Beamish and Lupton 2009, Cas- (Denove and Power 2007, Roman 2010). Online brand communities
siman and Veugelers 2006, von Hippel 1988). In light of this trend, allow participants to share information about a brand, and retain
Chesbrough (2003b) proposed a new concept of open innovation to the history and the culture of the brand without limitations of time
mean a paradigm where valuable ideas can come from inside or and distance (Füller et al. 2006, Iansiti and MacCormack 1997).
H. Lee et al. / Electronic Commerce Research and Applications 13 (2014) 205–219 207

Opinion leaders exist in online brand communities as in other tendency increases with customers who experience satisfaction
community types. Opinion leaders refers to a small part of a social or dissatisfaction from a gap between expectations and experi-
group who are active in an interpersonal communication network, ences of services or products provided by companies (Day 1980,
provide information, opinions, and suggestions, exert personal Landon 1980). Although no one theory of customer satisfaction
influences on others, and shoulder idea-guiding responsibility or dissatisfaction is universally accepted, the most widely applied
(Katz and Lazarsfeld 1955, Lazarsfeld et al. 1944). From a market- theory is the Expectancy Disconfirmation Theory. This theory has
ing perspective, opinion leaders are characterized by domain-spe- two prominent variables: expectation and perceived quality. These
cific innovativeness (Goldsmith 2001, Goldsmith and Hofacker variables are defined in two different time periods. Expectation is
1991), which refers to the tendency to learn about and adopt a related to the pre-purchase time, while perceived quality is related
new product within a specific product category (Goldsmith and to the post-purchase time. The gap between initial expectation and
Hofacker 1991). Innovativeness is positively correlated with prod- perceived quality is known as disconfirmation of expectation
uct involvement and product knowledge (Goldsmith 2001). By fre- (Oliver 1980, Patterson et al. 1996). When a customer perceives a
quently talking about product-related issues and seeking product higher or lower performance over the quality of a certain product
knowledge, opinion leaders can be more knowledgeable about a or service than customer expectation, positive or negative discon-
relevant product and show more involvement than other custom- firmation will occur and thus result in satisfaction or dissatisfac-
ers. Consequently, opinion leaders can provide information to fol- tion. In sum, expectation plays the role as a standard that
lowers and exercise great influence (Riecken and Yavas 1986, determines satisfaction or dissatisfaction, which may then
Venkatraman 1988). Moreover, opinion leaders are found to be influence customer behavior in online brand communities.
more influential on the opinions of individuals due to the credibil- Currently, an online brand community is vibrantly exploited as
ity they seem to have (Katz and Lazarsfeld 1955). Therefore, this a place for listening to VOC with various expectations. In this study,
study investigates the impact of opinion leaders on engagement we suggest five types of expectation collectively from prior studies.
in encouraging innovative ideas and the interaction among cus- These expectations are utilitarian, hedonic, ideal-based, product-
tomers within online brand communities. based, and brand-based (Chitturi et al. 2008, Magnusson 2009,
Miller 1977, Schindler 1998, Woodruff et al. 1983). Utilitarian
2.3. Voice of the customer refers to functional, instrumental, practical, and monetary related
to consumption offerings. Hedonic refers to aesthetic and enjoy-
VOC refers to a set of the needs and desires suggested by cus- ment-related expectation (Batra and Ahtola 1991, Chitturi et al.
tomers (Griffin and Hauser 1993) or a description of customer pref- 2008, Schindler 1998, Strahilevitz and Myers 1998, Dhar and
erences and aversions (Roman 2010). VOC contains diverse Wertenbroch 2000). Utilitarian and hedonic can cover any type of
contents such as customer benefits (e.g., free coupons and lowering expectation related to customer benefits (Magnusson 2009).
of prices), suggestions of ideas for new products or services, or Ideal-based expectation refers to customer needs in an ideal sense,
complaints about previous customer experiences. VOC includes which may not be associated with what a product provider prom-
valuable information for firms in many aspects. First, VOC provides ises (Miller 1977). Product-based expectation and brand-based
detailed understanding of customer perceptions on products and expectation arise from the comparison of the most recent purchase
services. This understanding provides key clues to identify cus- with a past purchase (Woodruff et al. 1983). Product-based expec-
tomer requirements that can be used to determine customer pref- tation rests on an experience related to a similar product or service
erences and needs (Griffin and Hauser 1993, Woodruff 1997). within the same brand. In contrast, brand-based expectation
Second, VOC can help to discover critical areas needed to evaluate comes from a similar product or service experience across different
such as the attributes of a product that a customer finds most brands (Woodruff et al. 1983).
important, and the attributes that need improvement (Manchu- In summary, we select an online brand community as an inno-
lenko 2001, Woodruff 1997). Third, VOC can suggest a strategic vation platform which is operated by a company in order to under-
direction for the progression of a firm by providing a shared com- stand their customer’s needs and wants. We then examine the
mon language (Hauser 1993, Kärkkäinen et al. 2001). VOC basically relationships between VOC characteristics, company’s innovation,
allows a firm to give customers what they really want most from and community’s interactions. We also investigate the influence
that firm, and prevents the annoyance and alienation that arises of opinion leaders and negative VOC on company’s innovation
when a firm provides a customer with what they do not want (Ro- and community’s interactions.
man 2010). Customers also need to be welcomed and encouraged
(Reynolds and Harris 2005) since they bring opportunities to en-
hance management effectiveness and ensure profitability through 3. Context domain
increased customer satisfaction (Huang et al. 1996). Thus, cus-
tomer voices should be viewed as a gift to the firm (Barlow 1996). The research context is MSI which is one of the most popular
Although VOC can be helpful, the consequences of VOC can be brand communities. Starbucks launched this community in March
harmful or even a threat to a firm. Those consequences include 2008, searching for a way to talk to its customers directly. As of
customer behavior such as switching brands or stores, filing a com- January 2013, more than 150,000 customer-generated comments
plaint to the seller, and disseminating negative WOM (Richins had been posted, which cover customers’ criticism of and prefer-
1983). Prior studies have shown that WOM has a significant impact ences for the ideas posted by other customers. Through this diverse
on customer decision-making (Arndt 1967, Katz and Lazarsfeld VOC, Starbucks gleaned what their customers are dissatisfied with
1955, Richins 1983). Similar to WOM, electronic WOM (eWOM) and expect from Starbucks.
can have higher credibility and relevance to customers than mar- The process of ideation in MSI can be explained as follows. Any
keter-created information (Bickart and Schindler 2001). Since a customer can register on the webpage to become a member. A
negative message has a stronger effect and can spread more Member can post whatever he/she wants to share as his/her own
quickly and widely than a positive message (Diener and Greyser opinion. Once an idea is posted, other members can vote on the idea
1978, Richins 1983), negative messages in an online community as either a promotion or a demotion of the idea. Each promotion or
should be handled attentively. demotion will result in a 10-point increase or decrease, respectively.
Generating VOC or telling others of their experiences related to Members also can comment on ideas to discuss in more detail. Opin-
past products or services is a natural response of customers. This ion leaders, who were selected as top contributors among ordinary
208 H. Lee et al. / Electronic Commerce Research and Applications 13 (2014) 205–219

members by Starbucks, usually join or nurture the discussion. There have been many arguments for and against the involve-
Through the discussion, the initial idea is evaluated, refined, and ment of customers in the innovation process (Magnusson 2009).
developed into a better idea. These interactions are important in Previous studies have shown evidences of success after customers
terms of promoting the community and developing VOC in more engage in the innovation process (e.g., Japanese mobile phone sys-
detail, which can lead to organizational innovation. tem I-mode and the development of equipment for outdoor sports)
Based on the votes and the discussion, some prospective ideas (Hienerth 2006, Lüthje 2004). In contrast, customers seem to have
go into the internal review process. Starbucks discloses the review a limited knowledge and awareness in the capability of the under-
process to the members allowing them to see the current status of lying technology associated with innovation and their contribution
each idea, whether it is under review, reviewed, coming soon, or to the innovation process has been doubtful (Christensen and
launched. Starbucks has implemented more than 200 ideas from Bower 1996, Martin 1995).
March 2008 to November 2011, amounting to more than one idea Previous research suggested numerous criteria to evaluate the
implemented every week since the community was launched. effectiveness of an idea for innovation (Balachandra and Friar
1997, Cooper 1993). Several researchers emphasized that the crite-
4. Hypothesis development ria should be chosen according to the context (Hart et al. 2003,
Hauser and Zettelmeyer 1997, Tzokas et al. 2004). However, cus-
The research model and hypotheses depicted in Fig. 1 include tomer benefits can be accepted as a criterion irrespective of busi-
the five types of VOC as independent variables, negative valence ness context (Magnusson 2009). Having examined the
of VOC as a moderator, and the engagement of opinion leaders as relationship between the types of customers and ideas, Magnusson
both an independent variable and a moderator. (2009) showed that more customer benefits should be included in
VOC contains customer expectations, which includes utilitarian ideas that are suggested by ordinary customers rather than profes-
and hedonic expectations. Examples of utilitarian expectation can sionals (Magnusson 2009). Since ideas are suggested by ordinary
be found in previous studies such as the request for free-WiFi ser- customers rather than by professionals in MSI, VOC would be more
vice, improved tumbler lid technology, lactose free milk option and likely to represent customer benefits, and such VOC could then
lower the price (Batra and Ahtola 1991, Chitturi et al. 2008, Schin- encourage organizational innovation. As mentioned, customer ben-
dler 1998, Strahilevitz and Myers 1998, Dhar and Wertenbroch efits cover utilitarian and hedonic benefits. Whereas exceeding util-
2000). These expectations are related to functional, instrumental, itarian expectation evokes customer satisfaction, exceeding
practical, and monetary benefits of the products or services pro- hedonic expectation leads to customer delight (Chitturi et al.
vided by Starbucks. On the other hand, asking for a more comfort- 2008). Thus, firms which aim for customer satisfaction and delight
able sofa or mix up the classics and new music can be an example will be attentive to the VOC with utilitarian and hedonic expecta-
of hedonic expectation, which refers to customer aesthetic and tions. Consequently, such VOC is more likely to be adopted by firms.
enjoyment-related benefits. Therefore, we propose the following hypotheses:
The utilitarian and hedonic expectations can be grouped as cus-
tomer benefit-based expectations (Magnusson 2009). As adverting H2a. VOC with utilitarian expectation is positively associated with
studies show, customer benefits are major predictors of customer organizational innovation.
responses (Varnali et al. 2012). When members of a brand commu-
nity read ideas relating to customer benefits, they may feel that the
requirements are directly linked to their own gains. The results are H2b. VOC with hedonic expectation is positively associated with
increased discussions about such ideas and the generation of addi- organizational innovation.
tional comments. Consequently, customer benefit-based expecta- Ideal-based expectation is related to what a customer ideally
tion can significantly increase communication among customers needs, and hence, the expectation is not required to be feasible or
in a community. Therefore, we hypothesize: reasonable (Boulding et al. 1993). In this context, community mem-
bers freely convey comments on a posted message because the mes-
H1a. VOC with utilitarian expectation is positively associated with sage is already beyond feasibility and the comments are not required
the interaction among brand community members. to be practical. Thus, interactions among customers increase.
These expectations would likely be stable over time because
ideal-based expectations represent continuing desires and needs
H1b. VOC with hedonic expectation is positively associated with (Boulding et al. 1993, Liljander and Strandvik 1993). Posted mes-
the interaction among brand community members. sages on the online community express what customers continu-
ally want even though the message may not be practical. Since
these voices or expectations are likely to express real customer
Negative Valence of
VOC VOC needs, the company may try to evaluate and examine whether or
not to implement the messages. Accordingly, ideal-based VOC
Utilitarian Expectation H5 (+) would be more likely to lead to innovative products or services
Interaction of a company. Therefore, we propose the following hypotheses.
Hedonic Expectation
H1 (+)
H7 (+) H1c. VOC with ideal-based expectation is positively associated
Ideal-based Expectation
with the interaction among brand community members.
Product-based Expectation H2 (+)

H6 (+)
Brand-based Expectation
H2c. VOC with ideal-based expectation is positively associated
H3 (+)
with organizational innovation.
Experience can cause expectations to arise. Such experience-
Engagement of Opinion H4 (+) Innovation
Leaders based expectation includes product-based expectation and brand-
based expectation (Woodruff et al. 1983). In MSI, ‘‘. . . I want dark
Fig. 1. Research model. chocolate mocha as part of a new beverage menu that is seen in other
H. Lee et al. / Electronic Commerce Research and Applications 13 (2014) 205–219 209

coffee shops. . . .’’ and ‘‘. . . I really loved those cookies and am frus- Furthermore, opinion leaders are likely to be innovative with
trated that they are gone here in this shop. . . .’’ are examples of prod- regard to new products or ideas (Myers and Robertson 1972).
uct-based expectation and brand-based expectation, respectively. Accordingly, we propose that opinion leader involvement and nur-
Most MSI members have sufficient customer experiences and are turing can foster the launching of new products or services in MSI.
familiar with Starbucks products. Accordingly, when a member ex- Furthermore, opinion leadership can encourage other MSI mem-
presses an experience-based expectation, most other members are bers to submit their ideas for the innovation of products or ser-
likely to have a similar experience. Since experience-based expecta- vices. Therefore, we propose the following hypotheses:
tions are usually understandable, it facilitates other members in
deciding to agree or disagree to such VOC. Therefore, experience- H3. The engagement of opinion leaders is positively associated
based expectation will cause discussion to easily take place. As a with the interaction among brand community members.
result, we can expect that VOC characterized by product-based
and brand-based expectations will expedite the interaction among
the customers. Thus, we propose hypotheses as follows: H4. The engagement of opinion leaders is positively associated
with organizational innovation.
H1d. VOC with product-based expectation is positively associated
Negative messages have been shown to have a stronger impact
with the interaction among brand community members.
on the recipients than positive messages. Previous studies have
discussed the negative bias of messages (Anderson 1965, 1971;
Boucher and Osgood 1969; Fiske 1980; Jones and Davis 1965;
H1e. VOC with brand-based expectation is positively associated Kanouse 1984). Jones and Davis (1965) argue that negative infor-
with the interaction among brand community members. mation tends to be considered more credible than positive infor-
mation. There are common normative pressures for people to say
Feasibility has been studied as one of the characteristics of cus-
positive things (Jones and Davis 1965). Thus, when hearing nega-
tomer-generated ideas that determine idea quality. Feasibility can
tive things, it may be considered sincere, which may mean that
be classified into technical feasibility (i.e., the degree to which the
the information has not been influenced by normative pressure
organization can implement the idea concerning organizational
(Jones and Davis 1965). Anderson (1965, 1971) proposed an aver-
capabilities and resources) and economic feasibility (i.e., the de-
aging model, which explains the reason why people tend to accept
gree to which the organization can implement the idea within an
averaged individual impressions in an integrative decision process,
acceptable budget) (Dean et al. 2006, Faure 2004, MacCrimmon
and give more weight to negative messages. As Boucher and Os-
and Wagner 1994, Mumford et al. 2001). Feasibility is an important
good (1969) discussed, positive words are more likely to be used
factor in making decisions on whether or not to adopt an idea. In
than negative words, which contrasts the distinctiveness of nega-
our context, product-based and brand-based expectations origi-
tive messages. The distinctiveness may increase the probability
nated from the past customer experiences. As these requirements
of the message being remembered and attended to (Fiske 1980).
have been implemented in other brands or other products within
Recently, negative bias was empirically tested in online com-
the same brand, implementation of product-based or brand-based
munities. Naveed et al. (2011) showed that negative messages in-
expectations is mostly feasible. Accordingly, brands and products
creased the engagement of members. Consequently, such
are more likely to be adopted by the associated company (i.e. Star-
messages are transferred wider and faster than positive or neutral
bucks). Thus, we propose the following hypotheses:
messages. Similarly, Chmiel et al. (2011) found that negative opin-
ions boost member interaction in an online forum. In our context,
H2d. VOC with product-based expectation is positively associated
therefore, we expect that VOC with negative messages is more
with organizational innovation.
likely to promote conversations on a topic rather than VOC without
them. Thus, we propose the following hypotheses:

H2e. VOC with brand-based expectation is positively associated H5a. The relationship between VOC with utilitarian expectation
with organizational innovation. and the interaction among brand community members is moder-
ated by the negative valence of VOC.
Community engagement consists of community involvement
and community nurturing, which are identified as community
building activities (Hallahan 2004). Community engagement entail
H5b–H5e. H5b–H5e are similarly constructed, for each of hedonic,
facilitating organizational communication or causing participation,
ideal-based, product-based, and brand-based expectations,
and fostering economic, social and cultural vitality of communities,
respectively.
respectively (Hallahan 2004). Opinion leaders are characterized by
domain-specific knowledge and are often involved in product Opinion leaders have great influence on common people in indi-
development (Goldsmith 2001, Goldsmith and Hofacker 1991). vidual acceptance of information (Lazarsfeld et al. 1944). People’s
They demonstrate high levels of product familiarity (Bloch et al. opinions and attitudes can be driven by opinion leaders, and this is re-
1986, Goldsmith et al. 2003) and frequently raise product-related lated to the people’s perception on opinion leaders (Katz and Lazars-
issues (Feick and Price 1987). They are also highly motivated to feld 1955). Opinion leaders do not tend to be seen as forcing a notion
seek product-related information (Myers and Robertson 1972). In on the public, instead, they seem to be trustworthy and non-purpo-
MSI, there is a ‘‘Leader board’’ that displays scores of submitted sive (Katz and Lazarsfeld 1955). This can make people more attentive
ideas, comments, and votes as well as the members who submitted to specific opinions in which an opinion leader is engaged.
the ideas. Members listed on the board are considered opinion The concept of opinion leadership is recently used in strategic
leaders because they were judged to have Starbucks-related spe- attempts to boost involvement of the public in social issues like cli-
cific knowledge and are involved in product development by the mate change campaign. Such attempts show that increased
involvement of submitting ideas and opinions. Consequently, opin- engagement of opinion leaders can promote people to accept infor-
ion leader’s involvement and nurturing may lead to active discus- mation and participate in an issue.
sion among other community members by giving and seeking In our context, we expect that as the engagement of opinion
information. leaders on a certain VOC increases, more attention from MSI
210 H. Lee et al. / Electronic Commerce Research and Applications 13 (2014) 205–219

members might be drawn. The VOC consequently can motivate the opinion leaders, i.e., the idea launcher whose idea has been adopted
customers to participate in evaluation and discussion of such VOC. and the top commenter who actively participates in the commen-
In case that a VOC with utilitarian expectation is developed with tary of the ideas. The engagement of an opinion leader is measured
additional relevant information by opinion leaders, customer inter- in terms of the number of opinion leaders’ comments posted for
action related to the VOC is more likely to increase. This tendency each idea. In summary, we have seven independent variables: (1)
might be similar across all types of expectations contained in VOC. utilitarian expectation, (2) hedonic expectation, (3) ideal-based
Therefore we hypothesize as follows; expectation, (4) product-based expectation, (5) brand-based expec-
tation, (6) negative valence, and (7) opinion leader’s engagement.
H6a. The relationship between VOC with utilitarian expectation Dependent variables are Innovation and Interaction. Innovation
and the interaction among brand community members is moder- is a binary variable which indicates whether the idea is adopted
ated by the engagement of opinion leaders. by Starbucks. Interaction is measured by the number of comments
for each idea and the number of promotion or demotion points.
MSI members can vote on the idea as either the promotion or
H6b–H6e. H6b–H6e are similarly constructed, for each of hedonic, demotion of the idea. Each promotion or demotion will result in
ideal-based, product-based, and brand-based expectations, a 10-point increase or decrease, respectively. Definitions and acro-
respectively. nyms for all variables are described in Table 1.
Opinion leaders are enthusiastic about seeking brand/product
5.2. Qualitative contents analysis
information (Myers and Robertson 1972). They are likely to have
serviceable ideas in regard to new product development (Myers
We analyzed the contents of VOC to assess the characteristics of
and Robertson 1972). Generally, opinion leaders are more knowl-
VOC. Table 2 shows several examples of VOC characterized by cus-
edgeable and innovative than ordinary community members. Most
tomer expectations. VOC could be characterized by several expec-
ideas in their initial stage tend to be rough and need to be advanced.
tations of a customer. VOC, which receives high points and a
Thus, opinion leaders usually have knowledge and innovativeness,
number of comments, would be adopted by Starbucks.
which can play important roles as nutrition in idea evolvement.
Starbucks suggests 15 topic categories in MSI. Table 3 illustrates
In MSI context, increased engagement of opinion leaders can
the number of VOC that we evaluated and the launch rates by cat-
provide more chances for VOC to be developed. As a result, we ex-
egory. ‘Starbucks Card’, ‘Atmosphere & Locations’, and ‘Coffee &
pect that more developed VOC is more adopted by Starbucks.
Espresso Drinks’ are ranked the top three most posted topic cate-
Therefore we hypothesize as follows;
gory. ‘Merchandise & Music’ is ranked the top launch rate (54.7%).
Generally, the number of VOC with a negative message is rela-
H7a. The relationship between VOC with utilitarian expectation
tively small (see Table 4). Such VOC is less likely to be launched (13
and organizational innovation is moderated by the engagement of
out of 75 ideas) than non-negative VOC (318 out of 1365 ideas).
opinion leaders.
Fig. 2 illustrates an example of VOC that includes negative mes-
sages. The poster ‘vinnyalt’ criticizes the Starbucks gold card pro-
H7b–H7e. H7b–H7e are similarly constructed, for each of hedonic, gram and Starbucks marketing. The poster was disappointed by
ideal-based, product-based, and brand-based expectations, the new program because the program does not have any improve-
respectively. ment from the customer viewpoints. The VOC received 390 voting
points with 48 comment threads. However, the VOC is rejected by
Starbucks. The VOC criticizing Starbucks can cultivate negative
5. Analysis WOM through the interactions among community members,
which can be a challenge.
5.1. Data An opinion leader’s engagement in MSI can often be found. Fig. 3
illustrates an example of VOC which claims separated lines of
We developed a Web crawler and collected VOC text data brewed coffee during the morning rush with an opinion leader’s
(110,045 ideas) from the MSI Website. 1526 ideas were selected comments on the idea. As shown in Fig. 3, this idea received 30,690
which are confirmed either as adopted or rejected. We examined points and 36 comment threads. ‘Melody’ is one of the top comment-
the data to detect outliers and to check for normality prior to run- ers and idea launchers in MSI and is considered an opinion leader in
ning the regression. Consequently, we discarded 86 outliers and this study. Opinion leaders can show their simple agreement or dis-
1440 VOC data were selected for our study. The gathered data also agreement, or provide relevant information. Melody shows an opin-
includes customer comments on each idea and some related infor- ion leader’s role in the community by providing information related
mation such as posting date and member identification. to a posted idea (e.g., a space of store and possible consequence of the
The customer’s idea (i.e., VOC) is the focal unit of our analysis. To separate line). The opinion leader’s engagement helps the posted
assess the attributes of each idea and its comment threads, we re- ideas become feasible by providing more information and adjusting
lied on subjective assessment by two experts who have been active incorrect information. The involvement of opinion leaders can also
at MSI for two years. The experts examined 1440 ideas to evaluate lead to frequent conversations among the members.
whether or not they contain these: (1) utilitarian expectation, (2)
hedonic expectation, (3) ideal-based expectation, (4) product- 5.3. Analysis method
based expectation, and (5) brand-based expectation. They were
asked to read each idea description and associated comments. They We employed three regression models: zero-inflated negative
evaluated using a binary scale (e.g., 0 = contains no utilitarian binomial (ZINB) model, ordinary least squares (OLS) for dependent
expectation of a customer and 1 = contains utilitarian expectation). variable interaction, and the binomial logistic model for dependent
Difference in evaluation by the two experts was adjusted through variable innovation.
discussion. The experts also evaluated the negative valence of Zero-inflated count models such as ZINB or zero-inflated Poisson
VOC and the opinion leader’s engagement in an idea. The negative (ZIP) are applied to count variables with excessive zeros (Ridout
valence of VOC represents whether VOC includes or does not in- et al. 2001). The excess zeros are generated by a separate process
clude negative messages. MSI officially designates two types of from the count values and thus can be modeled independently.
H. Lee et al. / Electronic Commerce Research and Applications 13 (2014) 205–219 211

Table 1
Operationalization of variables.

Variable Definition Abbreviation


Organizational innovation Whether or not VOC is accepted by Starbucks for its innovation Innovation
Interaction among a community The number of comments for each idea Interaction_c
members
Promotion (demotion) points Interaction_p
Utilitarian expectation Whether or not VOC contains practical expectations including customer’s monetary benefits Utilitarian
Hedonic expectation Whether or not VOC contains enjoyment-related expectations of customer benefits Hedonic
Ideal-based expectation Whether or not VOC contains customer expectations in an ideal sense for which Starbucks does not Ideal_based
provide
Product-based expectation Whether or not VOC contains customer expectations related to the product in which customers Product_based
experienced
Brand-based expectation Whether or not VOC contains customer expectations related to brand in which customers experienced Brand_based
Negative valence of VOC Whether or not the contents of VOC is negative without any suggestion Neg_val
Opinion leaders’ engagement The number of comment threads in which a top commenter or an idea launcher post OL_eng

Table 2
VOC launched by Starbucks.

VOC title ‘‘VOC description’’ Points No. of (1)a (2) (3) (4) (5)
comments
Starbuck’s card
‘‘From the My Starbucks Idea team: Please see ‘‘Punch Card System’’. I’ve had, and used, a Starbuck’s card for quite a 10,010 18 Y N Y N Y
while. However, I have always felt that there should be some reward for continuing to put money on it, and using it.
Perhaps a free drink after each 10 purchases...or after certain money amount. . .’’
Happy Hour
‘‘. . . Why could you not do something similar, even if its 1.50 off any Venti, $1 off Grande, .75 off tall and .25 off Kids from 6160 23 Y N Y N N
1–3 everyday? The 1/2 price would be much better you could make it 1/2 price any drink from 3–4 or 3–5:30. . .’’
Hawaiian Kona Coffee
‘‘I was wondering if you would like to see Starbucks offer Kona Coffee in more stores for customers to buy. It would be 2040 39 N Y Y N N
neat to be able to buy a coffee from Hawaii, because it is the only State in the US that grows coffee.’’
Sugar free syrups
‘‘We also need more sugar free flavors. There are those of users who need the sugar free syrups and the selections are 39,680 11 N Y N Y N
very limited. . . .. SO....MORE sugar free syrups PLEASE!!’’
Recycling
‘‘. . .Starbucks should offer nationwide recycling at their stores. A lot of their competitors do it (aka Coffee Bean and Tea 90,490 2 N N Y N N
Leaf). Seeing as Recycling is the new hot thing and going green, it would really boost PR and be a great new ad
campaign.’’
a
(1) Utilitarian, (2) Hedonic, (3) Ideal_based, (4) Product_based, (5) Brand_based.

Table 3
The launch rate of VOC for each topic category.

Topic category Launched VOC Rejected VOC Subtotal Launch rate (%)
Coffee & Espresso Drinks 50 151 201 24.9
Frappuccino Beverage 12 45 57 21.1
Tea & Other Drinks 13 28 41 31.7
Food 40 89 129 31.0
Merchandise & Music 47 39 86 54.7
Starbucks Card 56 158 214 26.2
New Technology 11 31 42 26.2
Other Product Ideas 17 39 56 30.4
Ordering & Payment & Pick-Up 13 53 66 19.7
Atmosphere & Locations 14 197 211 6.6
Other Experience Ideas 21 102 123 17.1
Building Community 7 49 56 12.5
Social Responsibility 19 75 94 20.2
Other Involvement Ideas 5 47 52 9.6
Outside USA 6 6 12 50.0
Total 331 1109 1440 23.0

The bold numbers on the Subtotal column are the top three most posted topic category.The bold numbers on the Launch rate column are the top three most launched topic
category.

The dependent variable interaction_c has a mean of 7.32 and a var- come from a different generating process. The Vuong test compares
iance of 105.02 (see Table 5). The number of zero observations with- the zero-inflated model to a standard NBM.1 The zero-inflated count
in the data is large (i.e., 328) (see Fig. 4). Based on the characteristic models assume two latent groups: group A as an always 0 group
of the dependent variable, we can use either the negative binomial
model (NBM) and the Poisson model, or the ZINB and ZIP models
1
(Ridout et al. 2001) to quantify the effects VOC types on the interac- If the test statistics Z is greater than 1.96, the zero-inflated model is preferred. If Z
is less than 1.96, the NBM is preferred. The value Z between 1.96 and 1.96 indicate
tion among community members. The Vuong test (1989) deter-
that neither model is preferred, which means that the excessive zeros for the model
mines whether the data are negative binomial or the excess zeros response are not sufficient to warrant adjustment by a zero-inflated model.
212 H. Lee et al. / Electronic Commerce Research and Applications 13 (2014) 205–219

Table 4
Negative valence of VOC versus non-negative valence of VOC.

Topic category Negative valence of VOC Non-negative valence of VOC Total


Subtotal Launched Rejected Subtotal Launched Rejected
Coffee & Espresso Drinks 19 2 17 182 48 134 201
Frappuccino Beverage 4 1 3 53 11 42 57
Tea & Other Drinks 0 0 0 41 13 28 41
Food 5 3 2 124 37 87 129
Merchandise & Music 3 2 1 83 45 38 86
Starbucks Card 20 4 16 194 52 142 214
New Technology 0 0 0 42 11 31 42
Other Product Ideas 2 0 2 54 17 37 56
Ordering & Payment & Pick-Up 3 0 3 63 13 50 66
Atmosphere & Locations 2 0 2 209 14 195 211
Other Experience Ideas 10 0 10 113 21 92 123
Building Community 1 0 1 55 7 48 56
Social Responsibility 4 1 3 90 18 72 94
Other Involvement Ideas 2 0 2 50 5 45 52
Outside USA 0 0 0 12 6 6 12
Total 75 13 62 1365 318 1047 1440

Gold Card changes

Posted on 11/22/2009 11:51 AM

by vinnyatl

“Your email dated 11/2/09 announcing "More great bene fits are coming your way" was an insult to

the intelligence of anyone who read it. This was not an improvement in a loyalty program, … But I

have no tolerance for marketing gibberish that's supposed to make me feel like the change was an

improvement when in fact it represented a significant reduction in benefits for high-spending, loyal

customers. ”

Fig. 2. Example of negative valence of VOC (390 points and 48 comment threads).

(inflation group); and group B as a not always 0 group. Entities in 5.4. Results
group A have an outcome of 0 with a probability of 1, while entities
in group B might have a zero count. However, there is a non-zero 5.4.1. Zero-inflated negative binomial (ZINB)
probability of being a positive count in group B (Scott and Freese We ran several ZINB models in terms of independent variables.
2006). Model (1-1) only uses expectations and opinion leader’s engage-
OLS is used to evaluate the influence of independent variables ment. Model (1-2) adds the moderating of opinion leader’s engage-
on log points of VOC (interaction_p). When a log value is used as ment to Model (1-1). Model (1-3) adds the moderating of negative
a dependent variable, the results can be interpreted such that the valence of VOC to Model (1-1). Finally, Model (1-4) incorporates all
dependent variable (valued in points) changes by 100  (coeffi- independent variables and moderators. We have 1440 complete
cient) percent for a one unit increase in the independent variable observations in this analysis.
while all other variables in the model are held constant. The Vuong test showed that the negative binomial regression
A binomial logistic model is used to model the dichotomous model would be significantly worse statistically than the ZINB
outcome variables. In the logistic model, the log’s ‘‘odds of the out- model, since the observed number of zeros is significantly higher
come’’ is modeled as a linear combination of the predictor vari- than the number expected by the negative binomial model in all
ables. Our research model has a dummy variable to indicate models (z = 8.30, Pr > z = 0.000; z = 8.19, Pr > z = 0.000; z = 8.33,
whether or not Starbucks adopted or rejected the VOC for a new Pr > z = 0.000; z = 8.24, Pr > z = 0.000). In addition, the likelihood
product or service (1 = adopted, 0 = rejected). The regression coef- ratio test showed that all models with ZINB fit more significantly
ficients estimate the impact of the independent variables on the than ZIP (Prob > chi2 = 0.000), confirming our self-inspection of
probability that the company will launch VOC. We selected the the data (Greene 1994).
binomial logistic model because our method is considered more The results of running the ZINB include the negative binomial
effective than other methods in the goodness-of-fit of the analysis. regression of the output and the logistic regression for excess zeroes
The model is also appropriate to explain the relationship between called ‘‘inflate’’ (Table 6). In Model (1-1), the coefficient for the vari-
the dependent and independent variables where the dependent able utilitarian expectation is 0.507. We exponentiated the coefficient
variable is binary (Hosmer and Stanley 2000). We applied STATA to obtain the value of 1.66 in order to interpret the effect of utilitarian
11.0 for the ZINB and binomial logistic regression in this study. expectation on the dependent variable. Specifically, the number of
H. Lee et al. / Electronic Commerce Research and Applications 13 (2014) 205–219 213

Fig. 3. Example of an opinion leader’s engagement (30,690 points and 36 comment threads).

comments on VOC with utilitarian expectation is approximately 1.66


Table 5 times the number of comments on the other VOC. This difference is
Descriptive statistics of variables.
statistically significant (p < 0.001). We can obtain a 95% confidence
Variable Observation Mean Std. dev. Min. Max. interval for the coefficient as Bi ± 1.96 SE(Bi) = 0.507 ±
Innovation 1440 0.230 0.421 0 1 1.96  0.088 = (0.335, 0.679), which when exponentiated, results in
Interaction a confidence interval of (1.397, 1.972) for the rate ratio.
Interaction_c 1440 7.320 10.248 0 50 Table 6 illustrates additional output under the title ‘‘inflate.’’ This
Interaction_pa 1371 5.099 1.974 2.303 11.413 part of the output describes the logistic regression analysis for pre-
Utilitarian 1440 0.217 0.412 0 1
Hedonic 1440 0.510 0.500 0 1
dicting whether the outcome variable must be equal to zero or
Ideal_based 1440 0.493 0.500 0 1 whether it has the negative binomial distribution. This output is of
Product_based 1440 0.112 0.315 0 1 interest in itself. The output differentiates between VOC that had
Brand_based 1440 0.116 0.320 0 1 zero comments without the possibility of a positive number of com-
Neg_val 1440 0.052 0.222 0 1
ments, and VOC that resulted in the negative binomial distribution
OL_eng 1440 2.873 6.080 0 52
of the number of comments including the possibility of zero com-
a
Log-transformed points are used. ments. Specifically, it identifies the factors associated with whether
VOC had only the possibility of zero comments or resulted in the
negative binomial distribution of comments. For example, the coef-
ficient for utilitarian expectation is 2.761. The VOC with utilitarian
expectation has a lower likelihood of having only zero comments. In
other words, the VOC with utilitarian expectation is more likely to
have the negative binomial distribution, and is more likely to have
additional comments than the VOC with the lower level of utilitarian
expectation. We can interpret the exponentiated coefficient as an
odds ratio in logistic regression. Specifically, the odds of claiming
utilitarian expectation that have only zero comments are
e 2.761 = 0.063. The inverse of this quantity is 1/0.063 = 15.816. For
VOC with zero comment, the odds that VOC with utilitarian expec-
tation resulted in a negative binomial distribution are 15.816 times
greater than the odds that VOC without utilitarian expectation re-
sulted in a negative binomial distribution. Moreover, we note from
the output that this association is statistically significant.

5.4.2. OLS regression


Table 7 describes the results of the log promotion points for the
four OLS regression models. Model (2-4) is the best model to ex-
Fig. 4. Frequency in the number of comments of VOC. plain the log promotion points with the highest R2 of 0.183. All
214 H. Lee et al. / Electronic Commerce Research and Applications 13 (2014) 205–219

Table 6
ZINB regression: The number of comments.

Interaction (the number of Model (1-1) Model (1-2) Model (1-3) Model (1-4)
comments)
Negative binomial Inflate Negative binomial Inflate Negative binomial Inflate Negative binomial Inflate
portion portion portion portion portion portion portion portion
Utilitarian 0.507 2.761 0.492 2.900 0.471 2.728 0.472 2.822
Hedonic 0.235 0.019 0.290 0.039 0.200 0.007 0.269 0.032
Ideal_based 0.168 0.434 0.164 0.434 0.176 0.435 0.171 0.435
Product_based 0.197 0.418 0.128 0.457 0.196 0.411 0.150 0.437
Brand_based 0.109 0.595 0.090 0.601 0.053 0.630 0.052 0.620
OL_eng 0.085 24.853 0.088 24.821 0.080 24.515 0.085 24.315
Utilitarian  OL_eng 0.005 0.002
Hedonic  OL_eng 0.0139 0.0172
Ideal_based  OL_eng 0.001 0.001
Product_based  OL_eng 0.018 0.013
Brand_based  OL_eng 0.005 0.001
Utilitarian  Neg_val 0.514 0.510
Hedonic  Neg_val 0.911 0.957
Ideal_based  Neg_val 0.603 0.592
Product_based  Neg_val 0.109 0.167
Brand_based  Neg_val 0.274 0.241
Constant 1.380 0.446 1.367 0.452 1.360 0.428 1.332 0.440
Prob > chi2 0.000 0.000 0.000 0.000
Vuong test z = 8.30 z = 8.19 z = 8.33 z = 8.24
Pr > z = 0.000 Pr > z = 0.000 Pr > z = 0.000 Pr > z = 0.000
Log likelyhood 3923.37, p < 0.01 3922.26, p < 0.01 3890.64, p < 0.01 3889.37, p < 0.01

p < 0.01.

p < 0.05.

Table 7
Results of OLS: Log promotion points.

Interaction (log points) Model (2-1) Model (2-2) Model (2-3) Model (2-4)
  
Utilitarian 2.0488 1.9780 2.0344 1.9619
Hedonic 0.6264 0.5407 0.5910 0.5332
Ideal_based 0.2313 0.1994 0.2363 0.1975
Product_based 0.6308 0.7307 0.6510 0.7407
Brand_based 0.2130 0.1879 0.1658 0.1994
OL_eng 0.0508 0.0294 0.0420 0.0251
Utilitarian  OL_eng 0.0230 0.0243
Hedonic  OL_eng 0.0298 0.0226
Ideal_based  OL_eng 0.0139 0.0150
Product_based  OL_eng 0.0383 0.0446
Brand_based  OL_eng 0.0082 0.0158
Utilitarian  Neg_val 0.3387 0.2996
Hedonic  Neg_val 1.4754 1.3737
Ideal_based  Neg_val 0.6894 0.6976
Product_based  Neg_val 0.7470 0.2979
Brand_based  Neg_val 0.4171 0.5469
Constant 3.9600 4.0275 3.9636 4.0165
Observations 1371 1371 1371 1371
R-squared 0.167 0.170 0.180 0.183

p < 0.01.

p < 0.05.

expectations excluding the brand-based expectation statistically the VOC with opinion leader’s engagement increase by a factor of
influenced the interactions (promotion points) within MSI. We 0.996, which is insignificant.
did not find a moderating effect of OL_eng on the relationship be- As shown in Models (3-2), (3-3), and (3-4), the moderating ef-
tween expectations of customers and their interactions. The nega- fects of OL_eng and Neg_val are non-existent for innovation. Ex-
tive valence of VOC has a moderating effect only on the cept for brand-based expectation, the other expectations show a
relationship between hedonic expectation and interaction. significant effect on organizational innovation.
Table 9 illustrates the best models (1-4, 2-4, and 3-4) among
competing regression models. Except for the brand-based expecta-
5.4.3. Binomial logistic regression tion, the VOC characterized by claiming customer expectations
The results of the binomial logistic regression are shown in Ta- positively affects the interaction among the MSI members and
ble 8. In Model (3-1), for every single unit change in VOC with util- the adoption of VOC. The engagement of the opinion leaders only
itarian expectation, the log of odds for adopting the VOC increases influenced the interaction among the MSI members. The moderat-
by 1.013. For every single unit increase in VOC with product-based ing effect of negative valence of VOC shows positive for the rela-
expectation, the log of odds for launching the VOC increases by tionship between customer expectation and their interactions.
0.814. To interpret the coefficient of OL_eng, we exponentiated However, it is shown that the moderating effect of the engagement
the coefficient and obtained a value of 0.996. For every single unit of the opinion leaders is insignificant. A summary of hypotheses
increase in an opinion leader’s engagement, the odds for adopting test results are shown in Table 10.
H. Lee et al. / Electronic Commerce Research and Applications 13 (2014) 205–219 215

Table 8
Results of binomial logistic regression: Adopted versus rejected.

Innovation Model (3-1) Model (3-2) Model (3-3) Model (3-4)


  
Utilitarian 1.013 1.075 1.018 1.083
Hedonic 1.101 1.121 1.092 1.119
Ideal_based 0.505 0.579 0.513 0.583
Product_based 0.814 0.999 0.820 0.996
Brand_based 0.100 0.302 0.123 0.295
OL_eng 0.004 0.025 0.001 0.027
Utilitarian  OL_eng 0.018 0.018
Hedonic  OL_eng 0.004 0.002
Ideal_based  OL_eng 0.025 0.025
Product_based  OL_eng 0.072 0.082
Brand_based  OL_eng 0.090 0.093
Utilitarian  Neg_val 0.115 0.165
Hedonic  Neg_val 0.056 0.328
Ideal_based  Neg_val 0.436 0.334
Product_based  Neg_val 0.095 0.645
Brand_based  Neg_val 0.366 0.223
Constant 2.426 2.526 2.424 2.523
Observations 1440 1440 1440 1440

by community members. Utilitarian and hedonic expectations rep-


Table 9
Comparison of the effects of IV on DVs.
resent customer benefits (Magnusson 2009). When VOC claims cus-
tomer benefits, community members are likely to be interested in
the VOC and would likely support and comment on the idea.
Interaction
Models Innovation Ideal-based expectations represent requirements which are not
Interaction_c Interaction_p provided by Starbucks. Although customers do not experience such
requirements at Starbucks, they can discuss and share their own
Utilitarian + + + opinions that can lead to a more concrete shape of the
requirements.
Hedonic + + +
While product-based expectation is positively associated only
Ideal_based + NS + with the promotion points, brand-based expectation does not affect
the interaction among the community members on the contrary to
Product_based NS + + our expectation. The reason why these expectations do not drive an
active discussion seems to be the transparency of VOC. The posted
Brand_based NS NS NS VOC is too clear to discuss within a community because of customer
experiences. For example, one VOC in our dataset is asking to re-
OL_eng + NS NS
store the original gold card program that offers members free syrup
Utilitarian * OL_eng NS NS NS and coffee refills. Only two comments on the VOC are registered,
because the VOC has a clear meaning. Another possible explanation
Hedonic * OL_eng NS NS NS exists. The numbers of product-based and brand-based expecta-
Ideal_based * tions are less than those of the other expectations (Table 5). The
NS NS NS mismatch between the VOC characteristics and the goal of ideation
OL_eng
Product_based * sites may have occurred because these expectations often result
NS NS NS
OL_eng from customer experiences at Starbucks while Ideation sites aim
Brand_based * to collect ideas rather than experiences.
NS NS NS
OL_eng
Utilitarian * Neg_val + NS NA 6.2. Customer expectations and organizational innovation

Hedonic * Neg_val + + NA
Customer expectations in VOC are found to be the key determi-
Ideal_based * nants of organizational innovation. Utilitarian, hedonic, ideal-
+ NS NA
Neg_val based, and product-based expectations are positively related to
Product_based *
NS NS NA launching an idea. Many companies have tried to hear customer
Neg_val
Brand_based * voices because VOC includes customer needs and requirements.
NS NS NA Starbucks would like to adopt and serve the posted VOC since it
Neg_val
Significant positive (+); Nonsignificant (NS); Not applicable (NA) represents customer needs. However, brand-based expectation
does not affect the adoption of VOC, which may be due to the lim-
Significant positive (+); nonsignificant (NS); not applicable (NA).
itation of a small dataset (see Table 5). Brand-based expectations
demands new products or services which have been provided by
6. Discussion other companies such as Coffee Bean & Tea Leaf. These demands of-
ten depend on subjective personal experiences. Thus, it could be a
6.1. Customer expectations and online community interactions tough decision for Starbucks to adopt the brand-based VOC, when
considering technical and economic feasibilities. Only a limited
The expectations of customers contained in VOC are found to be number of ordinary customers understand the underlying financial
the key determinants of the interaction among the community structure of organizations (Christensen and Bower 1996, Martin
members. Utilitarian, hedonic, and ideal-based expectations are 1995). While investigating why brand-based expectations are not
positively related to the comment threads and the points promoted influential, we found one VOC that asked for comfortable sofas
216 H. Lee et al. / Electronic Commerce Research and Applications 13 (2014) 205–219

Table 10
Summary of results.

and chairs, which corresponds to brand-based expectation. There tions, resulting in the cultivation of negative WOM. The negative
was a comment on the VOC from an idea partner who represented valence of VOC shows a moderating effect on the relationships be-
Starbucks in MSI: ‘‘We will take this into consideration when tween utilitarian, hedonic, or ideal-based expectations and com-
building and remodeling stores. Another consideration is the ment threads. The partially significant moderating effect of the
amount of seating too. We try to balance both soft seating and negative valence on the relationship between customer expecta-
the number of seats to meet all the customer needs.’’ From this tions and interactions may be caused by the limitation of a small
comment, we can see that Starbucks thought that it was worth- data set (see Table 5). Norms such as constructive suggestions
while to implement the VOC, but not at that moment. Thus, Star- rather than criticism in ideation sites may affect the amount of
bucks decided to reject the VOC. the negative valence of VOC.
When a VOC claims these expectations with a negative valence,
community member opinions on the VOC can be divided into two
6.3. Moderating effect of negative valence of VOC
groups and show stark differences between the two groups (Ri-
chins 1983). Consequently, member interactions are more likely
We examined the moderating effect of negative valence of VOC
to increase. Thus, the involvement of Starbucks could be required
on the relationships between customer expectations and interac-
to deal with the VOC before producing negative WOM. Starbucks
tion among the community members in order to observe potential
can explain the stance of the company on the issue.
threats. VOC with negative messages may produce active interac-
H. Lee et al. / Electronic Commerce Research and Applications 13 (2014) 205–219 217

For example, there is a complaint in our dataset about the taste organizations that accumulate a tremendous amount of VOC
of new Frappuccino. A customer who disappointed the new Frap- through booming online communities are better off classifying
puccino requested for a recipe of the old Frappuccino. The VOC VOC before they are too abundant. By adapting the Expectancy Dis-
earned 2150 points and 50 comment threads. Many of the com- confirmation Theory, we suggested five types based on the con-
ments agreed with the VOC. A person from the Frappuccino team tents of VOC. In addition, used data is collected using a web
explained the reason why Starbucks would not provide the Frap- crawler, and thus, results without method bias are obtained. We
puccino recipe. She suggested community members to share per- examined the effect for each VOC type on organizational innova-
sonal Frappuccino recipes. After her comment, some members tion and interaction among brand community members using
posted their own recipes to share and other members were able qualitative content analysis and quantitative analysis.
to create their own tastes following the shared recipes. Although In the context of open innovation, we investigated the two-
the VOC was not adopted by Starbucks, the Frappuccino team sided VOC as a positive reward and a potential threat. The results
member dealt with the VOC with negative valence to prevent the show that VOC can be a source of organizational innovation and
negative WOM from being produced. brand community promotion. At the same time, VOC can be a po-
tential threat when it contains negative valence. Presently, there
6.4. Opinion leader’s engagement are no empirical studies that examine a two-sided VOC. Even
though limitations of the dataset exist, our examination can con-
The engagement of opinion leaders is shown to have a positive tribute to VOC research and online brand communities. Further-
effect on the interaction among the community members. The more, we employed the number of comments and promotion
interaction is measured in terms of the number of comment points for each idea obtained as measurements of interaction. An
threads and promotion (or demotion) points earned. While the di- extended measurement of these interactions could provide more
rect effect of opinion leader’s engagement on interaction is signif- rigorous evidence and justification of our study results.
icant, the moderating effect is insignificant for the relationship
between customer expectations and interactions. We also realized 7.2. Practical contributions
that opinion leader’s engagement has an insignificant effect on
organizational innovation. Although it is usual for the opinion lea- The customer’s voice affects organizational innovation and
der group to contribute more to a company than other ordinary interaction within a brand community in different ways depending
member groups, there may only be a minor difference between on their contents. VOC that includes customer benefits, ideal-
the two groups in terms of knowledge related to Starbucks coffee based, and product-based expectations are more likely to
products services, and the company. Starbucks coffee brand does contribute to organizational innovation and the interaction among
not require much expertise for ideation and most ordinary mem- community members. Contrary to our expectations, VOC that in-
bers already have enough knowledge and experiences related to cludes brand-based expectation is unlikely to associate with the
Starbucks products and services. Opinion leader engagement may interaction among community members. The insignificant rela-
have an insignificant effect on organizational innovation. This tionships between brand-based expectation and organizational
interpretation is consistent with prior research suggesting that innovation and between brand-based expectation and interaction
the correlation between opinion leader engagement and organiza- may be caused by the lack of feasibility of ideas, limitation of a
tional innovation of consumer products is not significant (Sum- small dataset, and the mismatch between VOC characteristics
mers 1971). The relationship could depend on the context or and the goal of ideation sites. The experimental results imply that
industry. For future research, we can control the characteristics companies can filter VOC based on the contents and further ana-
of an industry and add asymmetric information to our research lyze valuable VOC.
model to further investigate this issue. Specifically, the associations between expectations of VOC and
community member interaction are affected by the negative va-
lence of VOC. Companies that operate an online brand community
6.5. Interaction and organizational innovation
should focus on negative messages and eliminate the threat of neg-
ativity. As shown in the previous example of MSI, a person repre-
We examined the effect of VOC contents on the interaction
senting a company can actively explain company policy and plan,
among the community members and organizational innovation.
and change online conversations from a negative to constructive
By evaluating the ideas posted on MSI, we found that there were
reinforcement. With an increasing interest in open innovation
several examples where active interactions develop a posted idea
environments, we revisited the opinion leader role within an on-
and resulted in organizational adoption. For example, we asked
line brand community. Opinion leaders can contribute to promote
for a ‘‘splash stick’’ instead of a sticker to cover lid openings. There
an online community rather than to lead to organizational innova-
were 128 comments registered to elaborate on the original idea.
tion by posting their opinions and product knowledge. We recom-
After the idea was refined through members’ communications,
mend that organizations should recognize the role of opinion
Starbucks finally launched the idea. Although the scope of our
leaders, encourage organization involvement, and provide opinion
study does not examine the relationship between interactions
leaders with the appropriate knowledge regarding company po-
and organizational innovation, it may be worthwhile to examine
lices, products, and services.
the relationship between interactions and organizational innova-
tion with the former as a predictor of the latter.
7.3. Conclusions

7. Contributions and conclusions There are a few limitations for this study. First, the dataset was
collected from an online brand community that is created and sup-
7.1. Theoretical contributions ported by Starbucks. Further studies should try to validate and
elaborate the findings by analyzing VOC in other online brand com-
Our research extends VOC research by identifying the types of munity types (e.g., communities supported by third parties such as
customer expectations and their association with the interaction professional bodies or customer associations). Second, we opera-
among community members and organizational innovation. Prior tionalized the variables using qualitative content analysis of cus-
studies have focused on gathering and exploiting VOC. However, tomer postings. Although we applied great care to resolve all
218 H. Lee et al. / Electronic Commerce Research and Applications 13 (2014) 205–219

coding discrepancies, the possibility exists that certain customer Chesbrough, H.W., 2003a. The era of open innovation. MIT Sloan Management
Review 44 (3), 35–41.
voices could be omitted from our coding process. Such potential
Chesbrough, H.W., 2003b. Open Innovation: The New Imperative for Creating and
omissions can affect the overall validity and generalizability of Profiting from Technology. Harvard Business Press, Boston, MA.
the findings. Third, a number of customer-related variables in the Chitturi, R., Raghunathan, R., Mahajan, V., 2008. Delight by design: the role of
proposed research model may not have been included. For exam- hedonic versus utilitarian benefits. Journal of Marketing 72 (3), 48–63.
Chmiel, A., Sobkowicz, P., Sienkiewicz, J., Paltoglou, G., Buckley, K., Thelwall, M.,
ple, customer product- or brand-related engagement may affect Hołyst, J.A., 2011. Negative emotions boost user activity at BBC forum. Physica
customer contributions to the brand community. For future re- A: Statistical Mechanics and its Applications 390 (16), 2936–2944.
search, customer-related variables and a new path from interaction Christensen, C.M., Bower, J.L., 1996. Customer power, strategic investment, and the
failure of leading firms. Strategic Management Journal 17 (3), 197–218.
to innovation can be added to our research model to examine the Cooper, R.G., 1993. The performance impact of product innovation strategies.
potential effect of interaction. Fourth, Starbucks’s adoption of European Journal of Marketing 18 (5), 5–54.
VOC can be determined by customer voices and company factors Day, R.L., 1980. Research perspectives on consumer complaining behavior. In: Lamb,
C.W., Jr., Dunne, P.M. (Eds.), Theoretical Developments in Marketing. American
such as internal resources and strategic goals. For future research, Marketing Association, Chicago, pp. 211–215.
our research model can be extended to include specific company Dean, D.L., Hender, J.M., Rodgers, T.L., Santanen, E.L., 2006. Identifying quality,
factors. novel, and creative ideas: constructs and scales for idea evaluation. Journal of
the Association for Information Systems 7 (10), 646–698.
This study offers several new insights that enhance our under- Denove, C., Power IV, J.D., 2007. Satisfaction: How Every Great Company Listens to
standing of VOC within an online brand community. Companies the Voice of the Customer. Penguin.com, New York.
can obtain advice on how to make use of VOC that demonstrate Diener, B.J., Greyser, S.A., 1978. Consumer views of redress needs. Journal of
Marketing 42 (4), 21–27.
double-edged sword characteristics. This study provides several
Dhar, R., Wertenbroch, K., 2000. Consumer choice between hedonic and utilitarian
types of VOC based on the Expectancy Disconfirmation Theory. goods. Journal of Marketing Research 37 (1), 60–71.
VOC shows different influences on the interaction within a brand Enkel, E., Gassmann, O., Chesbrough, H., 2009. Open R&D and open innovation:
community and organizational innovation depending on the con- exploring the phenomenon. R&D Management 39 (4), 311–316.
Enkel, E., Gassmann, O., 2010. Creative imitation: exploring the case of cross
tents. VOC claiming utilitarian, hedonic, ideal-based, and prod- industry innovation. R&D Management 40 (3), 256–270.
uct-based expectations could affect organizational innovation Füller, J., Bartl, M., Ernst, H., Mühlbacher, H., 2006. Community based innovation:
and interaction among the community members. These expecta- how to integrate members of virtual communities into new product
development. Electronic Commerce Research 6 (1), 57–73.
tions can be exploited as criteria for filtering and analyzing VOC. Franke, N., von Hippel, E., Schreier, M., 2006. Finding commercially attractive user
The negative valence of VOC is considered a potential threat and innovations: a test of lead-user theory. Journal of Product Innovation
could arise from active community interaction. Thus, companies Management 23 (4), 301–315.
Faure, C., 2004. Beyond brainstorming: effects of different group procedures on
need to handle the negative valence of VOC carefully. Moreover, selection of ideas and satisfaction with the process. Journal of Creative Behavior
opinion leaders play an important role in boosting brand commu- 38 (1), 13–34.
nity interaction rather than leading to organizational innovation. Feick, L.F., Price, L.L., 1987. The market maven: a diffuser of marketplace
information. Journal of Marketing 51 (10), 83–97.
Gaining clarity about the effect of VOC on an online brand commu- Fiske, S.T., 1980. Attention and weight in person perception: the impact of negative
nity will be important to ensure if a VOC acts as a gift or threat to a and extreme behavior. Journal of Personality and Social Psychology 38 (6), 889–
company. 906.
Fournier, S., 1998. Consumers and their brands: developing relationship theory in
consumer research. Journal of Consumer Research 24 (4), 343–353.
Gangi, P.M.D., Wasko, M., 2009. Steal my idea! Organizational adoption of user
References innovations from a user innovation community: a case study of Dell IdeaStorm.
Decision Support Systems 48 (1), 303–312.
Gloet, M., Terziovski, M., 2004. Exploring the relationship between knowledge
Ahlstrom, D., 2010. Innovation and growth: how business contributes to society.
management practices and innovation performance. Journal of Manufacturing
Academy of Management Perspectives 24 (3), 11–24.
Technology Management 15 (5), 402–409.
Algesheimer, R., Dholakia, U., and Hermann, A. Interplay between brand and brand
Goldsmith, R.E., 2001. Using the domain specific innovativeness scale to identify
community: evidence from European car clubs, 2004. Available at SSRN 534542.
innovative internet consumers. Internet Research 11 (2), 149–158.
Anderson, N.H., 1965. Averaging versus adding as a stimulus-combination rule in
Goldsmith, R.E., Flynn, L.R., Goldsmith, E.B., 2003. Innovative consumers and market
impression formation. Journal of Experimental Psychology 70 (4), 394–400.
mavens. Journal of Marketing Theory and Practice 11 (4), 54–65.
Anderson, N.H., 1971. Integration theory and attitude change. Psychological Review
Goldsmith, R.E., Hofacker, C.F., 1991. Measuring consumer innovativeness. Journal
78 (3), 171–206.
of the Academy of Marketing Science 19 (3), 209–221.
Arndt, J., 1967. Role of product-related conversations in the diffusion of a new
Greene, W. H. Accounting for excess zeros and sample selection in Poisson and
product. Journal of Marketing Research 4 (3), 291–295.
negative binomial regression models. Working paper EC-94-10, Leonard N.
Bagozzi, R.P., Dholakia, U.M., 2006. Antecedents and purchase consequences of
Stern School of Business, New York University, 1994.
customer participation in small group brand communities. International Journal
Griffin, A., Hauser, J.R., 1993. The voice of the customer. Marketing Science 12 (1), 1–27.
of Research in Marketing 23 (March), 45–61.
Hallahan, K., 2004. ‘‘Community’’ as a foundation for public relations theory and
Balachandra, R., Friar, J.H., 1997. Factors for success in R&D projects and new
practice. Communication Yearbook 28, 233–280.
product innovation: a contextual framework. Engineering Management, IEEE
Hart, S., Jan Hultink, E., Tzokas, N., Commandeur, H.R., 2003. Industrial companies’
Transactions on 44 (3), 276–287.
evaluation criteria in new product development gates. Journal of Product
Barlow, J., 1996. A Complaint is a Gift: Using Customer Feedback as a Strategic Tool.
Innovation Management 20 (1), 22–36.
Berrett-Koehler Publishers.
Hauser, J.R., 1993. How Puritan-Bennett used the house of quality. Sloan
Barlow, J., Moller, C., 2008. A Complaint is a Gift: Recovering Customer Loyalty
Management Review 34 (3), 61–70.
When Things Go Wrong. Berrett-Koehler Publishers.
Hauser, J.R., Zettelmeyer, F., 1997. Metrics to evaluate R, D&E. Research Technology
Batra, R., Ahtola, O.T., 1991. Measuring the hedonic and utilitarian sources of
Management 40 (4), 32–38.
consumer attitudes. Marketing letters 2 (2), 159–170.
Hienerth, C., 2006. The commercialization of user innovations: the development of
Beamish, P.W., Lupton, N.C., 2009. Managing joint ventures. Academy of
the rodeo kayak industry. R&D Management 36 (3), 273–294.
Management Perspectives 23 (2), 75–94.
Hosmer Jr., D.W., Stanley, L., 2000. Applied Logistic Regression. Wiley.
Bernoff, J., Li, C., 2008. Harnessing the power of the oh-so-social web. MIT Sloan
Huang, J.-H., Huang, C.-T., Wu, S., 1996. National character and response to
Management Review 49 (3), 36–42.
unsatisfactory hotel service. International Journal of Hospitality Management
Bickart, B., Schindler, R.M., 2001. Internet forums as influential sources of consumer
15 (3), 229–243.
information. Journal of Interactive Marketing 15 (3), 31–40.
Iansiti, M., MacCormack, A., 1997. Developing products on Internet time. Harvard
Bloch, P.H., Sherrell, D.L., Ridgway, N.M., 1986. Consumer search: an extended
Business Review 75 (5), 108–117.
framework. Journal of Consumer Research 13 (1), 119–126.
Jones, E.E., Davis, K.E., 1965. From acts to dispositions: the attribution process in
Boucher, J., Osgood, C.E., 1969. The pollyanna hypothesis. Journal of Verbal Learning
person perception. In: Berkowitz, L. (Ed.), Advances in Experimental Social
and Verbal Behavior 8 (1), 1–8.
Psychology, vol. 2. Academic Press, New York, pp. 219–266.
Boulding, W., Kalra, A., Staelin, R., Zeithaml, V.A., 1993. A dynamic process model of
Kärkkäinen, H., Piippo, P., Tuominen, M., 2001. Ten tools for customer-driven
service quality: from expectations to behavioral intentions. Journal of
product development in industrial companies. International Journal of
Marketing Research 30 (1), 7–27.
Production Economics 69 (2), 161–176.
Cassiman, B., Veugelers, R., 2006. In search of complementarity in innovation
Kanouse, D.E., 1984. Explaining negativity biases in evaluation and choice behavior:
strategy: internal R&D and external knowledge acquisition. Management
theory and research. Advances in Consumer Research 11 (1), 703–708.
Science 52 (1), 68–82.
H. Lee et al. / Electronic Commerce Research and Applications 13 (2014) 205–219 219

Katz, R., Allen, T.J., 2007. Investigating the Not Invented Here (NIH) syndrome: a Ramaswamy, V., 2009. Co-creation of value – towards an expanded paradigm of
look at the performance, tenure, and communication patterns of 50 R&D Project value creation. Marketing Review St. Gallen 26 (6), 11–17.
Groups. R&D Management 12 (1), 7–20. Reichwald, P., Piller, F., 2009. Interactive Worth Creation: Open Innovation,
Katz, E., Lazarsfeld, P., 1955. Personal Influence: The Part Played by People in the Individualization and New Form of Labor Division, 2nd edition. Gabler,
Flow of Mass Communications. Free Press, New York. Wiesbaden.
Lüthje, C., 2004. Characteristics of innovating users in a consumer goods field: an Reynolds, K.L., Harris, L.C., 2005. When service failure is not service failure: an
empirical study of sport-related product consumers. Technovation 24, 683–695. exploration of the forms and motives of ‘‘illegitimate’’ customer complaining.
Landon, E.L., 1980. The direction of consumer complaint research. Advances in Journal of Services Marketing 19 (5), 321–335.
Consumer Research 7 (1), 335–338. Richins, M.L., 1983. Negative word-of-mouth by dissatisfied consumers: a pilot
Laursen, K., and Salter, A. My precious technology: the role of legal appropriability study. Journal of Marketing 47 (1), 68–78.
strategy in shaping innovative performance. Working paper, Tanaka Business Ridout, M., Hinde, J., DemeAtrio, C.G., 2001. A Score test for testing a Zero-Inflated
School, Imperial College London, 2006. Poisson Regression Model against Zero-Inflated Negative Binomial Alternatives.
Lazarsfeld, P.F., Berelson, B., Gaudet, H., 1944. The People’s Choice; How the Voter Biometrics 57 (1), 219–223.
Makes Up His Mind in a Presidential Campaign. Sloan and Pearce, New York. Riecken, G., Yavas, U., 1986. Seeking donors via opinion leadership. Journal of
Lettl, C., Herstatt, C., Gemuenden, H.G., 2006. ‘Users’ contributions to radical Professional Services Marketing 2 (1–2), 109–116.
innovation: evidence from four cases in the field of medical equipment Rogers, E.M., 1962. Diffusion of Innovations. Free Press.
technology. R&D Management 36 (3), 251–272. Roman, E., 2010. Voice-of-the-Customer Marketing: A Revolutionary 5-Step Process
Lichtenthaler, U., 2011. Open innovation: past research, current debates, and future to Create Customers Who Care, Spend, and Stay. McGraw-Hill.
directions. Academy of Management Perspectives 25 (1), 75–93. Schindler, R.M., 1998. Consequences of perceiving oneself as responsible for
Liljander, V., Strandvik, T., 1993. Estimating zones of tolerance in perceived service obtaining a discount: evidence for smart-shopper feelings. Journal of
quality and perceived service value. International Journal of Service Industry Consumer Psychology 7 (4), 371–392.
Management 4 (2), 6–28. Schouten, J.W., McAlexander, J.H., 1995. Subcultures of consumption: an
MacCrimmon, K.R., Wagner, C., 1994. Stimulating ideas through creative software. ethnography of the new bikers. Journal of Consumer Research 22 (4), 43–61.
Management Science 40 (11), 1514–1532. Scott, L.J., Freese, J., 2006. Regression Models for Categorical Dependent Variables
Magnusson, P.R., 2009. Exploring the contributions of involving ordinary users in Using Stata. Stata Press Books.
ideation of technology-based services. Journal of Product Innovation Stoetzel, M., Wiener, M., and Amberg, M. Key differentiators of open innovation
Management 26 (5), 578–593. platforms – a market-oriented perspective. In Wirtschaftinformatik Proceedings,
Manchulenko, N. Applying Axiomatic Design Principles to the House of Quality. PhD 2011.
thesis, University of Windsor, 2001. Stokburger-Sauer, N., 2010. Brand community: drivers and outcomes. Psychology
March, J.G., 1991. Exploration and exploitation in organizational learning. and Marketing 27 (4), 347–368.
Organization Science 2 (1), 71–87. Strahilevitz, M., Myers, J., 1998. Donations to charity as purchase incentives: how
Martin, J., 1995. Ignore your customer. Fortune 131 (8), 121–126. well they work may depend on what you are trying to sell. Journal of Consumer
McAlexander, J.H., Schouten, J.W., Koenig, H.F., 2002. Building brand community. Research 24 (4), 434–446.
Journal of Marketing 66 (1), 38–54. Summers, J., 1971. Generalized change agents and innovativeness. Journal of
McAlexander, J.H., Koenig, H.F., Schouten, J.W., 2006. Building relationships of brand Marketing Research 8 (3), 313–316.
community in higher education: a strategic framework for university Teece, D., Pisano, G., 1994. The dynamic capabilities of firms: an Introduction.
advancement. International Journal of Educational Advancement 6 (2), 107–118. Industrial and Corporate Change 3 (3), 537–556.
Miller, J.A., 1977. Studying satisfaction, modifying models, eliciting expectations, Tzokas, N., Hultink, E.J., Hart, S., 2004. Navigating the new product development
posing problems, and making meaningful measurements. In: Hunt, H.K. (Ed.), process. Industrial Marketing Management 33 (7), 619–626.
Conceptualization and Measurement of Consumer Satisfaction and Varnali, K., Yilmaz, C., Toker, A., 2012. Predictors of attitudinal and behavioral
Dissatisfaction. School of Business, Indiana University, pp. 72–91. outcomes in mobile advertising: a field experiment. Electronic Commerce
Muniz Jr., A.M., O’guinn, T.C., 2001. Brand community. Journal of Consumer Research and Applications 11 (6), 570–581.
Research 27 (4), 412–432. Venkatraman, M.P., 1988. Investigating differences in the roles of enduring and
Mumford, M.D., Feldman, J.M., Hein, M.B., Nagao, D.J., 2001. Tradeoffs between ideas instrumentally involved consumers in the diffusion process. Advances in
and structure: individual versus group performance in creative problem Consumer Research 15 (1), 209–303.
solving. Journal of Creative Behavior 35 (1), 1–23. von Hippel, E., 1988. The Sources of Innovation. Oxford University Press, New York.
Myers, J.H., Robertson, T.S., 1972. Dimensions of opinion leadership. Journal of von Hippel, E., von Krogh, E., 2006. Free revealing and the private-collective model
Marketing Research 9 (1), 41–46. for innovation incentives. R&D Management 36 (3), 295–306.
Naveed, N., Gottron, T., Kunegis, J., and Alhadi, A. C. Bad news travel fast: a content- Vuong, Q.H., 1989. Likelihood ratio tests for model selection and non-nested
based analysis of interestingness on twitter. Available at http:// hypotheses. Econometrica: Journal of the Econometric Society 57 (2), 307–333.
www.websci2011.org/fileadmin/websci/Papers/2050_paper.pdf. Retrieved on Woodruff, R.B., 1997. Customer value: the next source for competitive advantage.
July 16, 2011 Journal of the Academy of Marketing Science 25 (2), 139–153.
Oliver, R.L., 1980. A cognitive model of the antecedents and consequences of Woodruff, R.B., Cadotte, E.R., Jenkins, R.L., 1983. Modeling consumer satisfaction
satisfaction decisions. Journal of Marketing Research 17, 460–469. processes using experience-based norms. Journal of Marketing Research 20 (3),
Patterson, P.G., Johnson, L.W., Spreng, R.A., 1996. Modeling the determinants of 296–304.
customer satisfaction for business-to-business professional services. Journal of Wyld, D.C., 2010. Speaking up for customers: can sales professionals spark product
the Academy of Marketing Science 25 (1), 4–17. innovation? The Academy of Management Perspectives 24 (2), 80–82.

You might also like