Lead users’ motivations to communicate in online communities

1

Lead users’ motivations to communicate

in online communities

Thomas Bollen

S2044242

t.a.h.j.bollen@student.rug.nl

MSc. Business Administration – Strategic Innovation Management

University of Groningen

June 2014

Supervisors: T.L.J. Broekhuizen

_{and R.A. van der Eijk}

Abstract

This study examines the motivations of lead users to communicate in online communities by using a survey sample of 271 respondents from an online community that focuses on computers and consumer electronics. The concept of the lead user is compared to that of the opinion leader. Two different statistical methods are used to test the model: through structural equation modelling in Amos and with the Preacher & Hayes PROCESS tool. Furthermore, the communication in- and outflows from users with a different level of technical expertise are examined. Results show that lead users are motivated by entertainment, sense of belonging and learning in using the online community and that lead userness is positively associated with making contributions to the online community. There are partial mediation effects through sense of belonging and through learning on respectively outgoing and incoming communication. Opinion leaders are also motivated by reciprocity, but lead userness and opinion leadership are not associated with gaining status. Users with a high level of expertise give relatively more advice and this advice is equally distributed among the other groups of users. The results also show that lead users consult primary information and visit other online communities more often than non-lead users. The findings of this study can be used to identify and stimulate lead users in their actions, and to get a better understanding in how knowledge is diffused in an online community. Implications and directions for future research are discussed.

2

Table of Content

2.4 Motivations to communicate in online communities ... 10

2.4.1 Entertainment (hedonic) ... 11

2.4.2 Sense of belonging (social integrative) ... 11

2.4.3 Learning (cognitive) ... 12

2.4.4 Status (personal integrative) ... 12

2.4.5 Reciprocity ... 12

2.5 Types of communication in online communities ... 12

2.6 Communication flows ... 13

2.6.1 Role of expertise of other users ... 13

2.6.2 Types of information sources ... 14

2.6 Conceptual model and development of hypotheses ... 15

2.6.1 Motivations ... 16

2.6.2 Communication behavior ... 16

3. METHODOLOGY ... 19

3.1 Context for empirical research ... 19

3.2 Data collection ... 20

3.3 Sample ... 20

3.4 Measurement instruments ... 21

3.4.1 Lead userness ... 21

3.4.2 Opinion leadership ... 22

3.4.3 Motivations to use the online community ... 22

3.4.4 Communication behavior ... 23

3.4.6 Direction of communication flows... 24

3.4.7 Usage of information sources ... 24

3.5 Measurement model ... 25

4. RESULTS ... 28

4.1 Structural model ... 28

3

4.2.1 Mediating effects of motivations ... 28

4.2.2 Differential effect of lead userness and opinion leadership ... 33

4.4 Communication flows ... 35

4.5 Usage of information sources ... 37

4.6 Overview results ... 38

5. DISCUSSION AND CONCLUSION ... 39

5.1 Limitations and directions for further research ... 42

6. REFERENCES ... 44

7. APPENDIX ... 48

4

1. INTRODUCTION

Users can be a valuable source of knowledge and expertise for new product development (Franke et al., 2006; Jeppesen & Molin, 2003). Their knowledge and opinions can be implemented in the development process which in turn may lead to more successful products. In the past the Research and Development process was closed and internalized. However, it became apparent that including users in the innovation process and by a more open collaboration, more novel and divers insights into customers’ needs could be gathered (Franke et al., 2006). The way in which a firm can identify and integrate users with a lot of expertise and knowledge can lead to new products and services that offer a competitive advantage (Mahr & Lievens, 2012). Von Hippel focused on this aspect and was the first to the introduce the phenomenon of the ‘lead user’ in 1986 (Von Hippel, 1986).

A lead user often has three characteristics: first, they experience the need for a given innovation earlier than the majority of the target market. Second, they are users who expect attractive innovation-related benefits from a solution to a problem (Franke et al., 2006; Morrison, Roberts & Von Hippel, 2000; Von Hippel, 1986). Thirdly, in order to have a good understanding of the possibilities and limitations of products or services, lead users also have a high degree of technical expertise (Spann et al., 2009).

Lead users do not only play a role in the fuzzy front-end phase, but also in the commercial phase: lead users are often early adopters of the product or service (Morrison et al., 2004). This is one of the reasons why lead users are so important in the diffusion process: they influence later adopters by word and deed. Their power comes in part from two principles of social influence: authority and social validation (Clark & Goldsmith, 2006). Authority (credibility) comes from expertise and social validation comes from the visibility of many adoptions, where others can see the early adopters exhibiting and using the new product. In an online context this is also possible by lead users giving their advice and opinion through forum posts or product reviews. Lead users can therefore act as catalysts of the diffusion of a product or service (Morrison, Roberts & Midgley, 2000).

5

Laursen, 2009; Mahr & Lievens, 2012; Spann et al., 2009). Jeppesen & Laursen's (2009) results also show that lead users are more likely to possess more relevant solution knowledge and thus make more contributions to the community compared to normal users. Due to the increase of internet usage and e-commerce during the past years, the involvement of lead users from online communities in the products development process becomes more important (Jeppesen & Laursen, 2009; Mahr & Lievens, 2012).

Most research on lead users has investigated field dependent variables (e.g. expertise, early adopting behavior and multi brand loyalty) (Franke et al., 2006; Ozer, 2009; Spann et al., 2009). Other research showed that field independent variables, such as the Big Five personality traits (Alers, 2011), which are apparent independent from the context, did not have a high explanatory power of lead userness. Past research studied these behaviors and characteristics, but not the motivations of lead users for determining their distinctive behavior. Since lead users are active in online communities, it is interesting to research what drives lead users to communicate the way they do.

The motivations of lead users to participate in online communities are unclear and inconclusive, although propositions are made that lead userness is associated with constructs such as altruism, a social identity, knowledge development and gaining status (Dholakia et al., 2004; Jeppesen & Laursen, 2009; Mahr & Lievens, 2012). Lead users are attracted to visiting and to making contributions to online communities because it fulfils their needs to maintain and update their product related knowledge, to get insight in developments in the market and to help other members (Jeppesen & Laursen, 2009; Mahr & Lievens, 2012). Lead users can therefore not only be seen as users that absorb information, but also as gatekeepers, linking their community to the world at large (Jeppesen & Laursen, 2009). The aspect of motivational reasons to participate in online communities is treated in literature, but it is often not focused on the lead user (Jeppesen & Laursen, 2009; Mahr & Lievens, 2012). Furthermore, research on lead users has primarily focused on communication channels such as face-to-face workshops or specific electronic tools, thus further research is needed what their role is within an online community context (Mahr & Lievens, 2012).

6

communicate with other lead users or do they also communicate with members that have a lower status/reputation?

The purpose of this study is to research these ambiguities. How are lead users motivated to participate in a community and share knowledge? How do lead users interact and communicate in the communities they participate in and with whom? This leads to the following research question: how do lead users’ motivations influence their communication behavior in online communities?

When the answer to this question becomes more clear, then lead users could be approached more properly according to their specific motivational needs for the involvement in new product development processes or to use them as catalysts for the diffusion process of products or services. To better understand the concept of the lead user and distinguish it from other concepts, this study will compare lead userness to opinion leadership, a strongly related concept. An opinion leader has a great potential of influencing others in their purchasing behavior (Iyengar, 2011; Rose & Kim, 2011; Shoham & Ruvio, 2008). Due to inconsistent results and the potential overlap of the concepts, there is a discussion about whether opinion leadership is distinct from lead userness (Ozer, 2009), that it is a dimension of lead userness (Iyengar, 2011; Goldsmith & Witt, 2005; Spann et al., 2009) or that opinion leadership stems from a high degree of lead userness (Morrison, Roberts & Von Hippel, 2000; Morrison et al., 2004). In this study we will treat opinion leadership as a separate variable, next to lead userness, in order to compare lead users with opinion leaders in their motivations and communication behavior in online communities. This provides insights into whether opinion leaders are for example more motivated by the gained status, whereas lead users are more motivated by the enjoyment of helping people and learning new things. These differences in motivational reasons can lead to different outcomes in communication behavior.

This study consists out of two parts: the first part focuses on the motivational reasons of lead users and opinion leaders and their effects of communication behavior in the online community. The measurement items of this part are mainly based on previous research. The second part is more explorative and will go further in-depth into the communication behavior of users; how users differ in incoming and outgoing communication and with whom, based on expertise, they communicate.

7

2. THEORETICAL BACKGROUND

2.1 Lead userness

Developing new products requires accurate insights in the needs of customers and the product usage. The outside-in strategic view, which is based on developing from the customer needs, is a very common term for this user integration (Poetz & Schreier, 2012; Von Hippel, 1986). The lead users method, introduced by Von Hippel (1986), is an integration process that has conceived notable attention during the past two decades.

The idea behind this is that these special kind of users, the leading edge user, is involved in the fuzzy front end phase of the new product development process, where customer needs are not very clear. Companies can learn the needs and solutions from these leading edge users, assuming these are representable for the future mainstream market, which in turn will lead to better products and firm performance.

An additional point is that lead users tend to adopt products early on and are willing to share experiences (Von Hippel, 1986; Jeppesen & Laursen, 2009). They can also function as a catalyst of the diffusion process of innovations. A lead user can influence the buying behavior of other users, by telling about the product, or by providing signals to potential adopters (observational learning).

A lead user distinguishes himself from other users with mainly three main characteristics (Von Hippel, 1986; Morrison, Roberts & Von Hippel, 2000; Franke et al., 2006; Spann et al., 2009).:1

1. They experience the need for a given innovation earlier than the majority of the target market (ahead of market trend);

2. They are users who expect attractive innovation-related benefits from a solution to a problem (high expected benefits);

3. They have a high degree of technical expertise with products and services within the specified area of interest (technical expertise).

Sometimes ‘level of innovation’, the degree in which users perform innovative activities, is also included as an characteristic, while others see this aspect as an behavioral outcome of the three mentioned characteristics (Morrison et al., 2004).

The concept of lead userness is mainly based on Von Hippel’s ideas. The variables, being “ahead of the market”, “high expected benefits” and “technical expertise”, can be seen as continuous variables rather than binary ones; i.e. a person has a degree over lead userness and is not simply a lead user or not.

1 _{The original concept of the lead user consisted of two characteristics, ahead of market trend and high expected benefits.}

8

Ahead of market trend refers to lead users’ sensing a certain need before the majority of the market requires it (Franke & Shah, 2003; Franke et al., 2006; Luthje & Herstatt, 2004; Von Hippel, 1986). However, not all users facing new needs are lead users, as Luthje & Herstatt (2004) argues. Lead users must also realize that most customers in the market will face the need in the future. This is an important condition, since a manufacturer that aims to develop profitable products for tomorrow’s market must have a profitable potential market in order to have a commercial success (Luthje & Herstatt, 2004).

High expected benefits is about the innovation likelihood of an user. It reflects the investment users are willing to make in order to obtain the benefit, the dissatisfaction with the current offerings and the speed of adoption (Bilgram et al., 2008; Franke et al., 2006; Luthje & Herstatt, 2004).

Both original lead user characteristics rather concentrate on motivational qualities. The actual product-related abilities and the knowledge of users are not explicitly included in the original lead user criteria. Therefore several researchers have argued to include the technical expertise characteristic, since it is a prerequisite for lead userness, to have a more comprehensive view of the lead user (Spann et al., 2009).

Technical expertise refers to the degree of knowledge and experience the person has in a particular field (Ozer, 2009). Technical expertise refers to the ability of an user to actually accomplish modifications/changes to products of the particular field (Franke et al., 2006). The high degree of expertise enhances lead userness, and helps lead users to quickly gain new knowledge and determine the value of it; i.e. the absorptive capacity of these users is higher due to the higher expertise (Cohen & Levinthal, 1990; Mahr & Lievens, 2012).

Lead users possess great product knowledge about products, brands and industries. In order to maintain this knowledge they will participate in discussions with other users (Mahr & Lievens, 2012). These discussions are often about new products and developments in the market. Since lead users tend to have a high technical expertise (Franke et al., 2006) and adopt products early, like Jeppesen & Laursen (2009) argue, it can be proposed that they are able to provide solutions better, more often and faster than other users.

9

2.2 Opinion leadership

One of the main players in the market that accelerates the diffusion process of products is the opinion leader, a player that has a major role in the activation of diffusing networks because the opinion of opinion leaders can strongly influence others in their adopting behavior. (Burton & Khammash, 2010; Iyengar, 2011; Myers & Robertson, 1972; Shoham & Ruvio, 2008). This type of user is characterized by his/her knowledge, social influence, innovativeness and interpersonal factors (Kratzer & Lettl, 2009; Gnambs & Batinic, 2011; Goldsmith & Witt, 2005; Morrison, Roberts & Midgley, 2000).

Some researchers, like Myers & Robertson (1972), say that opinion leaders are not the main innovators and that there is a moderate, rather than strong, relationship opinion leadership and innovative behavior. Goldsmith & Witt (2005) argue otherwise and say that there is strong relationship.

There is no consensus whether opinion leadership is distinct from lead userness (Ozer, 2009), that it is a dimension of lead userness (Iyengar, 2011; Goldsmith & Witt, 2005; Clark & Goldsmith, 2006; Spann et al., 2009) or that opinion leadership stems from a high degree of lead userness (Morrison, Roberts & Von Hippel, 2000; Morrison et al., 2004). Tolba & Mourad (2011) combine the two concepts and based on the degree of lead userness and opinion leadership, they formed four classifications of users: champions, promoters, inventors and followers.

2.3 Online communities

An online community is a place where users with a similar interest can interact by discussing various topics. These communities are often focused on a particular common interest of the users; e.g. cars, a certain sport, consumer electronics or even clothes. The concept of online communities appeared quickly after the emergence of the internet. At that time (± 2000) the user base consisted mostly out of very active internet users that were discussing the technical aspects of products. However, due to increase of internet usage and accessibility of the internet, many other consumers used the internet for entertainment, but also to solve problems that they encountered or to get advice with their buying decisions (Benlian et al., 2012; Hennig-Thurau et al., 2004). A popular example is the use of expert and consumer reviews that are available on the internet thereby consumers do not only rely on word-of-mouth of their direct environment (Lee et al., 2008; Zhu & Zhang, 2010). These new users also consult online communities when they face problems. In sum, the online communities gained a broader target audience that consisted of more mainstream users.

10

knowledge, the role of the community is pointless since there is no interaction between the users. There is no point in expressing knowledge when no one is interested in it. Likewise, it is impossible to seek knowledge when there is no one available that shares the required. Knowledge sharing in these online communities is important as it sustains learning, but also because it supports processes of technological innovation, adoption and usage (Jeppesen & Laursen, 2009).

2.4 Motivations to communicate in online communities

The frequency of visits to the community website and the degree in which users actively participate in the community by interacting and communicating with other members of the community, depends in a certain degree in how the user is motivated (Cheung & Lee, 2012; Dholakia et al., 2004; Füller, 2010; Nambisan & Baron, 2009; Yang & Lai, 2010). Other factors include, for example, the ease of use and availability of the website (Lee et al., 2014) and socio-demographic and situational factors (Wasko & Faraj, 2000).

In online communities users can have different motivational reasons to participate in and contribute to the online community. Motivations to participate in and share knowledge in online environments has already been discussed in the literature (Füller et al., 2004; Füller, 2010; Nambisan & Baron, 2009; Wasko & Faraj, 2000; Wasko & Faraj, 2005). These motivations are mostly based on a classification of intrinsic and extrinsic motivation. Intrinsic motivation refers to “the internal satisfaction received from the process of performing behaviors” (Dholakia et al., 2004; Füller, 2010; Yang & Lai, 2010), whereas extrinsic motivation is about “a goal-oriented motivation that refers to performing an activity in anticipation of obtaining a return such as pay or reputation” (Dholakia et al., 2004; Füller, 2010; Yang & Lai, 2010).

Nambisan & Baron (2009) studied the motivational reasons for users to participate in co-creation by making use of the Uses and Gratifications Framework, which treats the motivation by the concept of benefits (Cognitive or Learning Benefit, Social Integrative Benefit, Personal Integrative Benefit, Hedonic Benefit). These benefits have the intrinsic and extrinsic motivations as a basis, but are more thoroughly ordered into groups.

11

Taking the four motivations of Nambisan & Baron (2009) and adding the reciprocity motivation (Cheung & Lee, 2012; Wasko & Faraj, 2000; Wasko & Faraj, 2005; Wiertz & de Ruyter, 2007) it is possible to cover all segments in the individual-social "individual orientation" (directed to self) and "social orientation" (directed to others), and intrinsic-extrinsic spectrum (see Table 1). This study identifies the motivations in Table 1 to be important. The terms between parentheses are the terms used by Nambisan & Baron (2009) for the same construct.

Individual Social

Intrinsic Entertainment Learning

Sense of Belonging

Extrinsic Reciprocity Status

Table 1- motivation spectrum

2.4.1 Entertainment (hedonic)

Entertainment is one of the motivations whey people use an online community. Users can be motivated because they find it pleasurable in using the community by having interactions with fellow members, discussing the latest news and helping others. This entertainment motivation can also be described as a hedonic benefit (Nambisan & Baron, 2009). A common example is that users share their opinions about products and purchasing decisions because others have a need of it. Users obtain intrinsic enjoyment and satisfaction by helping others (Hennig-Thurau et al., 2004; Lampe et al., 2010).

2.4.2 Sense of belonging (social integrative)

The motivation of ‘sense of belonging’ mainly stems of the Theory of Social Identity, which was introduced in the seventies by Tajfel (1974), and is about the behavior of members within a group; why and how they interact with other members.

12

2.4.3 Learning (cognitive)

Learning and gaining knowledge about a product or industry can be another reason to visit, participate in and contribute to an online community. In the context of an online community that focuses on a particular product (category), the learning variable refers to cognitive benefits that reflect product-related learning. This allows users to obtain more knowledge and a better understanding about the products, their underlying technologies and their usage (Füller, 2010; Nambisan & Baron, 2009). With extended knowledge users are more able to create value for their peers and for themselves.

2.4.4 Status (personal integrative)

Social enhancements in reputation or status and the achievement of a sense of self-efficacy yield personal integrative benefits (Nambisan & Baron, 2009). Like Cheung & Lee (2012) mention, some users share and contribute their knowledge because they want to gain an informal recognition and they want to establish themselves as experts in the field. Their status or reputation is important to them and they share knowledge to gain or maintain the acquired status. Status can be seen as a more extrinsic motivation in the context of online communities.

2.4.5 Reciprocity

Related to the personal integrative benefit of status, reciprocity is also more extrinsically oriented. Reciprocity refers to the expectation that the sharer of the knowledge will get something in return; that there is a balance between the amount of knowledge that is shared and that is gained (Cheung & Lee, 2012; Franke & Shah, 2003; Wasko & Faraj, 2000; Wiertz & de Ruyter, 2007). The results of Franke & Shah (2003) show that the expectation of reciprocity is important in communities where innovation-related information is revealed freely with users that do not know each other. Wasko & Faraj (2005) also argue that that knowledge sharing in electronic networks of practice is facilitated by a strong sense of reciprocity.

2.5 Types of communication in online communities

The participation in and contributions to an online community can differ between users. Firstly, there is difference between outgoing and incoming communication within knowledge sharing clusters (also referred to as inbound and outbound by Giuliani & Bell (2005) or ‘Get information’ and ‘Give information’ by Lampe et al. (2010)). Outgoing communication is about sharing knowledge and opinions, whereas incoming communication is about seeking knowledge and opinions.

13

Objective knowledge mainly stems from examining factual attributes of a product or service and interpreting the specifications. Giving an opinion is more focused on subjective attributes that stem from a personal preference. Both types of knowledge can be related to product-related experience; e.g. an user is dissatisfied with a particular brand due to bad experiences in the past.

The similarity between the two is that in both cases knowledge is shared, but they also differ in the fact that giving advice is often a specific response that another user faces (Mahr & Lievens, 2012). This can be for example giving advice on a buying decision of the other user or that the other user has a problem with his product that needs to be fixed. On the forums this is clearly present since users open topics and address their problem and request help. Besides these problem-based topics, other topics are active that are about more general developments in the market in which more often an opinion is given by the user. This type of communication is less specific and is aimed at the general public, rather than a specific case.

2.6 Communication flows

2.6.1 Role of expertise of other users

In an online community the level of expertise between users can differ. Users with a high level of expertise are therefore sometimes confronted with problems that are quite simple or were even treated multiple times in the past. The study of Jeppesen & Laursen (2009) researches this by stating that given that an individual holds the relevant knowledge to the problem in question, since the user has encountered the problem in the past or he has the problem solving capacity, the cost of providing an answer to a question becomes lower.

14

2.6.2 Types of information sources

Apart from the active incoming communication within the online community, there are also more passive forms of knowledge seeking and an online community is not the only place where users can gather relevant information. For example, one could also consult the specifications of a product at the website of the manufacturer or visit other websites (Mahr & Lievens, 2012; Spann et al., 2009). Options for gathering information via online sources include, but are not limited to:

 Consult primary information of the manufacturer

 Consult expert reviews

 Visit multiple communities or websites to get better insights

 Search the archive of a forum for relevant information

15

2.6 Conceptual model and development of hypotheses

In order to answer the relevant research questions I will setup several hypotheses that are based on the discussed literature. A conceptual model can be seen in Figure 1, where the relationships between variables are visualized.

The model assumes that the person-specific traits (lead userness and opinion leadership) have a direct effect on the communication behavior in community websites, and an indirect effect on the communication behavior through the motivations. The communication behavior consists out of outgoing and incoming communication behavior.

MOTIVATION Entertainment Sense of belonging Learning Reciprocity Status COMMUNICATION BEHAVIOR Outgoing - Giving advice

Outgoing - Giving opinion

Incoming - Seeking advice Incoming - Seeking opinion

Incoming - Seeking news and updates Lead

userness / opinion leadership

Figure 1 - Conceptual model

16

2.6.1 Motivations

Lead users want to maintain product related knowledge, get insight in developments in the market and help other members of the community. Lead users tend to be more active in online communities, due to their interest in new innovative products and services (Mahr & Lievens, 2012). Lead users are intrinsically motivated to visit, participate in and contribute to an online community. These intrinsic motivations include entertainment, a sense of belonging and learning. It is proposed that the extrinsic motivation variable of status is not significantly related to lead userness, because lead users do not highly value this motivation (Jeppesen & Laursen, 2009; Wasko & Faraj, 2005). Reciprocity is of importance for lead users, since they want a sustainable environment where they can assimilate new knowledge (Jeppesen & Laursen, 2009; Mahr & Lievens, 2012). This requires that others also share their knowledge. This will lead to the following hypotheses:

H1a-d: Lead userness is positively associated with the motivations a) entertainment, b) sense of belonging, c) learning and d) reciprocity, to visit, participate in and contribute to an online community.

2.6.2 Communication behavior

Lead users have a high level of expertise (Mahr & Lievens, 2012) and it is therefore argued that they, compared to non-lead users, are more likely to provide solutions and participate in discussions about products and services (Jeppesen & Laursen, 2009). Lead userness will therefore have a positive effect on ‘giving advice’ and ‘giving opinion’.

H2a-b: Lead userness is positively associated with a) giving advice and b) giving opinion.

Because of their high level of expertise, the problems of lead users will probably be very specific and more complex than the problems of other users (Ozer, 2009; Mahr & Lievens, 2012). Additionally, they have a greater tendency to first try to solve the problem themselves before they ask for help. The research of Clark & Goldsmith (2006) shows that lead users appear to be less attentive observers of others’ product and brand choices and rely less on the opinions of others regarding their purchases. We argue that the net result will be a significant negative association between lead userness and seeking advice and seeking opinions of other users.

17

H2c-e: Lead userness is negatively associated with c) seeking advice2 and d) seeking opinion, but positively associated with e) seeking news and updates.

Mediating role of motivations

Besides the direct effects, which are stated at the previous hypotheses, I also propose mediation effects. The motivation variables will then act as mediator variables and lead userness will have an effect on communication behavior through these motivation variables. Complementary to H1a and H1b, entertainment and sense of belonging both have a significant positive effect on giving advice and giving opinion. Users that are entertained and have a sense of belonging by visiting and participating in the community are more active with outgoing communication behavior.

H3a-b: Lead userness is positively associated with giving advice and giving opinion through a) higher entertainment and b) a higher sense of belonging.

Like hypothesis H2e, it is proposed that lead userness is positively associated with seeking news and updates, through learning. Learning about new products and technologies is an outcome of the continual high expected benefits, to see whether new products can fulfil the unfulfilled needs of the lead user. They do this by actively seeking news and updates about products and technologies (Mahr & Lievens, 2012).

H3c: Lead userness is positively associated with seeking news and updates through learning.

Differences of lead userness with opinion leadership

Opinion leadership will be treated as a separate construct in this study. This will enable us to compare lead userness with opinion leadership and see if there are differences in the motivations and the communication behavior. The following hypotheses will focus on the expected key differences in this study. I argue that opinion leaders, given their extraversion nature, attain more value to their status in the community, which leads to that opinion leadership is positively associated with the motivation ‘status’ variable (Kratzer & Lettl, 2009; Rose & Kim, 2011). Rose & Kim (2011) also argued that “opinion leaders seem motivated to influence others, a motivation that should be associated with the motivation to occupy a dominant social position”. Opinion leaders are also positively associated with the entertainment, sense of belonging, learning and reciprocity, like lead users.

18

H4a-e: Opinion leadership is positively associated with the motivations a) entertainment, b) sense of belonging, c) learning, d) status and e) reciprocity to visit, participate in and contribute to an online community.

Like the name suggests, opinion leaders are more active with proclaiming their opinion (Iyengar, 2011). Lead users however are more interested in providing explicit, factual information (Mahr & Lievens, 2012). We therefore propose:

H5a: Lead userness is more strongly positively associated with giving advice than opinion leadership. H5b: Opinion leadership is more strongly positively associated with giving opinion than lead userness.

19

3. METHODOLOGY

This research will follow the design of theory testing empirical cycle. This type of research is chosen because there is already done some research on the discussed topics, but the topics are either not related to each other or the outcomes are inconclusive. Therefore validation through empirical testing, with large scale data, is needed. This study is focused on the role of the lead user in online communities. Previous studies that focused on the lead user or the motivation of users in online communities, used (online) questionnaires to collect the data. In order to answer the research question of this study - how do lead users’ motivations influence their communication behavior in online communities? – it is necessary to collect data in a similar way to connect the topics of the lead user, the motivations to communicate, and the communication behavior in online communities.

3.1 Context for empirical research

In order to test the proposed hypotheses, this study collects data from the online community Tweakers through an online questionnaire. This paper focuses on the motivations and behavior of lead users in online communities. Hence, a suitable online community must be selected where a vast amount of lead users and opinion leaders, but also unexperienced users are active.

The online community Tweakers - http://tweakers.net/ - (and forums, which are called ‘Gathering of Tweakers’) focuses on information technology (products) in general. Computer hardware, consumer electronics and peripherals are key subjects that are extensively discussed.

This research uses this community for the large scale data collection due to several reasons: First, the community is large as it has over 500,000 members that generate three million page views on the website and 500,000 page views on the forums. Several other questionnaires that were posted on the forums in the past often yielded sufficient users that filled in the questionnaire, because the members are in general seen as pro-active users and willing respondents that provide useful feedback and are willing to improve the community. Newcomers and experienced users are both active within this community, providing a varied population. Second, a large portion of the discussions is about new products, innovations and modifying/improving existing products. Lead users are highly interested in these subjects, thereby chances are higher that a large amount of lead users fills in the questionnaire.

20

information (specifications) and subjective information (opinions), 2) the users often give or seek advice and/or opinions about these products and 3) because these are the major topics on Tweakers, thereby increasing the chance that relevant users fill in the questionnaire.

3.2 Data collection

An online questionnaire was used to collect the data from the users of the Tweakers community. A topic with a link to the questionnaire was posted on the forums of Tweakers in March 20143_{. With} the help of the forum administrators it was possible to promote visibility of the questionnaire/study by highlighting the topic in several sections of the forums. In this way awareness of the study was created, yielding a higher amount of respondents. To stimulate participation even further, it was proposed to donate €50 to a specific charity fund chosen by Tweakers when 200 users would fully complete the questionnaire. The topic was checked daily and the community was given status updates about the number of respondents that already had completed the questionnaire. The questionnaire was available for respondents for two weeks.

3.3 Sample

The final net sample contains 271 respondents with an average age of 28 years (SD=10.2), of which 99.3 percent are men. Early and late respondents were compared to test for nonresponse bias, but no significant differences in the variables of interest were found.

The initial topic generated 3751 page views which results in a net response rate of 7.2%. The amount of respondents is quite satisfactory considering the time investment that must be made in completing the questionnaire. Figure 2 shows that the registration years of the members, the year in which a member registered at the community, of the sample is evenly distributed. The distribution of the education degrees of the respondents are shown in Table 2.

3_{http://gathering.tweakers.net/forum/list_messages/1585694}

EDUCATION

None 1.8%

High school 19.2%

Community College (mbo) 14.8% University of Applied Sciences (hbo) 35.8% University (wo) 25.1% Doctor / PhD or higher 1.5% 0% 2% 4% 6% 8% 10% 12%

Registration year

21

3.4 Measurement instruments

All questions in this study use a five point Likert scale and where available empirically validated scales were used. Respondents needed to give a score from highly disagreeing to highly agreeing (1 = strongly disagree and 5 = strongly agree). See the appendix for the questionnaire and item labels. The measurement items are adjusted to the research context. To improve validity and verify that the questions are well structured and clear to the audience, a pre-test was conducted. This pre-test was held among a group of five crew members of Tweakers. They can be seen as experts in the field online communities and know familiar habits within the Tweakers forum. The outcome was that some sentences were simplified and some became shorter so they would be more clear and better understandable for the respondents.

3.4.1 Lead userness

This study follows the way of thinking of Spann et al. (2009) and measures lead as a reflective, second-order factor, consisting of three first-order constructs: ahead of market trend, high expected benefits and technical expertise

Ahead of market trend was measured by asking users to what extent they 1) find out about new computers / computer parts, consumer electronics and peripherals earlier than others, 2) find out about solutions to problems with computers, consumer electronics and peripherals earlier than others, 3) tested prototype versions of new computers, consumer electronics and peripherals for manufacturers and 4) have improved and/or developed new features for computers, consumer electronics and peripherals (Franke & Shah, 2003; Franke et al., 2006; Ozer, 2009).

High expected benefits consisted of three questions, if users 1) were dissatisfied with some pieces of commercially available equipment, 2) had already encountered problems with equipment which could not be solved with the manufacturer's conventional offerings and 3) were constantly looking for improved computers, consumer electronics and peripherals (Franke & Shah, 2003; Franke et al., 2006; Ozer, 2009).

22

3.4.2 Opinion leadership

Opinion leadership was measured by three items 1) whether users have the impression that they are regarded by friends and acquaintances as a good source for tips and advice, 2) if friends and acquaintances ask the respondent for advice before they buy new products and 3) if people in the social circle of the user frequently act upon their advice. These measurement items are based on the work of Gnambs & Batinic (2011) and Jadin et al. (2013), that develop and test measurement items for the construct of opinion leadership.

3.4.3 Motivations to use the online community

Entertainment was measured by asking to what degree users used the website and participated in discussions 1) for entertainment and stimulation of the mind, 2) for fun and pleasure, 3) to spend some enjoyable and relaxing time and 4) to derive enjoyment from problem solving, idea generation, etc. (Cheung & Lee, 2012; Nambisan & Baron, 2009; Yang & Lai, 2010).

Sense of belonging was measured by asking to what degree users 1) were attached to the community, 2) enhance the sense of belonging with the community, 3) appreciate the friendships they have with other members and 4) see themselves as part of the community (Cheung & Lee, 2012; Füller, 2010; Nambisan & Baron, 2009).

Learning was measured by asking to what degree users used the website and participated in discussions to 1) enhance their knowledge about products and their usage, 2) obtain solutions to specific product-usage related problems, 3) enhance their knowledge about advances in product, related products, and technology and 4) to be up to date with developments of new products and technologies in the computers, consumer electronics and peripherals industry (Füller, 2010; Nambisan & Baron, 2009).

Status was measured by asking to what degree users used the website and participated in discussions to 1) share knowledge in the community to improve their image, 2) earn respected from others in the community, 3) improve the professional status in the community and 4) enhance their status/reputation as product expert in the community (Cheung & Lee, 2012; Jeppesen & Molin, 2003; Nambisan & Baron, 2009; Yang & Lai, 2010).

23

3.4.4 Communication behavior

In this study the communication behavior will be divided into two parts: 1) outgoing and 2) incoming communication (Giuliani & Bell, 2005).

Outgoing communication will consist out of 1) giving advice and 2) giving an opinion. To make the difference between advice and opinion clear in the survey, definitions and examples were given. Advice was defined as an objective rating of product feature, brand or technology. The emphasis lies on giving factual information, like specifications of a product, brand or technology. An opinion was proposed to be a subjective rating of product feature, brand or technology.

1) Giving advice - E.g. giving advice on what computer hardware must be bought, which software must be used in a particular case, giving feedback on programming code etc. This type of advice is often trigged by a question of another user that has the problem in which others can respond.

2) Giving opinion - E.g. discussions about news, giving opinion about new products (product reviews). Not necessarily triggered by other users; more self-initiative. Incoming communication will consist of three parts: 1) seeking advice, 2) seeking opinion and 3) seeking news and updates. These types are ways to gather new knowledge, but they differ in how the quest for knowledge is initiated.

Seeking advice or help is a direct action taken to solve a problem or get advice. This often means that the user opens a new topic on the forums, that she/he asks his question in another topic or that she/he contacts other users directly. The latter one is more focused on being up to date and is undirected; the gained knowledge might be useful in the future and the user is more interested in the new knowledge, rather than the goal of the knowledge.

1) Seeking advice - E.g. when a user faces a problem he will ask other members of the forum to help him. See giving advice for examples.

2) Seeking opinion - E.g. seeking for the opinions of other users.

24

3.4.6 Direction of communication flows

The users were presented an illustration of a pyramid (see figure 3) that represented the online community Tweakers with a subdivision based on the expertise level of users. In this study the level of ‘expertise’ is used, since it is a clearer concept to the respondents compared to lead userness since lead userness consists out of three separate constructs. The level of expertise is also used in other research as an indicator of the hierarchy within an online community (Dholakia et al., 2004; Clark & Goldsmith, 2006). The pyramid was divided in five layers with each layer representing a group of users that had approximately the same level of expertise. Notable is that the percentages of each layer were not equal: it was argued that there are relatively less people with a very high level of expertise that belong to the top (A) than there are people that belong to groups that have a lower expertise level (B-E). In other words, on the top the experts and at the bottom the inexpert people. The percentages of each layer were chosen in such a way that a pyramid form was created and are not based on other literature. The size of the parts of the pyramid do not represent the percentages. The respondents must then indicate to what group they think they belong, so self-reported, and give a degree on a five point Likert scale (1 = little , 5 = much) on how much they 1) give advice to each level and 2) seek advice from each level.

Figure 3 - Expertise pyramid

3.4.7 Usage of information sources

25

3.5 Measurement model

Standard procedures were used to test the reliability and validity of the measurement items and a confirmatory factor analysis was performed. Some items had low loadings on their respective constructs which resulted in an average variance extracted (AVE) of the construct that was below the threshold of .50 (Hair, Black, Babin & Anderson, 2010). These items included INSN3, EH4, INSA3 and LU3, and were therefore dropped.

Further analysis showed that the items of the constructs "incoming - seeking advice" and "incoming - seeking opinion" were highly loading on one and the same latent factor, jeopardizing the discriminant validity of the constructs. Apparently the respondents did not see these as two significantly different variables, whereas there was a significant different between the outgoing communication behavior. To avoid the situation of highly correlated factors, the construct "incoming - seeking advice" (INSA) was dropped from the study since it only consisted out of two items. The final measurement model provided the following results: χ²/df = 1,703 (χ² = 1230; df = 721); CFI = 0.911; GFI = 0.817; RFI = 0.787 ; TLI = 0.899 ; RMSEA = 0.051, which indicates that the data fits the model well. The correlation matrix is shown in Table 3. The following guidelines of Hair, Black, Babin & Anderson (2010): to test reliability the composite reliability of each variable must be above .70. For convergent validity the composite reliability must exceed the average variance extracted and the average variance extracted of each variable must be above the threshold of .50. Discriminant validity is met when the average variance extracted is greater than the median shared variance and average shared variance. Table 3 shows that all requirements have been met. Each item loads significantly (p < .01) on its assigned factor and therefore proves to have both convergent validity and unidimensionality. There was sufficient evidence for discriminant validity, since the square root of the average extracted variance (AVE) of each latent construct always exceeds the correlations of the construct with any other construct in the model (Fornell & Larcker, 1981). There were no correlations between variables that exceeded the threshold of .80, indicating that the constructs are sufficiently distinct (Bagozzi et al., 1991).

26

27 Table 3 – Correlation Matrix.

CR = Composite Reliability, AVE = Average Variance Extracted, MSV = Maximum Shared Variance, ASV = Average Shared Variance (ASV). LU = Lead Userness, OL = Opinion Leadership, EH = Entertainment, SB = Sense of Belonging, LR = Learning, ST = Status, RC = Reciprocity, OUTGA = Outgoing Giving Advice, OUTGO = Outgoing Giving Opinion, INSO = Incoming Seeking Opinion, INSN = Incoming Seeking News and Updates.

Numbers in bold present the square root of the AVE. Standard errors were derived from bootstrapping with 500 samples. ** = p<0.01; * p<0.05

CR AVE MSV ASV LU OL EH SB LR ST RC OUTGA OUTGO INSO INSN

28

4. RESULTS

4.1 Structural model

The structural model with lead userness as the independent variable had the following model fit: χ²/df = 2.187 (χ² = 1392; df = 637); CFI = 0.856; GFI = 0.784; RFI = 0.742 ; TLI = 0.841 ; RMSEA = 0.66. Although several fit indices are below the threshold of .8 and below the desired value of .9, the fit seems reasonable. The results of the structural model can be found in table 4. The control variables gender, education level, age and time of membership did not have a significant effect on lead userness.

A structural model with both lead userness and opinion leadership as independent variables simultaneously was also created. This model had the following model fit: χ²/df = 2.162 (χ² = 1613; df = 746); CFI = 0.849; GFI = 0.770; RFI = 0.730 ; TLI = 0.834 ; RMSEA = 0.066. The model fit is similar to the model fits with the constructs tested isolated. We choose to perform isolated tests because opinion leadership and lead userness are highly correlated (near .6, see table 3), which could lead to multicollinearity and thus can influence the strength and significance level of relationships.

4.2 Hypotheses testing

Support is found for hypothesis H1a (β = 0.33, p<.001), H1b (β = 0.22, p<.01) and H1c (β = 0.34, p<.001), meaning that lead userness is significantly positively associated with respectively entertainment, sense of belonging and learning. No significant relationships were found between lead userness and status (β = 0.12, NS), and lead userness and reciprocity (β = 0.09, NS) (see Table 4). Hypothesis H1d is therefore not supported.

Further support is found for hypotheses H2a (β = 0.42, p<.001) and H2b (β = 0.39, p<.001), which means that lead userness is directly positively associated with giving advice and giving opinion. ‘Seeking advice’ was dropped when checking for validity and reliability of the measurement model. Therefore hypothesis H2c could not be tested. H2d, which stated that lead userness is negatively associated with seeking opinions, was not supported (β =-.010, NS). H2e (β = 0.38, p<.001) is also supported, indicating that lead userness is positively associated with seeking news and updates.

4.2.1 Mediating effects of motivations

29

method of Mathieu & Taylor (2006) for determining whether there are full mediation effects, partial mediation effects or indirect effects. In figure 4 the coefficient paths are shown in case of the inclusion of a mediator. Full mediation entails that without the mediator there is a significant relationship between the independent variable and dependent variable (c), but with the introduction of the mediator this path becomes insignificant (c’). Partial mediation means that with the introduction of the mediator the path from the IV to the DV is still significant (c’). Indirect effects are present when there is no significant relationship between the IV and the DV (c’), but that the paths from the IV to the mediator (a) and from the mediator to the DV (b) are significant.

Figure 4 - coefficient paths

Tests with bootstrapping were performed in Amos 21 to investigate the significance of the indirect effects. This happened in an isolated setting, with only one communication behavior variable at a time, in order to make comparisons of the results from the other technique, the PROCESS tool. With the bootstrapping method we can get indirect effects, which are the effects from the independent variable on the dependent variable through the mediator (Mathieu & Taylor, 2006).

30

test, works rather well, but that it is rarely used as a definitive test (Fritz et al., 2012). Therefore the bootstrapping technique is used to assess the indirect effects, since it takes error into account.

Preacher & Hayes PROCESS tool

Another way to test the mediation effects is by using the fairly new PROCESS macro tool from Preacher & Hayes (2014). This tool is plugin for SPSS and according to the authors, this macro/tool is “far superior to SOBEL, as it allows for more than one mediator and adjusts all paths for the potential influence of covariates not proposed to be mediators in the model” (Preacher & Hayes, 2014).

Prior to testing the variables in SPSS using the PROCESS tool, composite scores are calculated. These composite scores were computed in SPSS by taking the sum of all the values of the items of each construct and then divide them by the number of items; i.e. the mean. One of the disadvantages of the PROCESS tool is that it does not has the ability to explicitly account for measurement error, contrary to a structural equation program like Amos. The use of composite values can lead to different outcomes. Furthermore, this tool can only predict the indirect effects on one dependent variable at a time. Therefore the model in Amos is also tested with on dependent variable at time, so the results can be compared. Similar to the tests in Amos, the PROCESS tool is set to 1000 bootstraps with a 95% confidence interval.

The PROCESS tool only reports unstandardized coefficients. Amos 21 on the other hand can estimate the standardized coefficients. The standardized coefficients are shown in table 4, along with the significance level in Amos and with the PROCESS tool. With the indirect effects the PROCESS tool cannot calculate the exact p-value, but in that case the lower limit confidence interval of the bootstrap was assessed. When this lower value of the confidence interval is above zero then the relationship is significant, since then it is certain that the value will be nonzero. When this is the case ‘Sig’ is reported. By using the two techniques of assessing the direct and indirect effects the results are more robust and we have a higher degree of certainty whether a specific relationship is significant or not.

Results mediating effects

31

β = 0.27, p<.001) and giving opinion (without mediator: β = 0.39, p<.001; with mediator: β = 0.15, p<.05) with the mediators (path b), but in both casese the direct path is reduced in power when the mediator is introduced.

H3c is also supported, which stated that lead userness has a positive effect on seeking news and updates through learning. This path is also partially mediated (β = 0.14, p<.001), because the power of the effect of lead userness on seeking news and updates is reduced when introducing the mediator (i.e. from β = 0.38, p<.001 to β = 0.27, p<.001).

32

DV = OUTGA AMOS PRO AMOS PRO AMOS PRO

IV c M a b c’ LU 0.42 *** 0.27 *** *** EH 0.33 *** *** -0.07 -0.04 SB 0.22 ** *** 0.47 ** *** 0.10 *** Sig LR 0.34 *** *** 0.01 -0.01 ST 0.12 0.19 ** ** 0.02 RC 0.09 0.04 0.00

DV = OUTGO AMOS PRO AMOS PRO AMOS PRO

IV c M a b c’ LU 0.39 *** 0.15 * ** EH 0.33 *** *** -0.09 -0.03 SB 0.22 ** *** 0.33 ** *** 0.07 ** Sig LR 0.34 *** *** 0.14 * 0.04 ST 0.12 0.24 *** *** 0.03 RC 0.09 0.05 0.00

DV = INSO AMOS PRO AMOS PRO AMOS PRO

IV c M a b c’ LU -0.10 -0.10 EH 0.33 *** *** 0.12 0.05 SB 0.22 ** *** 0.08 0.02 LR 0.34 *** *** 0.17 * ** 0.05 ** Sig ST 0.12 0.01 0.00 RC 0.09 0.21 ** 0.02

DV = INSN AMOS PRO AMOS PRO AMOS PRO

IV c M a b c’ LU 0.38 *** 0.27 *** *** EH 0.33 *** *** -0.03 -0.02 SB 0.22 ** *** -0.07 -0.02 LR 0.34 *** *** 0.43 *** *** 0.14 *** Sig ST 0.12 0.17 * *** 0.02 RC 0.09 -0.03 0.00

Table 4 - Structural model results – Lead Userness

Notes:

A) IV = independent variable, DV = dependent variable, M = mediator, AMOS = standardized path coefficient and significance level from Amos, PRO = significance level from PROCESS tool

B) LU = Lead Userness, EH = Entertainment, SB = Sense of Belonging, LR = Learning, ST = Status, RC = Reciprocity, OUTGA = Outgoing Giving Advice, OUTGO = Outgoing Giving Opinion, INSO = Incoming Seeking Opinion, INSN = Incoming Seeking News and Updates.

C) c = IV  DV, a = IV  M, b = IV + M  DV, c’ = IV  M  DV (indirect)

33

4.2.2 Differential effect of lead userness and opinion leadership

The structural model with opinion leadership as the independent variable had the following model fit: χ²/df = 2.499 (χ² = 1011; df = 405); CFI = 0.865; GFI = 0.801; RFI = 0.767 ; TLI = 0.846 ; RMSEA = 0.075. The results of the structural model can be found in Table 5.

Hypotheses H4a-e were about the relations between opinion leadership and the motivations. It was argued that opinion leadership was positively associated with status and reciprocity. Support could be found for H4a (entertainment; β = 0.36, p<.001) , H4b (sense of belonging; β = 0.23, p<.001), H4c (learning; β = 0.24, p<.001 and H4e (reciprocity; β = 0.27, p<.001). Surprisingly, no full support could be found for hypothesis H4d (β = 0.13, p>.05), indicating that opinion leadership is not positively associated with the motivation variable status. Notable is that the relation is significant with the PROCESS technique, but not in Amos. The coefficients of the relationships between opinion leadership and the motivation variables, besides reciprocity, are similar to lead userness.

Hypothesis H5a proposed that lead userness is more strongly associated with giving advice than opinion leadership. Comparing the relations, (LU OUTGA, β = 0.42, p<.001 vs. OL  OUTGA, β = 0.25, p<.001), this study finds evidence supporting this reasoning, thereby supporting H5a.

34

DV = OUTGA AMOS PRO AMOS PRO AMOS PRO

IV c M a b c’ OL 0.25 *** 0.13 * EH 0.36 *** *** -0.05 -0.02 SB 0.23 *** ** 0.49 *** *** 0.11 * Sig LR 0.24 *** ** 0.04 0.01 ST 0.13 * 0.19 ** * 0.02 * Sig RC 0.27 *** *** -0.01 0.00

DV = OUTGO AMOS PRO AMOS PRO AMOS PRO

IV c M a b c’ OL 0.28 *** 0.14 * * EH 0.36 *** *** -0.10 -0.04 SB 0.23 *** ** 0.33 *** *** 0.07 *** Sig LR 0.24 *** ** 0.15 * * 0.03 ** Sig ST 0.13 * 0.25 *** ** 0.03 * Sig RC 0.27 *** *** 0.02 0.00

DV = INSO AMOS PRO AMOS PRO AMOS PRO

IV c M a b c’ OL 0.14 0.00 EH 0.36 *** *** 0.11 0.04 SB 0.23 *** ** 0.07 0.02 LR 0.24 *** ** 0.15 * ** 0.04 ** Sig ST 0.13 * 0.01 0.00 RC 0.27 *** *** 0.22 ** * 0.06 *** Sig

DV = INSN AMOS PRO AMOS PRO AMOS PRO

IV c M a b c’ OL 0.19 ** 0.10 EH 0.36 *** *** 0.00 0.00 SB 0.23 *** ** -0.05 0.00 LR 0.24 *** ** 0.47 *** *** 0.11 *** Sig ST 0.13 * 0.16 * ** 0.02 Sig RC 0.27 *** *** -0.08 -0.02

Table 5 - structural model results - Opinion Leadership

Notes:

A) IV = independent variable, DV = dependent variable, M = mediator, AMOS = standardized path coefficient and significance level from Amos, PRO = significance level from PROCESS tool.

B) OL = Opinion Leadership, EH = Entertainment, SB = Sense of Belonging, LR = Learning, ST = Status, RC = Reciprocity, OUTGA = Outgoing Giving Advice, OUTGO = Outgoing Giving Opinion, INSO = Incoming Seeking Opinion, INSN = Incoming Seeking News and Updates.

35

4.4 Communication flows

The results of the ‘expertise pyramid’ are shown in tables 7-10. Respondents indicated to which group they thought they belonged (A = high expertise, E = low expertise) and then they indicated with every group on how much advice they give or seek, with 5 being the highest value and 1 the lowest value of the intensity. Table 7 and 8 respectively show the computed averages of all respondents that belonged to a group for giving advice and seeking advice. To see whether the advice giving or advice seeking activities were focused on a specific group or rather equally distributed, the relative intensity was computed (table 9 and 10). In table 6 the distribution of the sample of 271 respondents among the groups is shown. The distribution is not similar to the proposed distribution (see figure 3). This can be due to that 1) mostly active users and users with a high level of expertise have filled in the questionnaire and/or 2) respondents have a tendency to think that they belong to a higher group than they actually belong to. Nevertheless the results show interesting findings.

In table 7 and table 8 the averages are shown of the intensity of communication (1 = little, 5 = much). The green gradients are added to the tables to get a visual interpretation of the effects. Table 7 clearly shows that members of group A give the most advice in total to other users. The averages decrease with each group. This indicates that the users with a higher level of expertise will also give more advice. Table 9 shows the relative intensity (= average/total) which shows that the highest expert group (A) gives approximately as many advice to every group (± 20%). The lower the expertise, the more focused the advice giving will be on the same group or a group with a lower level of expertise.

The opposite effect can be found for the advice seeking activities. Table 9 and table 10 show that members of the highest group are mainly focused on their own group for the advice seeking activities and pay little attention to groups with a much lower level of expertise. Notable is that the average intensity of the advice seeking activities is also higher than the advice giving averages, meaning that users tend to look more often for advice than that they will give advice.

36 GROUP # % A 23 8.5% B 113 41.7% C 85 31.4% D 43 15.9% E 7 2.6% Total 271 100.0%

Table 6 – Distribution of groups GROUP / TO  A B C D E Total A 3.00 3.35 3.74 3.83 3.57 17.48 B 2.04 2.76 3.46 3.65 3.41 15.32 C 1.78 2.17 3.10 3.60 3.73 14.40 D 1.30 1.60 1.98 2.84 3.30 11.02 E 1.14 1.14 1.29 2.00 2.43 8.00 Table 7 – Giving advice averages

GROUP / TO  A B C D E Total A 4.35 3.74 2.70 2.09 1.78 14.65 B 4.36 4.00 2.96 1.96 1.63 14.91 C 4.28 4.13 3.48 2.50 2.03 16.42 D 4.49 4.35 3.84 2.93 2.21 17.81 E 3.86 3.57 3.57 3.57 3.00 17.57 Table 8 – Seeking advice averages

GROUP / TO  A B C D E A 17.2% 19.2% 21.4% 21.9% 20.4% B 13.3% 18.0% 22.6% 23.8% 22.3% C 12.4% 15.1% 21.6% 25.0% 25.9% D 11.8% 14.6% 17.9% 25.7% 30.0% E 14.3% 14.3% 16.1% 25.0% 30.4% Table 9 – Giving advice relative intensity

37

4.5 Usage of information sources

Users can use different information sources to gain knowledge about a product or technology. To investigate this we will get a better understanding in how information and knowledge is gathered and shared throughout the community. A linear regression test was done using structural equation modelling using Amos between technical expertise (TE) and the information sources, and between lead userness (LU) and the information sources. In such a way the estimates are computed simultaneously. Both technical expertise and lead userness are used to determine if the innovative characteristic of the lead user, which stems from the characteristics ‘ahead of market trend’ and ‘high expected benefits’, leads to other usage of information sources.

Table 11 shows the structural equation results between technical expertise and the information sources used and lead userness and the information sources used. The findings confirm that technical expertise and lead userness are significantly positively associated with 1) consulting primary information of the manufacturer and 2) visiting multiple communities or websites to get better insights. The results of technical expertise and lead usernesss are fairly similar, although lead userness has stronger path coefficients for both consulting primary information of the manufacturer and visiting multiple communities or websites.

β - TE β - LU Consult primary information of the manufacturer 0.27*** 0.30*** Consult expert reviews 0.03 0.08 Visit multiple communities or websites to get better

insights 0.22*** 0.27*** Search the archive of a forum for relevant information 0.06 0.09 Ask questions on the forums / open a topic on the forums

to ask the question -0.12 -0.83

Table 11 – Usage of information sources

38

4.6 Overview results

Table 12 summarizes all the hypotheses and corresponding results in one convenient table, providing a quick overview of the results.

HYPOTHESIS RESULT

H1a Lead userness is positively associated with the motivation entertainment to visit,

participate in and contribute to an online community.

SUPPORTED

H1b Lead userness is positively associated with the motivation sense of belonging to

visit, participate in and contribute to an online community.

SUPPORTED

H1c Lead userness is positively associated with the motivation learning to visit,

participate in and contribute to an online community.

SUPPORTED H1d Lead userness is positively associated with the motivation reciprocity, to visit,

participate in and contribute to an online community.

NOT SUPPORTED

H2a Lead userness is positively associated with giving advice. SUPPORTED

H2b Lead userness is positively associated with giving opinion. SUPPORTED

H2c Lead userness is negatively associated with seeking advice. NOT SUPPORTED

H2d Lead userness is negatively associated with seeking opinion. NOT SUPPORTED

H2e Lead userness is positively associated with seeking news and updates. SUPPORTED

H3a Lead userness is positively associated with giving advice and giving opinion through

higher entertainment.

NOT SUPPORTED

H3b Lead userness is positively associated with giving advice and giving opinion through

a higher sense of belonging.

SUPPORTED

H3c Lead userness is positively associated with seeking news and updates through

learning.

SUPPORTED

H4a Opinion leadership is positively associated with the motivation entertainment to

visit, participate in and contribute to an online community.

SUPPORTED

H4b Opinion leadership is positively associated with the motivation sense of belonging

to visit, participate in and contribute to an online community.

SUPPORTED

H4c Opinion leadership is positively associated with the motivation learning to visit,

participate in and contribute to an online community.

SUPPORTED H4d Opinion leadership is positively associated with the motivation status to visit,

participate in and contribute to an online community.

NOT SUPPORTED

H4e Opinion leadership is positively associated with the motivation reciprocity to visit,

participate in and contribute to an online community.

SUPPORTED

H5a Lead userness is more strongly positively associated with giving advice than opinion

leadership.

SUPPORTED H5b Opinion leadership is more strongly positively associated with giving opinion than

lead userness.

NOT SUPPORTED