DISTINGUISHING LEAD USERS FROM OPINION LEADERS:
A CLUSTER ANALYSIS
By Rick Lotterman S1773038 University of GroningenFaculty of Economics and Business
MSc Strategic Innovation Management
February, 2015
Word count: 9753
Supervisor: Thijs Broekhuizen (1st) and Rene van der Eijk (2nd)
Abstract
This study examines the commonalities of and differences between opinion leaders and lead users. A cluster analysis was performed to categorize 235 users from an online community on consumer electronics in regards to their scores on an opinion leadership and a lead userness construct. A comparison of the mean differences of these clusters was executed using a set of field independent and field dependent, and behavioral factors. The most important result is that none of these factors proved suitable for distinguishing between opinion leaders and lead users, illustrating the similarity of the concepts. Furthermore, the field dependent antecedents expertise depth and expertise broadness prove to be requirements for becoming a lead user or an opinion leader. Finally, the more specific field dependent and behavioral factors are better suited for the identification process of lead users and opinion leaders in an online community, than the more generic field independent factors.
CONTENTS
1. Introduction ... 3
2. Literature Review ... 5
2.1 Lead Userness... 5
2.2 Opinion Leadership ... 6
2.3 Commonalities and Differences ... 6
2.4 Opinion Leaders and Lead Users in Online Communities ... 7
2.5 Research Framework Model ... 8
2.6.1 Field Independent Variables ... 9
2.6.2 Field Dependent Variables ... 10
2.6.3 Behaviors ... 11
2.6.4 Expectations ... 13
3. Methodology ... 14
3.1 Context of the Research ... 14
3.2 Data ... 14
3.2.1 Description ... 14
3.2.2 Construct Creation ... 15
3.2.2 Validity and Reliability ... 15
3.3 Method and Procedure of Analysis ... 15
3.3.1 Clustering Method ... 15
3.2.2 Procedure of Analysis ... 16
4. Results ... 17
4.1 Determining the Number of Clusters ... 17
4.2 Clustering... 17
4.2.1 The 3-Cluster Model ... 17
4.2.2 The 4-Cluster Model ... 20
4.3 Cluster comparison ... 21
4.3.1 Field Independent Variables ... 21
4.3.2 Field Independent Factors ... 21
4.3.3 Behaviors ... 22
1
5. Discussion ... 27
5.1 Discussion of Results and Managerial Implications. ... 27
5.2 Limitations and Future Research ... 30
3
1. INTRODUCTION
Firms benefit from targeting users that can accelerate an innovation’s diffusion. Fast diffusion of an innovation is a solid way to recapture investments. In order to catalyze this diffusion, it can be beneficial to get influential users to adopt a product and promote it to other potential adopters, influencing their buying behavior. Two types of users that fulfil this role are lead users (Schreier & Prügl, 2006) and opinion leaders (Iyengar, 2011; Goldsmith & Witt, 2005; Myers & Robertson, 1972; Shoham & Ruvio, 2008).
Lead users are often defined by two characteristics. First, they are users that are ahead of the market trend, facing specific needs before the rest of the marketplace. Second, they are users that are positioned to obtain relatively high benefits from finding a solution to their needs, making them more likely to innovate (Bilgram Brem & Voigt, 2008; Franke, Von Hippel & Schreier, 2006; Morrison, Roberts & Von Hippel, 2000; Von Hippel, 1986). Lead users are not only very useful in the early stages of the new product development (NPD) process, but also because they act as a role model for potential adopters (Schreier & Prügl, 2008). Since the rest of the market looks at lead users for advice, getting the lead user to adopt, test and support an innovation is highly beneficial for its diffusion (Urban & Von Hippel, 1988).
Opinion leaders are users that act as role models to which other people turn for advice on what product to buy or service to use, and play an important role in accelerating the diffusion process of products in a market (Goldsmith & Witt; Morrison et al., 2000; Shoham & Ruvio, 2008).
4 userness or opinion leadership. While the two constructs have been related to each other, no one has investigated systematically how the constructs come to be and how they are different from each other. In order to identify ways of properly differentiating between lead users and opinion leaders, the following research question was stated: “What factor(s) can be used to
distinguish lead users from opinion leaders in online communities?”
The factors that will be examined can be split up in antecedents and consequents. As antecedents, which are the factors influencing and determining the two constructs, both the Five Factor Model (FFM) and user expertise will be examined. The consequents, which are the results of said constructs, consist of a number of behavioral factors, including early adoption, online participation, the providing of solutions, and communication streams. These behaviors are mostly related to online communities, as this is where the data used in this research was collected. This context is also reflected in the research question.
With opinion leadership and lead userness being important constructs in social network analysis, being able to distinguish between the two user groups will allow researchers to properly attribute consequences to the right construct. This will enable managers to identify lead users and opinion leaders in online communities and it allows them to better facilitate and retain these users by appropriately adapting their online platforms.
5 Chapter Four. Finally, Chapter Five will discuss these results and provide managerial implications, followed by suggestions for future research.
2. LITERATURE REVIEW
2.1 Lead Userness
The two main characteristics differentiating lead users from non-lead users within a given market are that (1) lead users are ahead of important market trends and (2) that they experience a relatively high level of benefit from innovations, making them more likely to innovate (Bilgram et
al., 2008; Von Hippel, 1986; Morrison, Roberts & von Hippel, 2000; Spann et al., 2009; Urban &
von Hippel, 1988).
The first characteristic, being ahead of the market trend, means that lead users face certain needs earlier than most other users in the market (Bilgram et al, 2008; Franke et al., 2006; von Hippel, 1986). This allows lead users to foresee new market trends and help companies create commercially viable products. Especially in fast moving markets, such as the technology markets, it is essential for companies to develop successful innovative products in order to survive (Urban & Von Hippel, 1988).
The second characteristic, a relatively high level of expected benefits from innovations relates to a user’s likelihood to innovate. The greater the perceived reward for finding a solution to a need, the greater the effort a user is willing to make in order to achieve it (Franke et al., 2006; Schreier & Prügl, 2008).
6
2.2 Opinion Leadership
Like lead users, opinion leaders are individuals that play a role in the acceleration of the diffusion process of products in a market. They act as role models for other users in the market and strongly influence the adopting behavior of these users (Goldsmith & Witt, 2005; Iyengar, 2011; Myers & Robertson, 1972; Shoham & Ruvio, 2008). Myers & Robertson (1972) have found opinions leaders to be knowledgeable regarding a few related product domains, but did not find any proof that their knowledge reaches across all areas of interest relating to these domains. Furthermore, they identified a moderate relationship between opinion leadership and innovative behavior. Goldsmith and Witt (2005) show similar findings, with opinion leadership rarely extending across multiple product categories. However, they do find a strong relation between opinion leadership and innovativeness. Additionally, Jacoby & Hoyer (1981) identify a relation between opinion leadership and expertise. Finally, Lyons & Henderson (2005), who looked at the differences between opinion leaders and ordinary users in an online community, state that opinion leaders are more likely to be early adopters, are more innovative, and a have higher online participation than non-leaders, indicating that they mainly differ from other users in terms of their behavioral aspects.
2.3 Commonalities and Differences
Lead users and opinion leaders share many characteristics. According to Kratzer and Lettl (2009), opinion leaders and lead users are both able to influence other users in their purchasing behavior. Moreover, other users look to them for advice (Goldsmith & Witt, 2005; Morrison et
al., 2000; Shoham & Ruvio, 2008; Urban & von Hippel, 1988). Additionally, both lead users (Von
7 lead users a valuable source of information for non-lead users, causing them to also be opinion leaders. Finally, research shows that user expertise is important for both opinion leadership (Jacoby & Hoyer, 1981) and lead userness (Lüthje et al., 2005).
An essential difference between the two constructs relates to networking. Kratzer and Lettl (2009) demonstrate that lead users and opinion leaders both influence the purchasing behavior of others. However, they continue by stating that the two can easily be distinguished on the basis of their network characteristics. Lead users are often located between different social groups and perform a boundary spanning role, connecting different network domains, while opinion leaders tend to be located within a single network domain, displaying higher centrality and higher influence towards people within that domain. Unfortunately, the data used in this research does not include the required information to research the different network positions introduced by Kratzer & Lettl (2009). This type of network information is often not available to researchers and is rather difficult to collect in the context of online communities due to the large amount of users and communication.
In order to separate the constructs we will be looking at several variables, both antecedents and consequents. These variables include user expertise, the FFM of personality traits, and a number of behavioral aspects.
2.4 Opinion Leaders and Lead Users in Online Communities
8
2.5 Research Framework Model
The conceptual structure of this paper is illustrated in figure 1. It displays the central position of the opinion leadership (OL) and lead userness (LU) constructs, as well as the position of different variables in the model. The variables are categorized into three different aspects; (1) field independent variables, which are generic variables stemming from an internal context, such as the personality traits; (2) field dependent variables, which are variables stemming from external context, such as user expertise with products; and (3) behavioral variables, such as communication and multi brand loyalty. The field dependent and field independent factors are antecedents, which means that they influence and affect the constructs, while the behavioral variables are consequences resulting from the constructs. Additionally, the model illustrates the clustering of the constructs into the idealized high-low matrix, which is then coupled back to the antecedent and consequent variables in order to determine the variable scores for each cluster and identify how they differ from each other.
9 2.6.1 Field Independent Variables
In an effort to find ways of identifying the overall psychological traits that can potentially predict the level of opinion leadership and lead userness, this paper looks at the Five Factor Model (FFM), a model containing five constructs that can help identify a person’s personality traits (McCrae & John, 1992). These five constructs are extraversion, agreeableness, conscientiousness, neuroticism and openness. Batra et al. (2001) argue that personality traits can are formed during early childhood and have very enduring influence on social cognition. Furthermore they state that these traits operate hierarchically, meaning that a few fundamental traits influence a large set of secondary traits. This suggest that the personality traits of the FFM can be considered antecedents of opinion leadership and lead userness. Table 1 gives an explanation of the five constructs, each construct should be seen as a scale on which a person has a certain score.
Personality trait Explanation
Extraversion Includes sociability, activity, dominance and the tendency to experience positive
emotions.
Agreeableness Includes sympathy, trust, cooperation and altruism.
Conscientiousness Includes organization, persistence,
scrupulousness and need for achievement. Neuroticism Includes negative effects such as anxiety,
anger, depression and other cognitive and behavioral manifestations of emotional instability.
Openness to experience Includes Imaginativeness, aesthetic sensitivity, depth of feeling curiosity, and need for variety'
10 Table 2 shows whether, when compared to ‘ordinary’ users, the user types are likely to score higher, similar or lower for each personality trait. Because of the opinion leader’s communicative function as a role model (Shoham & Ruvio, 2008) one could expect a relatively high amount of extraversion, which includes sociability. However, lead users’ requirements to communicate with users and manufacturers across different product domains (Ozer, 2009) could also imply a relatively high level of extraversion. Additionally, since lead users are required to cooperate and share information with other users across the product domains, one could expect them to score higher on the agreeableness trait as this traits includes cooperation and altruism. Moreover, having a higher level of the openness to experience trait could be expected in lead users as this trait includes imaginativeness and a need for variety, and lead users are often innovators and operate across multiple product domains. For the conscientiousness and neuroticism personality traits, no meaningful expectations could be suggested for either user type.
Personality trait Expected in Lead users
Expected in Opinion leaders
Extraversion High High
Agreeableness High Similar
Conscientiousness Similar Similar
Neuroticism Similar Similar
Openness to experience High Similar
Table 2: Expected relative personality trait scores for both opinion leader and lead users.
2.6.2 Field Dependent Variables
User expertise is determined by the degree of knowledge and experience a user has in a particular
field (Ozer, 2009), it is also an important factor in the online identification of lead users (Bilgram
et al., 2008) According to Franke et al. (2006), technical expertise is related directly to the ability
11 required to modify a product, while expertise broadness refers to having knowledge across a variety of related product categories (Franke et al., 2006). Lead users require a relatively high level of expertise depth in order to predict future product demands (Lüthje, 2004; Lüthje et al., 2005; Ozer, 2009) and a relatively high level of expertise broadness in order to come up with new innovations (Lüthje et al., 2005). This suggests that relation is present between user expertise and lead userness. While Jacoby & Hoyer (1981) determined that opinion leadership and user expertise are also highly related, their definition of expertise did not consider depth and broadness. Consequently, while both user types are expected to score relatively high on user expertise, differences between people scoring relatively high on lead userness or opinion leadership in terms of expertise depth and broadness might become apparent.
2.6.3 Behaviors
Early adoption refers to whether a user adopts new innovations relative early when compared to
the rest of the market. Lead users are early adopters, which stems from their effort to fill the need gap they are experiencing. By adopting new innovations very early, they might be able to fulfill their needs (Lüthje, 2004; Schreier & Prügl, 2008). Opinion leaders have also been found to be early adopters, as this is necessary for their position as role models of less informed users (Lyons & Henderson, 2005; Rogers, 1995). This indicates that early adoption could be expected to be related to both lead userness and opinion leadership. By comparing the early adoptive behavior scores of people scoring relatively high on either the OL or LU constructs, the relevance of this variable for distinguishing between opinion leaders and lead users can be determined.
Multiple brand loyalty means that a user is familiar with a multitude of brands. Lead users
12
Active participation is the degree to which a user participates in the online environment.
Alers (2011) reports a strong correlation between the level of lead userness and the degree of participation in a given environment. Active participation exposes lead users to relevant information they can use in their innovative efforts (Bilgram et al., 2008). Furthermore, this active participation also means that lead users share their newfound innovations with the community, making them more visible to others. This means they go through a self-selection process that makes them easier to identify within a community (Belz & Baumbach, 2010; Spann et al., 2009). Lyons & Henderson (2005) show that opinion leaders also participate more actively in online communities than non-opinion leaders. These statements suggests that a relation with active participation exists for both lead userness and opinion leadership.
The likelihood of providing solutions and the speed of providing solutions measure the
likelihood and speed with which a user supplies solutions to other users in their community. In line with research executed by Spann et al. (2009), participants were questioned (1) to what extent they provided solutions to other users and (2) whether they were often the first one to post such a solution. These factors are related to both innovativeness and online participation. The fact that both lead users (Franke et al., 2006; Schreier & Prügl, 2008) and opinion leaders (Lyons & Henderson, 2005) are considered more innovative than ‘ordinary’ users suggests that both user types could be related to the likelihood and speed of providing solutions to other users.
Communication refers to the amount of communication of a user and whether this
13 (Ozer, 2009), they can be considered both opinion seekers and opinion givers. Additionally, the fact that both lead users (Schreier & Prugl, 2008) and opinion leaders (Lyons & Henderson, 2005) act as role model for less informed users is an indication that these user types would score relatively high on communication with people of lower status.
2.6.4 Expectations
Although extant research suggests certain relations between the field dependent and behavioral variables, this paper does not hypothesize specific relations. This research investigates the variable scores of both opinion leadership oriented and lead userness oriented people and tries to identify how these groups differ from each other. Table 3 shows the expectations whether the two user types are likely to score high, similar or low compared to ‘ordinary’ users, which are users showing limited to no signs of opinion leadership or lead userness.
Table 3: Expected variable scores relative to ‘ordinary’ users for both lead users and opinion leaders.
Variable Expected relative score for
Lead users
Expected relative score for Opinion leaders
Expertise depth High High
Expertise broadness High High
Early adoption High High
Multiple brand loyalty High Similar
Active participation High High
Providing solutions High High
Speed of providing solutions High High
Communication (Total) High High
Communication higher status High Similar
Communication same status High Similar
14
3. METHODOLOGY
3.1 Context of the Research
This study uses the dataset collected by Alers (2011) in an effort to simplify the identification process of lead users. This dataset fits this paper on distinguishing lead users and opinion leaders because (1) it includes variables relevant to both user types; (2) the online setting makes it easier to reach opinion leaders and lead users due to the self-selection process (Spann et al., 2009); (3) and it is a dataset limited to one product category, being consumer electronics, limiting the distortion of the data by other products.
3.2 Data
3.2.1 Description
The data set used in this study was collected from 235 users on the online community ‘Gathering of Tweakers’, a forum that is linked to the largest technology website in the Netherlands, http://tweakers.net. This is a forum where people come together to discuss leading edge technological innovations and share ideas about modifying and optimizing their electronic equipment. Due to these leading edge discussions, and the fact that many people frequent the forum for information on what consumer electronics to purchase, this community is expected to have both lead users and opinion leaders present in its user base. In an attempt to distinguish these types of users from each other and from ordinary users, this paper tries to identify differences between users in several field dependent, field independent and behavioral variables.
The dataset consists of three parts; (1) the Five Factor Model (FFM) (McCrae & John, 1992), a field independent factor which measures five personality traits; (2) field dependent factors, which are ‘lead userness’, ‘user expertise depth’ and ‘user expertise broadness’ (Franke
et al., 2006), ‘opinion leadership’ (Kratzer & Lettl, 2009; Schreier et al., 2007), and (3) behaviors,
which are ‘early adopting’ (Ozer, 2009), ‘multi brand loyalty’ (Stokburger-Sauer & Hoyer, 2009), ‘participation’, ‘rate of providing solutions to other online users’, ‘speed of providing solutions to
15 3.2.2 Construct Creation
In order to be able to cluster the data on the basis of opinion leadership and lead userness, two constructs of the variables corresponding to these two factors were made. Composite scores of the relevant variables for each construct were calculated using unweighted measures. The opinion leadership construct (OL construct) is based on the variables corresponding to opinion leadership. The lead userness construct (LU construct) is based on the variables corresponding to high expected benefits from innovations, and to being ahead of a market trend, which is in line with the characteristics Von Hippel (1986) uses to describe lead users (LU construct). Additionally, the same method was used to make constructs for several other variables, including each of the personality traits of the FFM, user expertise, communication streams and multiple brand loyalty.
3.2.2 Validity and Reliability
As the dataset was collected by Alers (2011), we can build on the validity and reliability of the data stated in his study. It states that standard procedures were used to test the reliability and validity of the measurement items and that the variables used in his study provided satisfactory validity, which can thus be extended into this research.
This study has formulated composite constructs, including the OL and LU constructs, each of the FFM’s personality traits, expertise depth, expertise broadness, multiple brand loyalty and communication. To test the reliability of these created constructs, Cronbach’s alpha score was calculated for each. Because all scores were higher than 0.7, this suggests that the internal reliability of these constructs was proven. The Cronbach’s alpha calculated for the OL construct and the LU construct were 0.782 and 0.735 respectively, indicating a high level internal consistency for each construct.
3.3 METHOD AND PROCEDURE OF ANALYSIS
3.3.1 Clustering Method
16 population and that the applied variables are not correlated. A Two-step clustering technique was applied to the constructs. The first step of this technique involves a nonhierarchical pre-clustering of the data that results in a number of pre-clusters that is much lower than the number of original cases, making it more suitable for hierarchical clustering. In step two of this technique, the hierarchical clustering technique was applied to the pre-clusters, resulting in a solution with an optimal division of heterogeneity between clusters and the homogeneity within clusters.
3.2.2 Procedure of Analysis
The first step in the analysis was to determine the optimal number of clusters. The number of clusters in an analysis is vital for effectively differentiating between respondent groups within the data. Too many clusters will result in a fragmented set with multiple marginal and meaningless clusters, while too few clusters will result in a large, dominant cluster that is equally meaningless. To determine the optimal number of clusters, the elbow method will be used. This method looks at the explained variance of the clusters sets the max numbers of cluster at the ‘elbow criterion’, which is the point where adding another cluster only marginally increases the amount of explained variance. By performing a hierarchical cluster analysis using the Ward’s method, the explained variance can be determined from the agglomeration schedule (table 4). By identifying the case at which the coefficients make a large jump, the most appropriate number of clusters can be determined (Hair et al., 2006).
The second step of the analysis procedure was to create the clusters, which was accomplished by performing a Two-Step Cluster analysis in SPSS. By performing this analysis on the OL construct and LU construct variables, each respondent became assigned to a cluster, based on their scores on these constructs. This technique resulted in several different clusters that are characterized with heterogeneity between the clusters and homogeneity within the clusters.
17 based on how they scored on these antecedent and consequent variables. When such a significant mean difference between the clusters can be identified, it indicates that the clusters can be differentiated based on their different scores for that variable.
4. RESULTS
4.1 Determining the Number of Clusters
Table 4 shows the agglomeration schedule. This schedule shows a ‘jump’ in coefficients between stage 230 and 231, out of a total 234 stages. According to the elbow method this means that the optimal number clusters for this model is 3. This indicates that a 3-cluster solution would the most adequate. However, because a 4-cluster solution would best fit the idealized high-low matrix presented in the conceptual model, this will report the findings of both the 3 and 4 cluster solution. Figures 2 and 4 show the median scores of the opinion leadership (OL) and lead userness (LU) constructs of the respective cluster models
Table 4 – Agglomeration Schedule
4.2 Clustering
In both cluster models the reported cluster scores on both the OL constructs and the LU construct are the median scores of these clusters. While these cluster scores can often not be considered high in terms of the range of the Likert-scale (1 through 5), they can be considered relatively high in terms of the overall median scores of the constructs, which are 2.33 for the OL construct and 2.5 for the LU construct.
4.2.1 The 3-Cluster Model
18 LU 2.12), which will be called the moderate cluster; (3) and a cluster scoring low on both constructs (OL 1.33, LU 1.62), which will be called the passive cluster. Figure 3 displays the distribution of the clusters in a scatter plot, illustrating the relative positioning of the clusters within the data. The scatter plot clearly shows the active cluster scoring higher on both constructs than the moderate and passive clusters. Furthermore, the scatter plot shows how the moderate cluster is dispersed more evenly over the two constructs, while the passive cluster seems to skew towards a higher score on the LU construct. The scatter plot shares the orientation of the idealized high-low matrix, giving an idea of which cluster can be found in which box of the matrix.
Figure 2: 3-Cluster model bar chart
Active cluster Moderate cluster Passive cluster
19 Figure 3: Scatter plot of the 3 cluster model.
20 Figure 5: Scatter plot of the 4 cluster model
4.2.2 The 4-Cluster Model
The 4-CM shows similar clusters to the 3-CM, with the exception that the first two clusters seem to be split up from the active cluster from the 3-CM. Cluster 1 can be characterized as a LU oriented cluster (OL 2.33, LU 3.25), while cluster 2 can be characterized as an OL oriented cluster (OL 3.00, LU 2.75). The remaining clusters, cluster 3 (OL 2.00, LU 2.12) and cluster 4 (OL 1.33, LU 1.62), appear to be the same clusters as cluster 2 and 3 from the 3-CM (fig. 2), and thus can be characterized as a moderate and passive cluster, respectively. A comparison of the scatter plots of the 3-CM (fig. 3) and the 4-CM (fig. 5) confirms the redistribution of the 3-CM’s active cluster over the 4-CM’s LU oriented and OL oriented clusters.
21
4.3 Cluster comparison
Table 6 displays the cluster sizes and cluster averages for both the 3-CM and the 4-CM, together with their average scores on the comparative variables per cluster. Furthermore, cluster differences significant at the 5% significance level are represented by the accordant cluster number in superscript.
The cluster sizes confirm that one of the cluster was split up. The construct active cluster from the 3-CM (n=110) roughly the size of the LU oriented (n=55) and OL oriented (n=60) clusters from the 4-CM put together. As shown in figure 4, the LU oriented and OL oriented clusters in the 4-cluster model have relatively high scores for lead userness and opinion leadership. This indicates that the differences between these two clusters could help identify variables in which lead users and opinion leaders differ from each other.
4.3.1 Field independent Variables
The extraversion variable does not show any significant mean difference between any of the clusters in both the 3-CM and 4-CM. This means that a person’s extraversion does not affect one’s level of opinion leadership or lead userness and thus cannot be used to distinguish between opinion leaders and lead users. The same results were found for the conscientiousness variable, the neuroticism variable and the openness to experience variable, meaning that neither of these variables affect the opinion leadership or lead userness of a person.
The agreeableness trait is the only variable that shows a weak significant mean difference between the moderate and passive clusters of the 3-CM. This indicates that people scoring high on lead userness and opinion leadership could be more agreeable than those scoring lower on these constructs. None of the personality traits in the five factor model are able to show a relation with either a person’s opinion leadership or lead userness, making them unsuitable for distinguishing between these two types of users.
4.3.2 Field Independent Factors
22 higher on the constructs. The lack of significant mean differences between the LU oriented and OL oriented clusters in the 4-CM suggests that these variables are unsuitable for distinguishing between lead users and opinion leaders. However, they could be useful in identifying these user types within a community.
4.3.3 Behaviors
The early adoption variable shows a significant mean difference between the active cluster and the moderate and passive clusters of the 3-CM, meaning that people scoring high on opinion leadership and lead userness adopt innovations earlier than people scoring moderate or low on these constructs. No such mean difference can be found between the LU oriented and OL oriented clusters in the 4-CM, indicating that early adoption is not a suitable variable to distinguish between lead users and opinion leaders. Similar as in the 3-CM, significant mean differences exist between the passive cluster and the other clusters, which indicates that people scoring low on opinion leadership and lead userness adopt innovations later than those scoring moderate to high on these constructs.
For the active participation variable, a significant mean difference was found in the 3-CM between the active cluster and the moderate and passive clusters This indicates that the higher people score on the OL and LU constructs, the more they actively participate in the online community. In the 4-CM, no significant mean difference was found between the LU oriented and OL oriented clusters, which means that no difference could be found between people scoring relatively high on either the OL or LU construct, indicating that active participation is not a suitable variable to distinguish between opinion leaders and lead users. However, like in the 3-CM, a significant mean difference with the moderate and passive were present, indicating that the active participation variable could be used to help identify lead users and opinion leaders in a community.
The tendency to provide solutions variable and speed of providing solutions variable both show similar significant mean differences between all clusters in the 3-CM, meaning that a higher score on opinion leadership and lead userness constructs suggests a higher tendency to and
speed of providing solutions to other users in an online environment. In the 4-CM, no such
23 indicating that the tendency to provide solutions variable and the speed to provide solutions
variable are not useful for distinguishing between people scoring relatively high on either the
opinion leadership or lead userness construct. However, similar to the 3-CM, the 4-CM shows significant mean differences with the moderate and passive clusters, suggesting that both the
tendency to provide solutions variable and the speed of providing variable can be useful in the
identification process of opinion leader and lead users.
The communication variables show a significant mean difference between all clusters in the 3-CM, meaning that the higher one’s opinion leadership and lead userness scores are, the more a person communicates overall in the online community. The 4-CM does not share these significant mean differences, and only shows significant differences between the passive cluster and the active and moderate clusters. Additionally, no evidence is found that the OL and LU oriented clusters differ regarding the direction of communication, which indicates that the higher people score on the OL and LU constructs, the more they communicate in all directions. While this demonstrates that the level of communication is not a relevant variable in distinguishing lead users and opinion leaders, it suggests that communication can be a useful variable in the identification of these two user types in an online community.
Finally, the multiple brand loyalty variable shows no significant mean differences between any of the clusters in both the 3-CM and the 4-CM, which means that no significant differences were identified in the multiple brand loyalty of people scoring either high, moderate or low on the opinion leadership and lead userness constructs. This means that this variable is unsuitable for identifying and distinguishing between lead users and opinion leaders.
4.3.4 Expectation Analysis
24 construct, and the ‘communication direction’ scores of people scoring relatively high on the OL construct. The findings indicate that the former group scores similarly to people scoring relatively lower on the LU construct, and that the latter group scores higher on communication in all directions than those scoring relatively low on the constructs. Furthermore, none of the variables of the personality traits showed any significant mean differences between the clusters.
25
Variable Expected relative score
for lead users
Expected relative score for opinion leaders
Extraversion High High
Agreeableness High Similar
Conscientiousness Similar Similar
Neuroticism Similar Similar
Openness to experience High Similar
Expertise (Total) High High
Expertise depth High High
Expertise broadness High High
Early adoption High High
Multiple brand loyalty High Similar
Active participation High High
Providing solutions High High
Speed of providing solutions High High
Communication (Total) High High
Communication higher status High Similar
Communication same status High Similar
Communication lower status High High
Table 6 - Cluster models and variable means per cluster. The numbers in superscript refer to the cluster numbers from which it is different at the 5% significance level.
3-cluster model 4-cluster model
Variables Active cluster (1)
Moderate cluster (2)
Passive LU oriented OL oriented Moderate Passive cluster (3) Cluster (1) Cluster (2) Cluster (3) Cluster (4)
5. DISCUSSION
5.1 Discussion of Results and Managerial Implications.
This study sheds light on the differences and commonalities of opinion leaders and lead users, using field independent, field dependent and behavioral factors. Extant research debates whether opinion leadership is a sub construct of lead userness, or that it is a separate construct. This ambiguity regarding the differences between the two constructs hampers researchers’ ability to understand whether consequences originate from either a person’s lead userness or opinion leadership. In order to help differentiate between the two constructs, the following research question was stated: “What factor(s) can be used to distinguish lead users from opinion leaders in online communities?” In order
to answer this question, several antecedent and consequent factors were examined. Even though
the 4-cluster solution presented in this study succeeded in identifying a lead userness oriented and an opinion leadership oriented cluster, no significant mean differences between these clusters were found for any of the examined factors. This means that no factors were identified that can be used to distinguish lead users from opinion leaders in online communities.
This finding naturally raises the question: do we need two constructs, while the current setting is not able to differentiate them well, or are they measuring the same thing? However, it must be noted that while no significant differences between the clusters were found, the ability to identify the two clusters suggests that they are conceptually different, and that there might be other variables that can help differentiate between the two.
28 Similar to the study of Alers (2011), this paper observes that the personality traits of the FFM introduced by Costa & McCrae (1992) are not very strong differentiators. No evidence was found that opinion leaders and lead users can be differentiated in terms of their psychological traits. The disparity in differentiating power between the personality traits and the other antecedents and behaviors suggests that the personality traits are too generic, and that more field specific factors and behaviors are better suited for differentiating between users. This is supported by the fact that the behavioral variables relating specifically to onsite behavior, which are online participation, the likelihood and speed of providing solutions, and online communication, show significant mean differences between all clusters of users in the 3-cluster solution. Additionally, the results regarding the two antecedents, expertise depth and expertise broadness, indicate that these two factors can be considered requirements for becoming lead users and opinion leaders.
The differentiating power of the variables relating to onsite behavior is connected to another finding of this study. While there is a lack of variables that are able to differentiate between lead users and opinion leaders, the findings in this study indicate that there are a multitude of factors that can help identify lead users and opinion leaders within an online community. The antecedent variables expertise depth and expertise broadness, and the consequent variables relating to onsite behavior and early adoption all appear to be useful in the identification process of lead users and opinion leaders within an online population. This finding is in line with the process of self-selection by lead users resulting from their active participation in online communities (Belz & Baumbach, 2010; Spann et al., 2009) and suggests that this process of self-selection is also applicable to opinion leaders.
29 Furthermore, once a manager has identified these valuable users, they can attempt to use this knowledge in an effort to improve advertising revenue. Online community sites are often limited to a specific product category, discussing the products in this category in great detail. This makes the active users in these communities very important to companies operating in the same product category, both in terms of product development and product diffusion. A community manager could, consequently, approach such a company with an offer to sample products and provide product reviews. They could potentially even develop tools to measure the popularity of product reviews on their sites, for example by measuring social media likes and shares. This would allow managers to illustrate and communicate their community’s impact to related companies.
Moreover, the findings of this study imply several ways for managers to improve their online communities and increase user retention. The results indicate that variables related to online participation are important in identifying lead users and opinion leaders. Rewarding the most active users in the community with either product samples or the opportunity to help companies develop new products would give a positive impulse to the users on the forum to be more active. Another way a community could be improved is by facilitating easy communication between users, which is suggested by the importance of the variables regarding online communication. Additionally, the relative importance of expertise in identifying valuable users implies that providing the users the possibility of showing their product expertise and experience on their profiles might make it easier for both community managers and other users to identify the valuable members in a community.
30
5.2 Limitations and Future Research
The data used in this study was collected for the purpose of identifying lead users in an online community. While it did include data on the opinion leadership of the respondents, the dataset might not be optimal for distinguishing lead users from opinion leaders. A survey designed specifically for this purpose might yield more conclusive results.
Future research should identify and look into additional variables that could help classify between lead users and opinion leaders. An example of this is the fact that the expertise in this data was measured in a very general way. By looking at more expertise specific variables such as IQ, education or occupation, more subtle differences between the two construct may become apparent. An idea for future research is to look at the network position of users within a community, as introduced in the research by Kratzer & Lettl (2009), in order to identify lead users and opinion leaders and then explore the differences between these two user groups.
Furthermore, while this research investigates how opinion leaders and lead users differ, it does not investigate the formation process of the two user types. What determines whether a person becomes a lead user or an opinion leader? The results of the antecedent variables regarding user expertise indicate that it plays an important role in the determination of being a lead user or opinion. While the other antecedents in this study, the personality traits of the FFM, were not useful in differentiating between the two user types, they might still play a role in the formation process of lead users and opinion leaders. This would be in line with the statement of Batra et al. (2001), that personality traits are formed during very early childhood, that they have an enduring effect on social cognition, and that they operate hierarchically, meaning that a few fundamental traits can influence a much larger set of secondary traits. A longitudinal study on the formation process of lead users and opinion leaders might identify how the user types are related to each other.
31 expertise. The focal point in this paper is a high tech forum, with a similarly high tech user base, which indicates that the homogeneity of the community in terms of user expertise might be higher than average. This makes it presumable that a lot of opinion leaders share characteristics, like user expertise, with the lead users and vice versa. A community where user expertise is less likely to be an important factor, such as a forum on fashion and lifestyle like https://viva.nl, may show more differences between the user types, as the community’s homogeneity is possibly lower.
32
6. REFERENCES
Alers, H. (2011). Finding a needle in a haystack: introducing a multi-level framework to identify lead users. Faculty of economics and business, University of Groningen.
Batra, R., Homer, P. M., & Kahle, L. R. (2001). Values, susceptibility to normative influence, and attribute importance weights: a nomological analysis. Journal of consumer
psychology, 11(2), 115-128.
Belz, F. M., & Baumbach, W. (2010). Netnography as a method of lead user identification. Creativity and Innovation Management, 19(3), 304-313.
Bilgram, V., Brem, A., & Voigt, K. I. (2008). User-centric innovations in new product development—Systematic identification of lead users harnessing interactive and collaborative online-tools. International Journal of Innovation Management, 12(3), 419-458.
Costa, P., & McCrae, R. (1992). Four ways five factors are basic. Personality and Individual
Differences, 13(6), 653-665.
Franke, N., von Hippel, E., & Schreier, M. (2006). Finding Commercially Attractive User Innovations: A Test of Lead-User Theory. Journal Of Product Innovation Management,
23(4), 301-315.
Goldsmith, R. E., & De Witt, T. S. (2003). THE PREDICTIVE VALIDITY OF AN OPINION LEADERSHIP SCALE. Journal Of Marketing Theory & Practice, 11(1), 28.
Hair, J. F., Tatham, R. L., Anderson, R. E., & Black, W. (2006). Multivariate data analysis (Vol. 6). Upper Saddle River, NJ: Pearson Prentice Hall.
Hippel, E. Von (1986). Lead users: a source of novel product concepts. Management
science, 32(7), 791-805.
33
Jacoby, J., & Hoyer, W. D. (1981). What if opinion leaders didn't know more? A question of nomological validity. Advances in consumer research, 8(1), 299-303.
Kratzer, J., & Lettl, C. (2009). Distinctive Roles of Lead Users and Opinion Leaders in the Social Networks of Schoolchildren. Journal Of Consumer Research, 36(4), 646-659.
Lilien, G., Morrison, P., Searls, K., Sonnack, M. & Von Hippel, E. (2002). Performance Assessment of the Lead user Idea-Generation Process for New Product Development. Management
Science, 48(8), 1042-1059.
Lüthje, C. (2004). Characteristics of innovating users in a consumer goods field: an empirical study of sport-related product consumers. Technovation, 24(9), 683–695.
Lüthje, C., Herstatt, C., & Von Hippel, E. (2005). User-innovators and “local” information: The case of mountain biking. Research policy, 34(6), 951-965.
Lyons, B., & Henderson, K. (2005). Opinion leadership in a computer‐mediated environment. Journal of Consumer Behaviour, 4(5), 319-329.
Morrison, P. D., Roberts, J. H., & Von Hippel, E. (2000). Determinants of User Innovation and Innovation Sharing in a Local Market. Management Science, 46(12), 1513.
Myers, J. H., & Robertson, T. S. (1972). Dimensions of Opinion Leadership. Journal Of Marketing
Research (JMR), 9(1), 41-46.
Ozer, M. (2009). The roles of product lead-users and product experts in new product evaluation.
Research Policy, 38(8), 1340-1349.
Rogers, E.M. (1995). Diffusion of innovations. New York: Free Press.
Schreier, M., Oberhauser, S., & Prügl, R. (2007). Lead users and the adoption and diffusion of new products: Insights from two extreme sports communities. Marketing Letters, 18(1/2), 15-30.
Shoham, A. & Ruvio, A. (2008). Opinion leaders and followers: A replication and extension.
34 Spann, M., Ernst, H., Skiera, B., & Soll, J. (2009). Identification of Lead Users for Consumer Products via Virtual Stock Markets. Journal Of Product Innovation Management, 26(3), 322-335.
Stokburger‐Sauer, N. E., & Hoyer, W. D. (2009). Consumer advisors revisited: What drives those with market mavenism and opinion leadership tendencies and why? Journal of Consumer
Behaviour, 8(2‐3), 100-115.
