CC BY
13
WHY PARTICIPATE IN THE COMMUNITIES OF INNOVATION? EMPIRICAL ANALYSIS OF FEEDBACK IN COMMUNITIES OF INNOVATION (Part II of the COI series)
by Alexander Pohl, Daniel Mühlhaus, Rolf Weiber, Maria Vola
In part I of the COI series, published in The World of New Economy, No. 1 (3), 2009, the authors presented a theoretical analysis framework for communities of innovation (COI). This framework is based on intensive literature review. Major elements of a COI were defined and especially the questions of success factors, participant groups and motivation drivers were analysed. As the reviewed literature shows, the typical motivations of different COI participants strongly correlate with the types of participants (actor groups) and their role within a COI.
In part II of the COI series the authors focus on the empirical analysis of the theoretically derived hypotheses of part I, and take it further to describe the interactive dynamics (feedback) as a central binding component of a COI. As the time dimension is added into the equation, COI development stages and relative significance of various actor groups in a certain stage are investigated within the context of feedback cycle of a COI. Successful progress and completion of all development stages leads to the overall success of a COI, thus fulfilling motivational needs of the actor groups involved. Hence the definition of a «vicious circle», where COI has to be perceived as successful by its participants during its development process to become successful in reality as a final result. Empirical analysis clearly shows that there are different motivational factors to be considered at each development stage, dependent upon the relative significance of a certain actor group for this particular stage.
This knowledge can be applied by firms, that are interested in fostering COIs around their products or services, or would like to create a culture of a COI within its own structures.
Keywords: Communities of innovation, online communities, open innovation, technological innovation, collective innovation process, innovation stages, empirical test, partial least squares (PLS) approach
1. INTRODUCTION
While Part I of the COI article series describes the major actor groups and their motivational factors, the process of COI development, forming of these actor groups and their interaction is of no less importance to the overall success. COI, as any other community, is a subject to the general laws of team-building and organization. The main difference to the firm-centered community is motivation, and therefore factors determining fulfillment of the motivational needs of the major actor groups. Consequently, as a first step to secure successful development of a COI it is crucial to gain understanding of the existing actor groups and their role in each stage of development. Degree of development success depends then on efficient catering to the specific motivational needs of the relevant actor groups.
In the following chapters we describe the stages of COI development and actor groups particularly relevant at a certain stage, as well as interaction dynamics (feedback) and challenges stemming from the voluntary nature of a COI. Furthermore, we present empirical study and analysis of the results, which confirm or contradict the assumptions of a theoretical COI model. Based upon the study we highlight the implications for firms, as well as areas to be researched in more depth in order to draw specific conclusions and draft the course of action to create and sustain a successful COI.
2. STAGES AND ACTORS IN A COLLECTIVE INNOVATION PROCESS (WHEN? WHO?)
2.1 Stages of the innovation process and specifics in communities of innovation
In order to structure open and community-based innovation projects, it is generally possible to revert to the same approaches that are used in traditional firm-centred innovation processes (cf. Füller et al. 2006; Scacchi et al., 2005). The following is based on an assumed five-stage innovation process that is also typical of firm-centred innovations (cf. similar to Cooper, 1996; Song, Thieme and Xie, 1998):
1. Idea-generation stage
2. Development stage
3. Test stage
4. Modification stage
5. Market launch stage (diffusion)
Here, the following special features of a COI innovation process should be considered in comparison with firm-centered innovation processes. Whereas creativity techniques are often used systematically for brainstorming purposes, in COIs the problem and/or the innovation idea usually already exists when the COI is established. It is beneficial to the success of subsequent process stages if the community pursues not only the ideas that are relevant for a small group with highly specific needs, but also ideas that are relevant for a large number of actors. For this reason, interaction by all actor groups appears to be a central aspect. In the development stage, an essential success factor of a COI project is modularization of the development, i.e. the degree to which a complex, overall task can be split into individual sub-tasks (cf. Rossi, 2006; MacCormack, Rusnak and Baldwin, 2006). If the task cannot be split easily into modules, the members of a COI can develop the innovation jointly only to a limited extent. Finally, testing of the concepts developed in the test stage is more difficult in COIs than in firm-centered innovation processes because prototypes are needed for this purpose. Problems occur here mainly with non-digital innovations, as those generally require more physical investments, which COIs alone may not be able to afford. Similarly, a
commercial actor is required in many cases to diffuse non-digitalized COI innovations, such as concepts for new sports equipment, clothing, or even a complete car.
2.2 Actor relationships in community-based innovation stages
In the following section we describe what types of actors are particularly important in the specific stages of the COI innovation process for its successful completion. These findings allow us further analysis of the feedback between the macro and the micro levels of COIs regarding the process stages (cf. Section 3).
(1) Idea-generation stage
Here, the ideas and any solutions already available are collected to provide a point of departure for joint development. The findings of innovation research show that this stage has a substantial influence on future diffusion of the solutions. Ideas that will only be important in the future for a small number of potential users are considered to be little more than technical gimmicks, and cannot be expected to achieve sustainable success on the market. This is why it is equally important that the users take active part in this stage. However, to put things into perspective we should note, that there is generally no systematic idea-generation stage in COIs. Existing solutions for improvement or problems to be solved are presented to the community right at the start and then become the point of departure for a COI. Besides the users, problems to be solved by a COI may also be presented by any of the other group of actors.
(2) Development stage
The success of the development stage is determined above all by functional interaction between developers and project leaders. As a general principle, it is assumed that an increasing number of developers and the resulting available vast pool of collective know-how will have a positive influence on the solution quality of innovations. A large number of potential problem-solvers with complementary skills results in a working environment of collective learning (cf. Sawhney and Prandelli, 2000). Since, however, having too many developers can lead to difficulties in coordination, it is particularly important to have acknowledged project leaders.
(3) Test stage
This stage is included to identify malfunctions on the one hand, and to align the solution with the need structures of potential users on the other. This is why the users and especially the facilitators play an important role here.
(4) Modification stage
The bugs identified in the test stage, problems of integration of the developed module, and the improvement suggestions all have to be incorporated in the modification stage of the tested prototypes. Thus, the developers once again take on a central function in this phase. The project leaders, however, also play an important role because they are the ones to decide specifically which adjustments are to be made (cf. Li et al., 2006).
(5) Diffusion stage
As the finished or functioning solutions are generally made available to a wide circle of users in the diffusion stage, the users and their decision on whether or not to make use of the solutions play a critical
role here. Particularly if there is no commercial actor to promote the solutions generated within the COI, the «word-of-mouth» concept becomes very important. In this case, the guiding principle is «the more, the better», because the more actors that recommend a product to others or assist less well-trained users in utilizing the product, the faster the solutions will achieve sustainable diffusion. It has also been observed that the people involved in product development or designing of marketing concepts are more willing to recommend the services/products to others (cf. Oetting and Jacob 2007).
The different levels of importance attributed to the individual actors in the phases of the innovation process play central role in the following observations.
3. FEEDBACK IN COMMUNITIES OF INNOVATION (HOW?)
Due to the critical mass nature of COIs, it is the feedback between the macro and the micro levels (refer part I of COI series; see fig. 1) that can lead to «critical occurrences», jeopardizing smooth flow of the innovation process and consequentially the success of a COI. This is why an analysis of the dependencies between the functional elements and feedback structures is essential in order to detect functional faults in a COI, and be able to take preventive measures against possible failure of a COI.
Motive-actor relationship
Fig. 1: Functional elements and feedback in communities of innovation
In order to provide a structure for the following considerations, the relationships between the central functional elements (motive categories, actor groups, innovation process, success dimensions) of a COI are illustrated in Fig. 1. Whereas the success dimensions (process quality, solution quality, and diffusion) of a COI are located on the macro level, motive categories, actor groups and innovation process form the micro level. Feedback is generated between the two levels because the macro level variables of «success dimensions» and «number of users» determine: a) the extent to which the actor-related motive categories are fulfilled, and b) the progression of the innovation process (cf. arrows in Fig. 1 - in bold). In the next section the following analysis is presented:
(1) Feedback and critical mass forming the «vicious circle» in COIs
(2) Analysis of feedback in the context of actor groups
3.1 The «vicious circle» in communities of innovation
Direct network effects and the resulting feedback are generated in COIs because a result has to be achieved on a collective basis. However, this means that a certain minimum number of actors must be engaged (critical mass) in order to enable collaborative work on a single problem. In COIs it is difficult to establish a collective of this kind because a certain collective structure must be available, as has already been demonstrated by the considerations on the various actor groups and their importance in the stages of the innovation process (cf. Section 3.2.).
Thus, the question of the critical mass takes several forms. Micro level: on the one hand the overall collective must achieve a minimum size so that the various motive categories can be satisfied (collective-related critical mass). On the other hand, a minimum number of actors is required within any particular actor group, if the various actor groups are to be able to complete their task in the specific innovation stages (actor group-related critical mass). Considering the informal, voluntary and unpaid nature of COI activities, the problem of critical mass is also present on the macro level: in particular, the «signal need» motive category can only be satisfied properly if the innovation result is also adequately diffused afterwards. However, since the extent of actual diffusion is only established ex post, the actors have to anticipate this development, which means that the actor-specific expectation on the extent of diffusion (critical user mass) has a substantial influence on fulfilment of signal need.
The various critical masses resulting from the direct network effects mean that COIs are subjected to the starting problem, which is typical of critical mass systems. In order to overcome this problem the COI must have at least one so-called basic contributor circle enabling collective activities. Due to the diversity of critical masses of a COI basic contributor circles are described here on the actor group level as «basic actor group-related collectives», and on the level of the overall collective as «basic COI collective». If actor group -related COIs emerge relatively quickly, the basic COI collective can be considered as secured. Here it is important that the actors themselves are located on the micro level, whereas the relevant basic collectives must be ascribed to the macro level as a result of aggregation of all actors in any one group. Since these (model-theoretical) macro variables are important for fulfilment of the contribution motives on the micro-level, feedback is generated here as well.
The possible ways of establishing the various basic collectives depend largely on the motive categories relevant for each actor group, and on their expectations in terms of evolution on the macro level. First of all, we can assume that the micro level has some influence on the macro level because the quality of
collective activities determines the quality of the process, and this in turn has considerable influence on the quality of the solution. On the other hand, the quality of the solution has a positive effect on diffusion of an innovation on the market. Feedback is only generated if the success dimensions on the macro level (solution quality and diffusion) influence the process quality success dimension on the micro level. This can be substantiated as follows:
Participation in a COI takes place on a voluntary basis and without payment. As a result, there are no extrinsic (monetary) incentives for contributing, which is why a COI will only experience «success» if the various relevant needs of the actors can be satisfied. Fulfilment of actors' needs can thus be considered a success in COIs, and is reflected in the three success dimensions.
Motive categories (1), (2), and (6) are not oriented by result, but focus on interaction with the members of a COI. Therefore we can assume that it is primarily the course of the innovation process that determines fulfilment of motives in these categories and that these needs are guaranteed to be satisfied if process quality is high. As a result, the motives in these categories can still be fulfilled even if the solution quality is not as high or the diffusion achieved is not as wide. In contrast, motive categories (3) and (4) are oriented towards achieving a concrete solution, which is why they can only be satisfied by a high solution quality. Similarly, signal need (5) can only be satisfied afterwards, when the generated innovation is launched and diffused broadly in a market. Furthermore, process quality should enforce signal need (5) because it is more rewarding to achieve recognition within a prospering COI. High solution quality also enhances signal need (5), because only remarkable results could increase one's reputation.
Since the success dimensions «solution quality» and «diffusion» only emerge ex post, however, it is only the expected values for these macro success variables that can motivate members to contribute continuously to the process. The anticipated values make these two macro-level success variables to be important determining factors of process quality, and largely account for the feedback dynamics in COIs. The feedback leads to a «vicious circle» (also known as the chicken and the egg problem) in a COI because there is circulating interaction between the micro and the macro levels. On the one hand, the process quality determines the solution quality and its diffusion. On the other hand, solution quality and diffusion are needed to achieve high process quality in the sense of process satisfaction due to the lack of extrinsic incentives. This applies in particular to those actor groups for which motive categories (3) and (5) are relevant.
3.2 Analysis of actor-specific feedback in communities of innovation
If the consideration in the previous section considering various actor groups and their importance in the individual innovation phases are analysed in relation to actors, we obtain the pattern shown in Fig. 2. Here the motive categories per actor group were assigned according to the results presented in Section 4.3.2 of part I of the COI series and importance of the actor groups in the various stages of the innovation process was assessed on the basis of the considerations in Section 2.2 of this paper. The «idealism» motive category is largely ignored in the following as it can be considered important in all phases of the process (lack of extrinsic incentives). A stage-related review of the classifications shown in Fig. 2 permits the following conclusions.
In terms of the risk of a COI being abandoned, the project leaders are considered to be the greatest problem source. On the one hand, they are of major importance in all phases of the innovation process and on the other, their need situation is concentrated on motivational factors (3) «Primary innovation
need» and (4) «Result-oriented interaction need». Ultimately, these needs can only be satisfied by the solution generated. Due to this result orientation and the specific need structure, the motivation of project leaders to participate in the innovation process is sustained primarily by their expectations of the «solution quality». The emergence of expectations with respect to solution quality in the course of the innovation process is also very important for the other actor groups, thus making expectation of «solution quality» a central critical variable for the overall success of a COI.
Micro level Macro level
«Process quality» «Solution Stage 5
Stage 1 Stage 2 Stage 3 Stage 4 quality» «diffusion»
idea-generation development test modification
Project leader 9 • • • • ©
Starting problem: utility handicap Motive categories: (3) and (4) Motive categories: (5)
Developer © • © • • ©
Motive categories: (1) and (2) Motive categories: (3) and (4) Motive categories: (5)
Facilitator © 9 • © 9 9
Motive categories: (1) and (2) Motive categories: (3) and (4) Motive categories: (5)
User • O © O © •
Motive categories neglectable Motive categories: (3)
Stage-specific actor importance: Ç^) none low ^^ medium ^^ high very ^^ high
Fig. 2: Stage-related importance of actor groups and result dimensions
Unlike the project leaders, the needs of developers and facilitators relevant to the innovation process (Stages 2 to 4) can be satisfied by means of a high process quality. Since these two actor groups also play a very important role in generating innovation, the main problem here is that of critical masses. If a basic collective can be established for these two actor groups within a short time, but the critical masses related to actor groups are not reached in the mid-term, there is also a high risk of the COI being abandoned. In contrast, achieving an actor-related critical mass of project leaders bears less relevance because only a small number of project leaders will usually suffice for most projects. For developers and facilitators, it is possible to establish actor group-related basic collectives through the relevant motive categories «(1) Hedonism» and «(2) Non-specific interaction need». The motive category «(5) Signal need» is of primary importance for the active actors' groups in particular. As these motives cannot be fulfilled until the diffusion of the innovation, the success dimension «diffusion» also generates feedback to the innovation process. Here the expectations of the developers in particular are very important for the success of a COI.
4. EMPIRICAL TEST OF FEEDBACK IN COMMUNITIES OF INNOVATION
In order to support the plausibility of our considerations presented so far, we conducted a survey for a prototype of a COI related to an online-browser application called «die staemme». Within this context we examined the members' perception of the three success dimensions, the motive categories and the planned future activity. Specifically, we investigated the subsequent dependencies in greater detail:
• Extent of interdependencies between success dimensions
• Impact of success dimensions on distinct motive categories
• Impact of strength of motive categories on planned activity
The model illustrated in fig. 3 shows the structure of success dimensions and motive categories as described in section 3.1. It is presumed that these motive categories cause the planned activity. In order to investigate differences across the four actor groups the depicted model is calculated for each group separately.
Fig. 3: Interaction of success dimensions and motive categories and their impact on planned activity 4.1 Sample
The investigated online browser application involved more than 600.000 participants with over 50.000 active members in the related community. They developed more than 100.000 contributions which finally resulted in improvements and extensions of the applications' features. Therefore the «die staemme»-community accounts for the great success of the application. We selected this community for a first empirical test because of the dynamic interaction within this community. We posted links within various discussion forums of the COI to an online questionnaire, which led to 1121 completed questionnaires. Based on self-judgement, the interviewees were assigned to the specific actor groups. Similar to other studies in the field of open source software communities, only a small proportion of actors (about 10%) could be assigned to the groups of project leaders and developers (cf. Hemetsberger, 2003). These actor groups are characterized by a durable membership (18.8 and 17.4 months) and a great amount of time they spend within the community (about 20 hours per week).
4.2 Measurement
Due to lack of widespread empirical studies of COIs and thus no validated and well cited scales (e.g. for the success of a COI) we based our measurement for the constructs on the meta studies (see part I of COI series, sections 4.1 and 4.3.2). Here we selected specific or typical items often mentioned in the relevant literature. Based on this extensive literature review and in-depth interviews with experts in the field of open innovation we created scales for success dimensions, motive categories and planned activity. All items used for scale development appear in the appendix. Using a six-point Likert scales (from «strongly agree» to «strongly disagree»), we measured each of the constructs with at least two items. We followed standard psychometric scale development procedures (Gerbing and Anderson, 1988).
To specify the measurement of the constructs we used the procedure proposed by Jarvis, MacKenzie and Podsakoff (2003, p. 201). Solely the motive categories were measured using reflective (effect) indicators (see appendix). For success dimensions and planned activity we used a formative scale, whereas the constructs are defined as a function of their (causal) indicators (cf. Diamantopoulos and Winklhofer, 2001).
Using confirmatory factor analysis, we assessed measurement reliability and validity for each of the 6 motive categories which were operationalized as reflective measures. Overall, the results indicate acceptable psychometric properties. Each construct had a composite reliability greater than the recommended threshold value of .6 (Bagozzi and Yi, 1988). In addition, for all constructs, the coefficient alpha values exceed .6, thus providing evidence for a significant degree of internal consistency among the corresponding indicators (Malhotra, 2004; Nunnally, 1967). Furthermore, item reliabilities are above the recommended value of .4 (Bagozzi and Baumgartner, 1994). For each pair of constructs, we assessed discriminant validity on the basis of Fornell and Larcker's (1981) criterion. The results indicate that there are no problems with respect to discriminant validity.
For the constructs with formative indicators (success dimensions and activity), we checked for multicollinearity. The calculated values of the variance inflation factor for the three attributes related to the constructs n1, n2, n9 and i1 were all below the threshold value of 10 which indicates that there are no problems with respect to multicollinearity (cf. Hair et al., 1998). With all path coefficients from the formative indicators to the composite latent construct having significant effect (all t-values above 1.98) item validity can be confirmed (cf. MacKenzie, Podsakoff and Jarvis, 2005).
4.3 Results
Due to the small sample size in the groups of project leaders and developers we used the partial least squares approach (PLS) to calculate the model estimates (cf. Bollen, 1989). Despite a large number of constructs and a high proportion of formative scales PLS provides robust parameter estimates. Chin and Newsted (1999) recommend a minimum sample size of tenfold the regression with the largest number of independent variables. The existing data satisfies this requirement even for the smallest subset, the project leader data (n=67) with the largest number of independent variables totaling six (P93 to p98).
All Q2 values of the Stone-Geisser criterion are above zero, indicating satisfactory forecast validity across all groups (cf. Fornell and Cha, 1994). The R2 values indicate that the data fits the model reasonably well. Only the motive categories in the facilitators and user data sets are clearly below the threshold of .33 proposed by Chin (1998), pointing out that the related constructs do not account for a substantial amount
of the indicators variance. One possible explanation for this is the pioneer status of the presented study using newly developed scales. Furthermore, the measurement of motives and motive categories is indeed an awkward task due to the multiple faucets of actor-specific singularity. Due to the process orientation of this study these characteristics are not taken into account.
Across all groups of actors the success dimensions are highly related. Process quality or more precisely perceived process success shows a significant effect (p < .01) on solution quality with 021 ranging from .69 to .78 and accounts for a weaker but still significant effect (p < .1) on diffusion. Perceived diffusion is influenced by solution quality with all p12 coefficients above .60. Even if these results are not startling they are in line with the few assumptions existent in the related literature (e. g. Franke and Shah, 2003).
Furthermore, the proposed dependencies were confirmed regarding the effect of success dimensions on distinct motive categories. Process quality has a positive impact on the motive categories (1), (2) and (6) which is significant (p < .05) and stronger for the groups of project leaders and developers. With path coefficients above .1 process quality has a considerable, but not critical effect (p > .05) on signal need, except for the user group. Solution quality consistently shows positive effect on the motive categories (3) and (4), indicating that higher quality of developed innovations enhances the result-oriented motives. The relationship between perceived solution quality and signal need is heterogeneous and even negative for the group of developers. Against the presented considerations of section 2.2, diffusion does not show any noteworthy effect at all.
Noticeable differences are disclosed between the four actor groups if we consider the parameter estimates for the effects of the various motive categories on planned activity (P93 to p98). This can be interpreted as the manifestation of importance of a particular motive. Project leaders are strongly driven by hedonistic motives, whereas for developers none of the motive categories seems to be dominant. This corresponds with the findings of our meta study presented in part I of the COI series (section 4.3.2) where all identified motive categories are of equal importance. For facilitators the motive categories (1), (5) and (6) are crucial for planned activity. This indicates that these actors are less driven by the aim of realizing better solutions, but rather focus on aspects such as collective interaction and learning. In contradiction to the results of our meta study the users are strongly driven by the motive categories (4), (5) and (6).
If we only consider the total effects of success dimensions on future activity, a significant effect (p < .05) results for process quality ranging from .32 to .53 can be identified for the first two actor groups. The total effects for the remaining groups of facilitators and users are significant (p < .05) for process and solution quality. Diffusion, on the contrary, does not account for a noteworthy amount of the variance of planned activity with values of .01 to .06 across all groups.
Altogether the results suggest that the effects of success dimensions on motive categories are comparable across different actor groups. Remarkable discrepancies occur in the motive-activity dependency, which was also implied by the meta study presented in section part I of the COI series, section 4.3.2. The project leaders are strongly motivated where planned activity is explained well by the model (R2=.73), whereas the main effect results from the hedonistic motives (P93=.51). For developers no distinct preference is shown, indicating that these actors are driven by a diversity of motives. The lower share of the planned activities' variance explained by motive categories for the last two actor groups (R2=.29 and .25), implying that their planned behavior could not really be explained by the proposed model.
Project leader (n=67): Process quality Solution quality Diffusion Activity
Months since registration: 18.8; Hours per week: 19.9 (PQ) (SQ) (D) (A)
R2 Q2 Path coefficient Path coefficient Path coefficient Path coefficient
(1) Hedonism .28 .25 .53 *** .51 ***
(2) Non-specific interaction need .41 .33 .64 *** .11 n.s.
(3) Primary innovation need .24 .18 .49 *** .08 n.s.
(4) Result-oriented interaction need .51 .42 »»» -.01 n.s.
(5) Signal need .46 .37 .41 n.s. .14 n.s. .19 n.s. .07 n.s.
(6) Idealism .28 .25 .53 *** .21 n.s.
R2 Q2 Path coefficient Path coefficient Path coefficient Total effects
Process quality (PQ) - .57 .78 *** .12 * .53 **
Solution quality (SQ) .60 .43 12 *** .08 n.s.
Diffusion (D) .68 .55 .01 n.s.
Aktivity (A) .73 .62
Developer (n=82): Process quality Solution quality Diffusion Activity
Months since registration: 17.4; Hours per week: 22.1 (PQ) (SQ) (D) (A)
R2 Q2 Path coefficient Path coefficient Path coefficient Path coefficient
(1) Hedonism .15 .06 .38 *** .06 n.s.
(2) Non-specific interaction need .21 .15 .45 *** -.03 n.s.
(3) Primary innovation need .06 .02 .25 * -.09 n.s.
(4) Result-oriented interaction need .14 .08 .38 *** .11 n.s.
(5) Signal need .35 .29 .35 n.s. -.31 n.s. .23 n.s. .25 n.s.
(6) Idealism .20 .14 .45 *** .33 n.s.
R2 Q2 Path coefficient Path coefficient Path coefficient Total effects
Process quality (PQ) - .54 .78 *** .21 * .32 **
Solution quality (SQ) .60 .25 .65 *** -.02 n.s.
Diffusion (D) .68 .43 .06 n.s.
Aktivity (A) .34 .19
Facilitator (n=423): Process quality Solution quality Diffusion Activity
Months since registration: 15.9; Hours per week: 14.4 (PQ) (SQ) (D) (A)
R2 Q2 Path coefficient Path coefficient Path coefficient Path coefficient
(1) Hedonism .07 .05 .26 *** .15 **
(2) Non-specific interaction need .09 .07 .31 *** .04 n.s.
(3) Primary innovation need .05 .03 22 *** .02 n.s.
(4) Result-oriented interaction need .06 .04 .25 *** .09 n.s.
(5) Signal need .13 .09 .22 n.s. .14 * .04 n.s. 22 ***
(6) Idealism .09 .06 .30 *** .16 **
R2 Q2 Path coefficient Path coefficient Path coefficient Total effects
Process quality (PQ) - .42 .69 *** 21 *** ^ Q A A A
Solution quality (SQ) .48 .32 .69 *** .06 **
Diffusion (D) .57 .39 .01 n.s.
Aktivity (A) .29 .21
User (n=549): Process quality Solution quality Diffusion Activity
Months since registration: 14.2; Hours per week: 11.1 (PQ) (SQ) (D) (A)
R2 Q2 Path coefficient Path coefficient Path coefficient Path coefficient
(1) Hedonism .07 .05 .26 *** -.07 n.s.
(2) Non-specific interaction need .08 .06 .28 *** .08 n.s.
(3) Primary innovation need .02 .02 .16 *** -.06 n.s.
(4) Result-oriented interaction need .04 .04 21 *** <|y ***
(5) Signal need .10 .08 .20 *** .05 *** .09 n.s. .16 ***
(6) Idealism .09 .07 .31 *** .28 ***
R2 Q2 Path coefficient Path coefficient Path coefficient Total effects
Process quality (PQ) - .52 74 *** 23 *** .16 ***
Solution quality (SQ) .55 .39 .61 *** .04 **
Diffusion (D) .64 .48 .01 n.s.
Aktivity (A) .25 .18
Notes: standardized coefficients are shown; *p=10; **p=.05; ***p=.01; n.s.=not significant
Fig. 4: Model parameter estimates
5. IMPLICATIONS FOR FIRMS AND FOR FUTURE ACADEMIC RESEARCH
Modelling of COIs shows that many interdependencies between actors and their motives influence the success of an innovation process. Firms are not typically the initiators of COIs, but they can make the innovation capability of COIs work for their own purposes with well-directed and careful interventions. The paper raises a number of implications for firms. For example, firms can derive hints as to how a COI can be guided to success as a whole and how to prevent a COI from «dying off». A COI can be initiated if a specific innovation process is introduced by a firm and three other active member groups. Considering special characteristics of COIs (particularly voluntary participation, joint actions, and absence of payment), the project's success is determined substantially by degree of motive fulfilment of the various actors. In addition to evaluation and perception of the success variables, the actors' expectations in terms of the COI's continued development is of central importance. Firms can select one of the three success dimensions to build on. If we take success dimension «expectations», firms can emphasize the relevance of the innovation. By announcing that a firm has established a project group with the same objective as a COI, and collaborating with the COI, it sends out a targeted signal to a COI. This indeed should have a positive effect on the expected process and solution quality, therefore impacting the planned activity. Also, it exerts a positive influence on the expectations emerging from the above mentioned motive categories.
Since activity in collaboration with others is one of the characteristic features of a COI, direct network effects are being formed, resulting in the problem of critical mass. Considering this, and the high innovation energy of COIs, the firm must decide how far it can support the innovation process in a COI by targeted actions, preventing it from being abandoned prematurely. When intervening in a COI, it is wise for firms to try to begin with the «critical abandonment variables» of a COI. The analyses performed have shown that this has to be done to achieve the following:
• Establishment of actor group-related basic collectives
• Solving the critical mass problem in the developer and facilitator groups
• Facilitate expectations in terms of the solution quality and diffusion success variables
To contribute to establishing actor-group-related basic collectives at an early stage, firms can delegate specialists and provide suitable development and test tools resulting in higher expectations on future solution quality. Considering high importance of project leaders, firms could also continue in this role by providing qualified project management. We must remember here, however, that the staff provided by firms is not acting on a voluntary basis and without payment from the COI's point of view, which undermines an essential motivation basis of a COI. Any «intervention» here must therefore be made with extreme caution. A particularly promising approach is that described by Dahlander and Magnusson (2005, p. 488) of a symbiotic relationship, where «the firm is focusing on the realization of mutual benefits for both the firm and its community». Firms are also limited with regard to achieving the critical masses. On the one hand, it may be too expensive to fall back on company employees here, and on the other hand, monetary incentives as motivation to contribute could undermine the primary motivation basis of the COI. A firm is much better off if it strengthens expectations relating to solution quality and diffusion. With non-digital innovations, in particular, the invention must also be manufactured at the end of the innovation process. A COI cannot afford to do this by itself and needs assistance from firms. Making the necessary production capacity available will solve the production problem and therefore enhance the solution quality. Also, it will have an additional motivating effect on the COI's contributors due to the higher expected solution quality, which will be assumed if a firm states its readiness to provide production facilities at an early stage. In terms of the
«diffusion» success dimension, firms can achieve a positive effect in fulfilment of the signal need through appropriate marketing measures and early announcements.
Finally, the following constraints should be mentioned in a critical review of the observations made in the present paper. The model approach shown in Fig. 1 is based on the integration of existing empirical studies and conceptual papers. This implies that the existing literature yields a reliable basis, well-suited to derive motive categories and success dimensions. However it is necessary to emphasize the pioneer nature of the present paper. Its purpose is to provide a firm basis for developing a theory on the functioning of COIs. Considering though the general explanatory approach that was selected here, the overall concept may require some modifications to suit the case at hand. Specifically, the structure of the innovation process may require some adaptation to the characteristic features of the corresponding object of the innovation. Depending on the extent to which the generated solutions can be implemented into prototypes and then actual products, the importance of the different process stages may change. This is the case for example if concept tests are only possible on a virtual level in 3D applications, or by falling back on a commercial manufacturer. Also, the substantial use of studies from the OSS sector requires critical review. There is a particular need to examine the extent to which other innovation sectors display similar motive structures and interdependencies. The predominance of references to OSS projects was due to the fact that this is the only field in which successful COIs exist, providing broad and diversified findings. Furthermore, the influence of a commercial firm on the motive structure, as perceived by the actors, should be taken into consideration as this could weaken idealistic motivation source (especially for facilitators and users). This plays an important role in ensuring that the actors continue to contribute to a COI particularly in its initial phase, when the number of contributors and the fulfilment of needs focused on interaction are still low. In addition, the empirical study conducted here is of an indicative nature only, and does not claim to be representative.
These limitations can be eliminated by conducting further empirical studies. Ideally, these studies should differentiate between project leaders, developers, facilitators and users, where a meta-study on the motive structures of individual actor groups can serve as a reference point. An interesting framework for an empirical survey would be a long-term examination of a COI, with surveys conducted at various stages of the innovation process. This could be used to further test the considerations on stage-specific importance of the actors.
APPENDIX
Constructs with formative indicators Measures and Actor groups PL D F U
Process quality (Source: based on Crowston et al., 2003; Franke and Shah, 2003)
•Within the community there is very intensive interaction among the members. R2 - . . .
•The individual members always work together very actively. Q2 .57 .57 .57 .57
•All members support each other when designing patterns or finding solutions to problems. t/VIF all t-values > 1.98 all VIF values < 10
Solution quality (Source: based on Franke and Shah, 2003; Lakhani et al., 2007)
•Patterns that are developed within the community are very innovative. R2 .60 .60 .48 .55
•The patterns that are developed within the community could be successfully marketed. Q2 .43 .25 .32 .39
•Problems with different members can always be solved within the community. t/VIF all t-values > 1.98 all VIF values < 10
Diffusion (Source: based on Crowston et al., 2003; Franke and Shah, 2003)
•The community has a large influence on for the game. R2 .68 .68 .57 .64
•Solutions that were developed in the community are used by many people. Q2 .55 .43 .39 .48
•Solutions of the community help a large amount of people. t/VIF all t-values > 1.98 all VIF values < 10
Activity (stated activity intentions) (Source: based on Lakhani and von Hippel, 2003; Butler, 2001)
•Time I will spent on the community. * R2 .73 .34 .29 .25
•Amount of information I will provide. * Q2 .62 .19 .18 .21
•My future effort to support other members. * t/VIF all t-values > 1.98 all VIF values < 10
Constructs with reflective indicators Measures and Actor groups PL D F U
Hedonism (Source: based on Hars and Ou, 2002; Lakhani and Wolf, 2005)
•Providing information is fun. 0 .85 .66 .77 .79
•It is challenging to interact with other members. CR .93 .85 .89 .91
AV .92 .74 .81 .83
Non-specific interaction need (Source: based on Ghosh et al., 2002; Hars and Ou, 2002)
•I want to improve my skills together with others. 0 .82 .87 .67 .83
•I want to compare my knowledge to that of other members. CR .92 .94 .86 .92
AV .90 .88 .75 .86
Primary innovation need (Source: based on Füller et al., 2006; Ghosh et al., 2002)
•I have a particular problem which has to be solved. 0 .82 .68 .59 .72
•I want to find better solutions for my own application. CR .92 .86 .83 .88
AV .90 .75 .71 .78
Result-oriented interaction need (Source: based on Hemetsberger, 2003; Constant, Sproull and Kiesler, 1996)
•If I provide information, I hope others will do the same. 0 .82 .82 .67 .79
•I provide my own patterns and hope that other members use CR .92 .92 .86 .91
the ideas and improve them as well. AV .90 .85 .75 .83
Signal need (Source: based on Ghosh et al., 2002; Lakhani and Wolf, 2005)
•By participating actively I can earn recognition within the community. 0 .84 .89 .79 .83
•Good ideas give me a good name on the community. CR .92 .95 .91 .92
AV .95 .90 .83 .86
Idealism (Source: based on Hars and Ou, 2002)
•It is fun to help other people. 0 .91 .82 .67 .74
•I want to be a part of this community. CR .96 .92 .86 .88
AV .91 .85 .75 .79
PL: project leader (influences the development of the community); D: developer (develops/provides own solutions); F: facilitator (tests solutions, discovers errors and suggests improvements); U: user (participates without being active);t: t-values for the path coefficients; VIF: variance inflation factorO: coefficient alpha; CR: composite reliability; AV: average variance extracted;* Six-point Likert scales (from «very high» to «very low»)
Fig. 5: Scale items for construct measures
REFERENCES
Bagozzi, R. P. and Baumgartner, H. (1994) The Evaluation of Structural Equation Models and Hypothesis Testing, in: Bagozzi, R. P. (Ed.): Principles of Marketing Research, Cambridge, Massachusetts, 386-422.
Bagozzi, R. P. and Yi, Y. (1988) On the Evaluation of Structural Equation Models, Journal of the Academy of Marketing Science, 16, 74-94. Bollen, K.A. (1989) Structural equations with latent variables, New York.
Butler, B. S. (2001) Membership Size, Communication Activity, and Sustainability: A Resource-Based Model of Online Social Structures, Information Systems Research, 12, 346-362.
Chin, W. (1998) The Partial Least Squares Approach to Structural Equation Modeling, in: Marcoulides, G.A. (Ed.), Modern Methods for Business Research, Mahwah.
Chin, W. and Newsted, P. R. (1999) Structural Equation Modeling Analysis with Small Simples Using Partial Least Squares, in: Hoyle, R. H. (Ed.), Strategies for Small Sample Research, Thousand Oaks, CA, 307-341.
Constant, D., Sproull, L. and Kiesler, S. (1996) The Kindness of Strangers: The Usefulness of Electronic Weak Ties for Technical Advice, Organization Science, 7, 119-135.
Cooper, R. G. (1996) Overhauling the New Product Process, Industrial Marketing Management, 25, 465-482.
Crowston K., Annabi H. and Howison J. (2003) Defining Open Source Software Project Success, 24th International Conference on Information Systems,
Dahlander, L. and Magnusson, M. G. (2005) Relationships between open source software companies and communities: observations from Nordic firms, Research Policy, 34, 481-493.
Diamantopoulos, A. and Winklhofer, H. M. (2001) Index Construction with Formative Indicators: An Alternative to Scale Development, Journal of Marketing Research, 38, 269-277.
Fornell, C. and Cha, J. (1994) Partial least squares, in: Bagozzi, R. P. (Ed.): Advanced Methods of Marketing Research, Oxford.
Fornell, C. and Larcker, D. F. (1981) Evaluating Structural Equation Models with Unobservable Variables and Measurement Error, Journal of Marketing
Research, 18, 39-50.
Franke, N. and Shah, S. (2003) How Communities Support Innovative Activities: An Exploration of Assistance and Sharing Among End-Users, Research Policy, 32, 157-178.
Fuller, J., Bartl, M., Ernst, H. and Muhlbacher, H. (2006) Community based innovation: How to integrate members of virtual communities into new product development, Electronic Commerce Research, 6, 57-73.
Gerbing, D. W. and Anderson, J. C. (1988) An Updated Paradigm for Scale Development Incorporating Unidimensionality and Its Assessment, Journal of Marketing Research, 25, 186-192.
Ghosh, R. A., Glott, R., Krieger, B. and Robles, G. (2002) Survey of developers. Free/Libre and Open Source Software: Survey and Study, FLOSS, Final Report, International Institute of Infonomics, Berlecom Research GmbH.
Hair, J. F., Anderson, R. E., Tatham, R. L. and Black, W. C. (1998) Multivariate Data Analysis, 5th Edition, New Jersey.
Hars, A. and Ou, S. (2002) Working for free? Motivations for participating in Open Source projects, international Journal of Electronic Commerce, 6, 25-39. Hemetsberger, A. (2003) When Consumers Produce on the Internet: The Relationship between Cognitive-affective, Socially-based, and Behavioral Involvement of Prosumers, http://opensource.mit.edu/papers/hemetsberger1.pdf.
Jarvis, C. B., MacKenzie, S. B. and Podsakoff, P. M. (2003) A Critical Review of Construct Indicators and Measurement Model Misspecifications in Marketing and Consumer Research, Journal of Consumer Research, 30, 199-218.
Lakhani, K. R. and von Hippel, E. (2003) How Open Source software works: «free» user-to-user assistance, Research Policy, 32, 923-943. Lakhani, K. R. and Wolf, B. (2005) Why Hackers Do What They Do: Understanding Motivation and Effort in Free/Open Source Software Projects, in: Feller, J., Fitzgerald, B., Hissam, S. and Lakhani, K. R. (Eds.) Perspectives on Free and Open Source Software, Boston.
Lakhani, K. R., Jeppesen, L. B., Lohse, P. A. and Panetta, J. A. (2007) The Value of Openness in Scientific Problem Solving, Working Paper, Harvard Business School, Boston, MA.
Li, Y., Tan, C.-H., Teo, H.-H. and Mattar, A. T. (2006) Motivating Open Source Software Developers: Influence of Transformational and Transactional Leaderships, Proceedings of the 2006 ACM SIGMIS CPR conference on computer personnel research, 34-43.
MacCormack, A., Rusnak, J. and Baldwin, C. Y. (2006) Exploring the Structure of Complex Software Designs: An Empirical Study of Open Source and Proprietary Code, Management Science, 52, 1015-1030.
MacKenzie, S. B., Podsakoff, P. M. and Jarvis, C. B. (2005) The problem of measurement model misspecification in behavoiral and organizational research and some recommended solutions, Journal of Applied Psychology, 90, 710-730. Malhotra, N. K. (2004) Marketing research: an applied orientation, 4th Edition, New Jersey. Nunnally, J. C. (1967) Psychometric theory, New York.
Oetting, M. and Jacob, F. (2007) Empowered Involvement and Word of Mouth: An Agenda for Academic Inquiry, Working paper No. 28, ESCP-EAP, Berlin. Rossi, M. A. (2006) Decoding the free/open source software puzzle: A survey of theoretical and empirical contributions, in: Bitzer, J. and Schroder, P. J. H. (Eds.) The Economics of Open Source Software Development, Amsterdam, 15-55.
Sawhney, M. and Prandelli, E. (2000) Communities of Creation: Managing Distributed Innovations in Turbulent Markets, California Management Review, 42, 24-54.
Scacchi, W., Jensen, C., Noll, J. and Elliott, M. (2005) Multimodal modeling, analysis and validation of open source software requirements processes, Proceedings of the First International Conference Open Source Software, Genova, July 1-8 2005.
Song, X. M., Thieme, R. J. and Xie, J. (1998) The Impact of Cross-Functional Joint Involvement Across Product Development Stages: An Exploratory Study, The Journal of Product Innovation Management, 15, 289-303.
Seattle.
41
