Master’s Thesis
MSc BA Strategic Innovation Management
“Firm-User Interaction in Online User
Innovation Communities”
A multi-method approach
By:
JONATHAN MELLINK
University of Groningen Faculty of Economics and Business
Abstract
Innovation has always been essential for the survival of organizations. Nowadays online crowdsourcing platforms enable organizations to engage with customers and gather ideas for innovations. More and more firms have started their own Online User Innovation Community (OUIC) to collect ideas generated by their customers. Firm employees interact online with the OUIC community on a regular basis; however this interaction has never been explored by researchers. In this study I take a multi-method approach to look into firm-user interaction in OUICs from a firm’s perspective. I use a qualitative approach to explore data from Starbucks’ OUIC and I find six major interaction types that firm employees use to interact with the users of OUICs. In addition, I set out to find out whether the six interaction types identified have an influence on the idea quality as perceived by the firm and the user, measured by how far an idea progresses in the innovation value chain and the number of votes of the community gives an idea. The results of the quantitative analysis show that idea related information interactions are of major importance for the idea quality to both the firm and the community. The other interaction types are found to have a lesser or negative influence on idea quality.
Introduction
For organizations it is of critical importance to innovate in order to stay ahead of the competition. However, for many organizations it is not easy to come up with a continuous stream of ideas for innovation. For this reason organizations are looking to explore open innovation by turning to external sources for innovative ideas. (Von Hippel, 2005) The rise of new information technologies such as the Internet and later Web 2.0 changed the way organizations could interact with their customers. Open innovation techniques such as crowdsourcing became increasingly popular due to the ease of reaching the customer, increased interaction with multiple customers and the relatively low costs associated with these methods. (Doan et al., 2011; Howe, 2006)
Crowdsourcing has been extensively researched on many different subjects among which the value of crowdsourcing (Afuah and Tucci, 2012; Poetz and Schreier, 2012), what drives user participation in crowdsourcing programs (Jeppesen and Frederiksen, 2006; Nambisan and Baron 2009) and how to organize crowdsourcing communities (Di Gangi et al., 2010; Nambisan and Baron, 2010).
One form of crowdsourcing which is getting increased attention is the use of Online User Innovation Communities (OUICs). (Di Gangi et al., 2010) In these OUICs users can post ideas for innovations, comment on these ideas and vote for or against the ideas. (Dong and Wu, 2015). Several major corporations such Dell and Starbuck have made extensive use of these OUICs for some years, generating thousands of new ideas and hundreds of innovations.
users, allow firms to take a proactive position in their crowdsourcing programs and possibly increase the value of their OUIC programs.
In extension, there is a lack of research and understanding on how these different firm-user interaction types influence the new product development process and specifically the idea generation stage. Crowdsourcing is mainly designed to strengthen the external phase of the idea generation stage of the innovation value chain (Hansen and Birkinshaw, 2007). Therefore it is of interest to both organizations and researchers to find out if employees can influence the quality of the idea generations stage by using different firm-user interactions types.
The goal of this study will thus be twofold. Firstly, I set out to discover the types of firm-user interactions used by firm employee in OUICs. Secondly, I will study whether firm-user interaction types influence the quality of ideas generated on OUICs.
This study sets out to answer the following research questions: What are the types of firm-user interaction firm employees use in OUICs? and: How do different types of firm-user interaction influence idea quality?
In order to fill the gaps identified above I will follow a mixed methods approach. Johnson and Onwuegbuzie (2004, p.17) define mixed methods research as “the class of research where the researcher mixes or combines qualitative and quantitative research techniques, methods, approaches, concepts or language into a single study”. Greene et al. (1989) describes different purposes of multi method studies. The purpose of using a multi method approach during this study is, according to Greene et al.’s (1989) classification, for development purposes. Due to the lack of research into firm-user interactions in OUICs, especially from a firm-side perspective, this study is aimed at creating as much understanding as possible. By using qualitative methods, this study develops constructs from the data so a clear insight into employee interactions in OUICs is achieved. By using the outcomes of the qualitative part of this research as input for the quantitative part of the study, statistical testing for the effects of the interaction types on idea quality is enabled. The mixed methods design allows this study to not only build theory but at the same time allows testing of the theory. By conducting a multi method study, a combination of the best of qualitative and quantitative approaches can be achieved.
evaluation of an idea’s quality by the community. This provides an interesting starting point for further research into the role of firm employees, OUIC communities, idea filtering and selection by the community, and the perception of idea quality.
The rest of the paper will proceed as follows. First, I will develop the theory and present the found categories and constructs of firm-user interaction types. Next, I will present the empirical evidence and hypothesis based on the constructs developed in the first part of the research. Methodology and results are described next. And finally the theoretical and managerial implications are discussed as well as the limitations of this study and the directions for further research.
Theory Development
Due to the lack of research on firm-user interactions by employees in Online User Innovation Communities, there is no existing theory which I can use to categorize the firm-user interactions. Therefore I will need to explore empirical data and use a grounded qualitative approach, such as a case study, to build my own categories and constructs.
According to Eisenhardt (1989) case studies are especially relevant where there is a lack of literature since case studies do not build on existing theory. In addition, one of the strengths of using case studies is the increased chance of creating novel theory. A second advantage is that “the emergent theory is likely to be testable with constructs that can be readily measured and hypotheses that can be proven false” (p.547). Since I want to develop theory that is testable, a case study is an appropriate approach.
MyStarbucksIdea
kept up to date on the success stories of the website. Since the launch of My Starbuck Ideas over 210.000 ideas have been posted by users and more than a thousand ideas have been put into action
In this study I have gathered 3311 ideas posted between 28th of March 2008 and 2nd of September 2011 in the “Building Communities” category. The “Building Communities” category is an interesting category to analyze due to two factors. First, this category has the lowest number of ideas posted per idea launched. This suggests that the ideas posted in this category are of relatively high quality. Second, the “Building Communities” category is used for ideas which are related to both the online and offline community of Starbucks. This makes that a lot of the ideas posted are more complex and of personal nature. This is ideal for interactions among users but also between users and the idea planners. The 3311 ideas have attracted 8266 comments and 367 of these ideas attracted a total of 453 comments from Starbucks idea planners.
The goal of this part of the research is to build theory which is grounded in the data. I am looking for firm-user interaction types from the firm’s perspective. Since my interest lies with the firms’ perspective I will analyze the comments of idea planners posted on the website. By making the unit of analysis is the entire comment, I will be able to perceive the same content as the people who the comment was originally meant for: the users of MyStarbucksIdea. Even though I might miss hidden meanings or messages of the idea planners, I am confident that the average user of the MSI website would also have missed these hidden meanings; therefore the comments are a valuable and valid data-source. Since the focus of this study is on firm-user interaction by firm employees, I only study the interactions by firm employees in the form of idea planner comments. I do not analyze the user comments on the MSI platforms during this study to allow full focus on the analysis of the idea planner comments.
Category Construction
There is a need to find out what kind of interaction types firms use to interact with users in their OUICs. Due to a lack of any existing theory, I will use a grounded approach to analyze the data and developed firm-user interaction categories.
According to Berg et al. (2004) and Silverman (2013) taking a grounded qualitative approach can provide great insights into the workings of social interactions. Specifically, by using a qualitative approach I will be able to closely inspect the data and with each iteration of analysis I will be able gain a better understanding of the data and thus better abstract the information into categories.
Data analysis process
Since there is no prior theory about firm-user interaction types in OUICs, an inductive content analysis is appropriate (Berg et al., 2004). The data analysis process will consist out of four general steps; preparation, categorization, abstraction and iteration (Spiggle, 1994).
Construct Category Description Idea related information
N = 265
Feedback on idea N = 79
A firm employee provides any kind of feedback on the idea. Feedback means that the employee states an opinion on the idea or a certain part of the idea.
Information provision - idea general N = 104
Information on the idea in general, systems already in place in relation to the idea and past experiences with similar ideas.
Information provision - idea status update N = 12
Information on the official idea status. Can be an advance in status, eg. from under review to coming soon. Can also be that the idea gets a certain status now, eg. under review.
Follow-up on the idea N = 57
Information regarding how the idea implementation or review has progressed or actual steps taken to see the idea implemented / reviewed.
Rectification N = 13
The employees point out and rectify wrong information believed true by the poster and / or commenters.
Community Moderation N = 125
Community Moderation N = 16
Commenting with the purpose of moderating the community’s behavior; to make sure the poster’s and/or commenter’s behavior is in line with the expected and desired norms and values of the community.
Thanking poster for contribution N = 54
Expressing gratitude for posting ideas / comments / votes / suggestions.
Thanking commenter for contribution N = 32
Expressing gratitude for comments / votes / suggestions.
Motivate commenters / posters N = 6
Motivating customer to post more ideas, vote more and / or comment more or return to the website. Often by done by the use of positive language.
Casual communication N = 17
Casual communication with member of the community which are not in relation to the idea or MSI in general.
Suggested solutions N = 230
Suggest follow up / further actions N = 150
Actions outside the MSI website that the commenter or poster can do to reach the implementation of the idea.
Solution provision N = 9
Provide a solution to the problem of which the staff thinks will solve the problem or need of the poster / commenter.
Making aware, promoting or inviting to V2V N = 71
Posting about Starbucks’ V2V website, encouraging participation on V2V and / or proposing the use of V2V often as a problem solution.
Probing N = 44
Idea Probing – poster N = 21
Employees ask questions to the idea poster in order to clarify or extent the idea. Employees might ask for more votes or comments from others on the idea.
Idea Probing – Commenter N = 23
Employees ask questions to commenters in order to clarify or extent the idea. Probing questions who address everybody fall in this category.
Non-idea related information N = 98
Information provision – general N = 68
General information provision such as on the coffee, Starbucks in general, the purpose of the website, etc.
Information provision – procedures N = 30
Information on the procedures followed on the website. Can be related to topics like reviewing and selection of idea, function of idea planners and compensation for ideas.
Answering Questions N = 27
Answering questions - Idea poster N = 7
Answering question asked by the idea poster.
Answering questions – Commenter N = 20
Answering question asked by a commenter. Table 1 – Categories of user-firm interaction
The second step is to form the first preliminary categories. During the process of categorization I first started with the open coding process by analyzing the data and writing down every code I felt necessary. During this step I also found that the comments posted by idea planners could be long and possibly have multiple purposes, however I found that the comments did not have more than two intended purposes and thus I decided to limit the number of possible categories assigned per comment to two categories. I freely started to put together the first few categories. The next step is the abstraction step. During this step I grouped the previously identified categories and tried to condense them into more useful conceptual categories. This was done on the basis of common features among the different comments in addition to what I already found and knew about the data.
Eventually the categorization and abstraction steps needed multiple iterations in order to fully code, understand and categorize the data. During these iterations I went back to the original codes and tried to assign each code to the appropriate categories. I added, condensed or removed categories and I got a clearer picture after each iteration. For some comments to be fully understood, I needed to revisit the entire original idea on the website and use the additional information about context, tone or purpose to understand the function of the idea planner’s comment. After three rounds of categorizing and abstraction, I came to a total of 19 different firm-user interactions categories. I decided to control if the coding was consistent and developed a definition for each of the 19 categories based on the common features in the comments. This enabled me to deductively go over all the comments again and check the consistency of each category assigned to the idea planner’s comments. I have included an example of multiple coded comments in appendix A.
One of the main weaknesses of this entire process is that it is inherently subjective and there is a chance of researcher bias. A second coder, a master student from the field of innovation management, was asked to independently code a random sample of the data. The second coder was provided with the coding scheme to help with the coding process (table 1). After a short instruction and an explanation of the definitions, the coder was presented with 50 randomly selected ideas (13.62% of total sample) and was left alone to independently code the comments. The coder has to assign each comment with one or two of the 19 categories from the coding scheme. Cohen's κ was run to determine whether there was agreement between the two coders on the coding of the categories of the 50 ideas. There was substantial agreement on the categories assigned between my coding and the coding of the second coder, κ = 0.755 (95% CI, .663 to .847), p<0.01 (Landis and Koch, 1977).
Category abstraction
To create higher level constructs I needed to continue the abstraction process until I felt that further abstraction of the categories would become unreasonable. I developed six higher level constructs that that incorporated at least two and at most five of the before identified categories. The categories showed what the comment actually said, while these constructs are meant to show the main purpose of a comment or a part of a comment. The categories and constructs found by analyzing the data can be found in table 1 together with the definitions I created for each of the categories.
Findings
I found six constructs by analyzing the 453 comments posted by Starbucks idea planners. This section will shortly describe each of the constructs.
The ‘idea related information’ construct is contracted out of 5 categories all categorized by the provision of information about the idea and everything that has to do with the idea. Several types of information about the idea fall under this construct. For example, information about the idea’s quality in the form of feedback as well as information about past experience with similar ideas are all examples of content that specifically refers to the idea posted. The purpose of this construct is to provide information and feedback on ideas to the community. This will enable users to come up with new, improved ideas or to better tailor ideas to Starbucks’ needs.
An interesting observation is that this construct includes the use of rectifications by idea planners, especially in two cases. There seems to have been a rumor that Starbucks was not supporting US troops deployed in Iraq or Afghanistan which caused a backlash on the MSI website with multiple users posting their feelings on MSI about this. Idea planners were quick to react to these allegations and provided proof that the allegations were false. In addition, a lot of misinformation about Starbucks stores closing in the US seemed to have spread. Again, idea planners quickly rectified the wrong information and informed customers about the truth. The rectification category is a part of the idea related information construct because the information provided in these employee comments still refers to the original idea posted and in a way it is a type of feedback on the user’s original idea post.
‘Community moderation’ refers to interactions by the idea planner with the community. The five underlying categories all have the purpose to motivate the community to participate, to grow a sense of community and to make sure the community’s behavior is in line with the norms and values of the website. Corrective actions such as deleting comments, removing the option to comment on an idea or reprimanding someone are also ways to show what behavior is tolerated on the website and thus these interactions fall under this construct. An interesting finding was that it seems like some ideas would get the status reviewed in order to close the idea and keep people from posting comments on that particular idea. However, if this was the purpose of the idea planners is unknown. A majority of the comments under this construct have a very positive tone and message and encourage discussions on ideas or thank users for their contributions on the website.
problem or need and ensure the satisfaction of the customers of Starbucks. It is especially interesting to note that during the start of the MSI website, Starbucks used another community website called V2V or “volunteer to volunteer”. This website was meant to be a community platform to bring initiatives of Starbucks’ customers together in one place. Idea planners made extensive use of the MyStarbucksIdea platform to make users aware of the V2V website and promote the use of the V2V platform. Comments of idea planners clearly show that the V2V platform has a different purpose than the MSI platform:
“sbx_nric - 7/24/2008 3:25 PM
Hi novophil! Thanks so much for your posts to this idea. My name is Nancy and I am a manager in digital strategy. I am writing to tell you about Starbucks V2V (www.v2v.net/starbucks). It's an online social network about community. You can get involved in many ways through hands-on activities or anytime/anywhere events or simply connect with partners and customers who share your similar interests. The site is currently by invitation only, so if you're interested in checking it out just send me a quick email to v2v@starbucks.com and I'll invite you. Thanks again for your post and please keep in touch. –Nancy” Idea ID-0478
– Comment 1
This in this comment, posted by idea planner sbx_nric, it is clearly stated that V2V is a social network platform for initiatives from the Starbucks community, while the MSI platform is designed for ideas collection from Starbucks customers. Even though the V2V platform was not designed specifically for Starbucks, they did use and promote the platform on the MSI website. A downside of the avid promotion of the V2V platform was that a lot of ideas that users posted on MSI got a very similar response from idea planners about V2V, even when the idea did not contain a problem or need that could be solved by the V2V platform. It even went as far as some responses of idea planners being exact copies of each other with a changed username. The V2V platform was discontinued somewhere around June of 2009 and from that moment on idea planners stopped the promotion of the V2V platform.
The ‘probing’ construct captures all interactions of idea planners that intent to clarify, extend or check the popularity of posted ideas. Often times idea are posted with limited explanation, which prompts probing interactions by the idea planners in order to fully understand the idea and correctly assess the idea’s quality. The purpose of this construct is to gain additional information on an idea to improve idea planners’ ability to evaluate, assimilate and implement the idea. The probing construct was often found together with idea related information construct; most of the times the idea’s quality was assessed and the idea was probed by asking for input or votes from the community. An example of this is the comment of sbx_omm:
“sbx_omm - 6/20/2008 3:46 PM
Hi Strivan. I've been keeping my eye on this idea because I think it is interesting. But I wish we could get a lot more votes so that we know other customers think it's interesting, as well. Can you [s]pread the word among your friends and neighbors to vote on this idea -- it would be neat to get momentum behind it. Once it gets a little momentum it could potentially get lots!” Idea ID-0981
– Comment 1
The ‘non-idea related information’ construct consists out of two categories and captures all the information provision that is not related to the posted idea. Information such as Starbucks’ history, procedures of the website, (international) laws and policies of Starbucks in general are provided by the idea planners. Often this information is provided to support an argument made by the idea planner in relation to the idea, therefore non-idea related information interactions are often found together with idea related information interactions. Questions about rewards for ideas and business proposals by organizations are also found in the ideas posted by users. Idea planner references to the terms and conditions to explain that there is no financial rewards for the users that had their ideas implemented and redirected the business proposals to customer care, are also examples of non-idea related information interactions. This construct is conceptually separated from the idea related information construct since the main purpose of the comments under this construct is to provide customers with general information to better understand the situation or position Starbucks and to enable customers to better understand the rules or procedures followed.
The last construct found in the data is ‘answering questions’. Many ideas and comments posted by users contain questions and idea planners often stepped in to answer the questions of users. The questions and thus the answers are of very diverse nature and can range from questions about the history of Starbucks to strategic choices made by Starbucks and from personal questions to idea planners to the existence of book clubs at Starbucks locations. The general purpose of this construct can thus be explained simply as answering the questions of users. Even though the answering questions construct is often found together with idea related information interactions or non-idea related information interactions, it is conceptually different since in order to qualify as an answering questions interaction a question must have been asked by a user. Therefore these interactions are the result of a user asking for information instead of passively being provided information by idea planners.
These six constructs provide a new insight into the workings of OUICs. The categories and the constructs found by analyzing the data show that firms and specifically their employees can and do proactively engage in different ways with users on crowdsourcing platforms.
Hypotheses
For organizations it is not only of interest to know what kind of interactions their employees can and do use during firm-user interaction, it is especially interesting to find out what the effects of these types of interactions are on the outcomes of the idea generation stage of the innovation value chain (Hansen and Birkinshaw, 2007).
diffused throughout the entire organization (Dong and Wu, 2015). Therefore the official status of an idea is a good indicator for the perceived idea quality by the firm. On the other hand, users can vote on ideas. The number of votes show what the community thinks are the highest quality and most valuable ideas. Therefore the number of votes can be seen as a rating of idea quality by the community. The influence of firm-user interaction types on both the official idea statuses and the number of votes is interesting since it provides a unique look into the relationship between employee interaction types and the quality of the idea generation process from two perspectives.
In this part of the paper I will link the different interaction types to both the firm-side and customer-side quality indicators and develop hypotheses for each of the interaction types. All hypotheses are shown in the conceptual model in figure 1.
Idea Related Information
The purpose of the idea related information construct was defined as to provide information and feedback on ideas posted by users. Feedback has been recognized as critical component to the performance of any complex process (Kluger and DeNisi, 1996). The idea creation process is inherently a complex process since it involve different kinds of information, tacit and explicit, and the recombination of information in various different ways when only a couple of recombinations are sought after by Starbucks (West and Bogers, 2014). For the users of the MyStarbucksIdea website this is especially difficult since users do not know what ideas will be in the interest of Starbucks and the idea planners. When idea planners give feedback on posted ideas, users learn what idea planners like or dislike and can adapt their idea accordingly. Other information provided by idea planners often has the same goal as feedback. Sigala (2012), who also studied the MyStarbucksIdea website, states that Starbucks should in invest in the information transfer between the experts on staff and the users so users can propose more appropriate and feasible ideas. Chen et al. (2012) found that firm feedback has a significant positive effect on participants’ contributions of high quality ideas in Figure 1 – Conceptual Model
Answering Questions Idea Related information
Online User Innovation Communities and Adamczyk et al. (2011) found that users who receive feedback come up with higher quality content than users who do not. Ideas of high quality are more likely to be selected by idea planners and proceed from the idea generation stage to the conversion and diffusion stage. In turn this would result in an idea getting any of the four statuses.
Hypothesis 1: An idea with idea related information firm-user interaction has a higher
likelihood of having an a) under review, b) reviewed, c) coming soon or d) launched status. As stated above, the provision of information and feedback shows that idea planners have interest in an idea and it also provides users with the opportunity to improve the idea. The community will think these ideas are more interesting relative to ideas without idea related information interactions and thus they are more likely turn their attention to the idea and positively evaluate the quality of the idea. As a result the community would award the ideas with this type of firm-user interactions with more votes.
Hypothesis 2: Idea related information firm-user interaction has a positive effect on the
number of votes an idea has. Community Moderation
For an Online User Innovation Community it is of major importance that the community remains a healthy group and that individual users do not ruin the community for other users and the organization. Managing the community is thus an important task of the idea planners on the MSI platform. In Füller et al.’s (2011) paper on design competitions, they state that the “co-creation experience significantly impacts participants’ quality and number of submitted designs. Further, it positively affects individuals’ participation frequency and interest in future competitions” (p.269). It is thus of importance that Starbucks fosters a community where users get a sense of belonging and identity, since this will increase the motivation of users and stimulates users to create high quality ideas (Nambisan, 2002). Kudaravalli and Faraj (2008) found that a key factor in the effectiveness of collaboration on electronic networks is the sustaining of dialogue, which is closely linked to the moderation of communities. In addition, general human interactivity and the creation of a connection to the organization was found to increase social integrative and hedonic benefits which increases customer participation in value creation in Virtual Customer Environments (Nambisan and Baron, 2009).
Idea planners do not only motivate customers, it is also their job to moderate any toxic behavior on the website. The dangers of toxic behavior, like flaming, are that the community will fail to develop constructive feedback on ideas, users will get demotivated and the ability of idea planners to value and assimilate ideas will decrease (Di Gangi and Wasko, 2009). Gebauer et al. (2013) advises that community managers should take an active role in innovation contests to resolve conflicts and ensure that the community’s behavior is in line with the desired norms and values of the website. Since community moderation comments from idea planner have the purpose of creating a community and motivating and increasing participation among users, idea planner’s comments will most like have a positive effect on the generation of high quality ideas which will result in a higher chance of ideas being selected for review and launch.
Hypothesis 3: An idea with community moderation firm-user interaction has a higher
The community moderation interactions provide positive and motivating signals to users and the community will most likely vote more on ideas with such positive and motivating language compared to idea without community moderation interaction. In addition, the community will recognize the high quality ideas that are a result of community moderation comments from idea planners.
Hypothesis 4: Community moderation firm-user interaction has a positive effect on the
number of votes an idea has. Suggested Solution
When users post ideas based on needs or problems that cannot be fulfilled or solved via the MyStarbucksIdea website, idea planners can provide users with directions on how to solve their problems or needs outside the MSI platform. Idea planner can do this to increase the satisfaction of the customers, to make sure that customers come back with ideas that are relevant, or keep to the MSI platform free of ideas that cannot be discussed or implemented on the platform. Ideas where such redirections and solutions are provided have a high chance of being of relatively low interested to idea planners since the solution to the problem or the fulfilment of the need can only be found outside the MSI platform, which is out of the idea planner’s reach. The ideas are evaluated as low quality idea by idea planners for Starbucks in general and thus these ideas have a lower chance of getting past the idea generation stage.
Hypothesis 5: An idea with suggested solution firm-user interaction has a lower likelihood of
having an a) under review, b) reviewed, c) coming soon or d) launched status.
The ideas with suggested solution interactions from idea planners can be high quality ideas but the ideas will not be handled on the MSI platform any further. Even though an idea may not be handled on the website in the future, the community can still recognize the quality of an idea. The suggested solution interaction of an idea planner can give the community a signal that the idea planner saw at least some potential quality in the idea. This can encourage the community to up vote the idea as a statement of agreement to the original idea and idea poster.
Hypothesis 6: Suggested solution firm-user interaction has a positive effect on the number of
votes an idea has. Probing
Idea generation is a complex process. After all, it is difficult to put a recombination of tacit and explicit information back into explicit information that everybody can understand. This often leads to ideas being posted that are difficult to understand for idea planners for one of the following reasons: a lack of details in the idea post or the high complexity of an idea.
Lack of details, as explained by Di Gangi et al. (2010) in his research on Dell’s OUIC IdeaStorm, is a problem that often occurs in OUICs. Many ideas posted consist out of only a couple of sentences without much explanation aside from the basic idea. For idea planners it is difficult to fully evaluate the quality of an idea without extra information. Therefore idea planners try to gain extra information by probing the author or the community for extra information on the idea.
IdeaStorm is based on the complexity of an idea and the ability to of Dell’s idea planners to reduce the complexity of an idea. This shows that idea planners should actively engage with users to reduce the complexity of ideas. Whether an idea is considered too complex is the result of the organization’s and idea planners’ absorptive capacity (Cohen and Levinthal, 1990; West and Bogers, 2014). When the absorptive capacity is too low, idea planners can engage into probing interactions with the idea posters to try to lower the absorptive capacity needed for the idea planners to correctly recognize the value of, assimilate and apply the new knowledge.
As stated before, probing includes the call for more votes or comments from the community. This can be seen as a way to perform a consumer or market analysis since it is a way of trying to gauge the interest of the market for the idea (Ozer, 2005). Overall, the goal of probing seems to be summarized by Sawhney et al. (2005) when they state that “firms seek to reduce uncertainty by interacting directly with customers to understand their needs and preferences” (p.8). Despite the evidence of probing having a positive effect on the idea planner’s ability to evaluate the quality of an idea, much depends on whether the idea poster or at least some community members respond to the probing interactions of the idea planners. While some of the probing interactions may lead to clarifications, many ideas will probably not be clarified (Di Gangi et al., 2010).
Ideas where probing interactions occur are often lacking detail or are highly complex. Ideas like this are difficult for idea planners to evaluate, assimilate and apply. This will result in the idea being perceived as low quality and thus the idea will have a lower chance of moving beyond the idea generation phase.
Hypothesis 7: An idea with probing firm-user interaction has a lower likelihood of having an a) under review, b) reviewed, c) coming soon or d) launched status.
Di Gangi et al. (2010) states that users mostly ignore posts that need clarification and thus require probing interactions. One could expect that via crowdsourcing, the community would fill in the gaps of the idea but according to the research of Di Gangi et al. (2010) this is seldom the case. Due to the lack of explanations, the quality of an idea is often perceived as low and thus the community will likely give the idea a low rating.
Hypothesis 8: Probing firm-user interaction has a negative effect on the number of votes an
idea has.
Non-Idea Related Information
As I stated before, the purpose of this construct is to provide customers with general information to better understand the situation or position of Starbucks and to enable customers to better understand the rules or procedures followed. The non-idea related information interactions can be divided into two categories. Firstly, interactions where general information about Starbucks is shared. Secondly, interactions where procedures are explained.
Secondly, non-idea related information interactions can provide information on procedures. Ebner et al. (2009) states that system trust is essential, together with interpersonal trust, to develop overall trust. Overall trust is a key factor in ensuring that people will participate in idea competitions. Key in developing system trust is making sure that users understand the procedures followed and providing updates to the community on ideas’ currently moving through the innovation value chain. By providing information on procedures, idea planners can thus increase the system trust of the users. Overall, I state that the effect of non-idea related information interactions on idea quality is very low. Therefore the chance of an idea moving beyond the idea generation stage is not influenced by this type of interaction.
Hypothesis 9: Non-idea related information firm-user interaction has no effect on an idea
having an a) under review, b) reviewed, c) coming soon or d) launched status.
The information shared during non-idea related information interactions is of little value for a large majority of the users active on the MSI website. Therefore the community will most likely perceived the quality of an idea with non-idea related information interactions lower and down vote the idea.
Hypothesis 10: Non-idea related information firm-user interaction has a negative effect on
the number of votes an idea has. Answering Questions
Some users post questions instead of ideas or comments. Idea planners answer these questions by providing the asked information and thereby help the users. Nambisan and Nambisan (2008) found that this kind of information provision increases the pragmatic experience of users in Virtual Customer Environments (VCE). In their paper, Nambisan and Nambisan (2008) categorize Starbucks’ MSI platform as a VCE for product concepts. They found that the two most important customer experiences for a product concept VCE are pragmatic and the hedonic experiences. Customers who express positive experiences are twice as likely to remain involved and increase both the quantity and quality of their contributions. I therefore expect that answering questions interactions increase the quality of idea on the long term; however for the short term and current idea I do not expect an increase in idea quality. Therefore I expect a reduced chance to see the idea move through the innovation value chain.
Hypothesis 11: An idea with answering question firm-user interaction has a lower likelihood
of having an a) under review, b) reviewed, c) coming soon or d) launched status.
For the answering questions interaction, the same logic as the non-idea related information interactions applies. The answers to the questions are of little value to the general community on the MyStarbucksIdea website. Therefore I expect the community to down vote ideas containing answering questions interactions.
Hypothesis 12: Answering questions firm-user interaction has a negative effect on the
Empirical Evidence
Data Collection
For the quantitative part of this study I use the same sample as for the qualitative part. However, instead of only analyzing the ideas with firm-user interactions by idea planners, I analyze all the ideas from the “Building Community” category on the MSI website. I collected all ideas posted between the launch of the MyStarbucksIdea website, the 18th of March 2008, up until the 2nd of September 2011. I have collected 3311 ideas and recorded multiple variables for each idea. During this part of the research, the unit of analysis will be the entire idea.
Since there is a lack of research on firm interaction in OUICs, this research can contribute most when the focus will be on the employee interactions and the influence of those interactions on idea quality. Therefore this study will only look at the interactions by firm employees in the form of idea planner comments and will not make in-depth analyses of user comments on the posted ideas.
Dependent variables
The dependent variable for this research is the official idea status and the number of votes an idea has. Starbucks’ idea planners can provide an idea with four different idea statuses. The four different statuses are: under review, reviewed, coming soon and launched. These statuses refer to how far an idea is in the innovation value chain (Hansen and Birkinshaw, 2007). When an idea is generated by users and the idea is considered of high value by idea planners, the idea will entre the selection phase, which is a part of the conversion stage in the innovation value chain. Ideas in this stage will get the ‘under review’ status. When an idea is rejected for whatever reason, it will receive the status ‘reviewed’. This shows users that the idea has been taken under consideration. Ideas that pass the selection phase will continue to the development phase, ideas under development will get the status ‘coming soon’. After the development phase, ideas will go to the spread phase of the diffusion stage. This is the final stage and ideas will be considered launched from here on and thus receive the status ‘launched’. To enable statistical analysis I created four dummy variables, one for each official status. The number of votes is the dependent variable to measure the user-side outcome of different firm-user interactions. Every registered firm-user for the MSI website can cast one vote per idea. This vote is a simple up- or down vote. Each vote is worth 10 point, this means that a down vote would add -10 points and an up vote would add 10 point to the score. The up votes and down votes are added together to result in one aggregated score which is shown next to each idea in the overview and on the idea page. This aggregated score is the input for the variable.
Independent variables
occurrences are the input for the six independent variables. For example, when there is an idea which has three idea planner comments and those three comments are fully coded with categories from the idea related information construct in an idea is given a value of six on the idea related information construct which is the input of the idea related information interaction variable.
Control variables
To control for variance, I collected several control variables. Some users in the community gain the Top Commenter status. These users are often key figures within the MSI community and have been around for quite a while. These users have commented on a lot of ideas, earning the Top Commenter status which can be recognized by a badge in the form of a small text balloon next to their username whenever they post comments or ideas. I record whether the original idea poster is has a Top Commenter badge or not. Likewise, a user can earn the Idea Launched status. When a user has posted an idea that was launched by Starbucks, the user will receive a badge in the form of a lightbulb next to its name which will be visible whenever the user posts a comment or new idea. I record whether the original idea poster is has an Idea Launched badge or not. Both the Top Commenter and the Idea Launched control variables were transformed into dummy variables to enable statistical analysis. In addition, the total number of comments on each idea as well as the number of idea planner comments per idea was collected. All the control variables will be used to gain additional insights on the factors driving the number of votes per idea.
N Minimum Maximum Mean Std.
Deviation
Variance
Statistic Statistic Statistic Statistic Std. Error Statistic Statistic
Number of votes 3311 -600 95190 30,65 32,221 1854,052 3437508,793 Status - under review 3311 0 1 ,00 ,000 ,025 ,001 Status - reviewed 3311 0 1 ,01 ,002 ,121 ,015 Status - coming soon 3311 0 1 ,00 ,000 ,017 ,000 Status - launched 3311 0 1 ,00 ,001 ,052 ,003 No status 3311 0 1 ,98 ,002 ,134 ,018 Top Commenter 3311 0 1 ,05 ,004 ,213 ,045 Idea Launched 3311 0 1 ,08 ,005 ,266 ,071 Number of comments 3311 0 1030 2,50 ,378 21,740 472,630 Number of idea planner comments 3311 0 13 ,14 ,009 ,499 ,249 Interaction – community moderation 3311 0 9 ,04 ,005 ,312 ,098 Interaction –probing 3311 0 3 ,01 ,002 ,132 ,017 Interaction – non-idea related info
3311 0 2 ,03 ,003 ,177 ,031 Interaction – answering questions 3311 0 4 ,01 ,002 ,111 ,012 Interaction – suggested solution 3311 0 3 ,07 ,005 ,306 ,094 Interaction – idea related info 3311 0 10 ,10 ,008 ,443 ,196 Valid N (listwise) 3311
Results
By performing several analyses I want to discover the influence of different kinds of firm-user interactions on the idea status and the influence of different kinds of firm-user interactions on the number of votes an idea receives. To find the effects on statuses of ideas I run binary logistic regressions, also called a logit regression, since the dependent variables are dummy variables and thus binary. I run four separate logistic regression analyses to test each of the four different statuses. To test for the effect of different kinds of firm-user interactions on the number of votes of an idea I run a linear regression with the number of votes as dependent variable. Table 2 shows the descriptive statistics for the independent, dependent and control variables.
In table 3 the results of the binary logistic regressions for the different firm-user interaction types and their influence on the different idea statuses are presented. The results show that hypothesis 1 is confirmed with the exception of hypothesis 1c. Idea related information interactions have a significant and positive effect on the likelihood of an idea having the under review (B=2.166; p<0.01), coming soon (B=2.779; p<0.01) and launched status (B=1.251; p<0.01). However, idea related information interactions have no significant influence on the likelihood of an idea having the reviewed status.
Hypothesis 3 is not confirmed since none of hypothesized relations of community moderation interactions were found to be significant.
Hypothesis 5 is completely rejected. The influences of suggested solution interactions, described in hypothesis 5, were all found to be non-significant with the exception of hypothesis 5d. However hypothesize 5d predicted that the relation between suggested solution firm-user interaction and the likelihood of an idea having a launched status would be negative. The results, however, state that this relation is positive (B=.929; p<0.10) which means that hypothesis 5d is also rejected.
None of the hypothesized relations of probing interactions are found to be significant. Therefore hypothesis 7 is completely rejected.
For non-idea related information interactions, I found that the interactions have a significant and positive effect on the likelihood of an idea having the status reviewed status (B=2.885; p<0.01). However, the hypothesis predicted that non-idea relation information interactions would have no effect on the likelihood of an idea having any status. Therefore, hypothesis 9 is confirmed with the exception of hypothesis 9b.
Hypothesis 11 is completely rejected since all of the effects related to answering questions interactions are found to be non-significant.
In table 4, the results of the linear regression analysis with the number of votes as dependent variable are shown. As predicted, idea related information interactions have a significant and positive effect on the number of votes (B=328.817; p<0.01), which confirms hypothesis 2.
Hypothesis 4 predicted that the relation between community moderation interactions and the number of votes would be positive, however the relation is found to negative (B=–582.317; p<0.01). Therefore hypothesis 4 is rejected.
Hypothesis 8; probing interactions (B=–1067.801; p<0.01), hypothesis 10; non-idea related information interactions (B=–375.634; p<0.01) and hypothesis 12; answering questions interactions (B=–481.138; p<0.05) all have a significant and negative effect on the number of votes an idea has. This confirms all these hypotheses.
As for the control variables of the linear regression analysis, the results show that both total the number of comments (B=79.220; p<0.01) and the number of comments from idea planners (B=227.713; p<0.05) have a significant and positive effect on the number of votes an idea will get from the community. Interesting to note is that while a users with the Idea Launched badge will receive more votes on their ideas (B=527.388; p<0.01), users that have the Top Commenter badge will see a massive decrease in the number of votes on their ideas (B=–1492.962: p<0.01)
Table 5 shows that the linear regression model for the number of votes has a R2 of 0.827 and is statistically significant (p<0.01). This means that the study explains most of the variance of the dependent variable, the number of votes, with this model and no major variance explaining variables have been left out of the model.
Extended Analysis
It can be argued that when an idea gets accepted by Starbucks, the idea has had each of the official statuses since the idea went through the entire innovation value chain (Hansen and Birkinshaw, 2007). Instead of looking at the likelihood of ideas with different interaction types getting a status, the data can also be interpreted in another way to create a new dependent variable; idea acceptance. By transforming each official status into a numerical value I can approximate an idea following the innovation value chain. An additional linear regression will be performed, with the dependent variable idea acceptance, to verify the results of the previous binary logistic regression analyses. Each idea status is assigned a value. Ideas without any status and ideas with the status reviewed are not implemented and thus get the value 1, under review receives the value 2, coming soon the value 3 and launched will receive the value 4. This creates a variable called idea acceptance, where a higher score means that an idea is further on its way to becoming accepted and thus further along the innovation value chain.
Idea Status Dependent variables B S.E. Wald Sig. Exp(B) Status – under review Interaction – community moderation -20,620 1911,604 ,000 ,991 ,000 Interaction – probing -17,443 3401,746 ,000 ,996 ,000 Interaction – non-idea related info 1,649 1,282 1,656 ,198 5,204 Interaction – answering questions -18,082 4656,924 ,000 ,997 ,000 Interaction – suggested solution ,938 1,136 ,681 ,409 2,554
Interaction – idea related info 2,166 ,709 9,326 ,002*** 8,724 Constant -9,656 1,718 31,601 ,000 ,000 Status – reviewed Interaction – community moderation ,579 ,381 2,308 ,129 1,785 Interaction – probing -,410 ,942 ,190 ,663 ,663 Interaction – non-idea related info 2,885 ,326 78,365 ,000*** 17,896 Interaction – answering questions -,671 ,960 ,489 ,484 ,511 Interaction – suggested solution ,533 ,342 2,421 ,120 1,703
Interaction – idea related info -,123 ,272 ,204 ,651 ,884 Constant -4,717 ,188 626,441 ,000 ,009 Status – coming soon Interaction – community moderation -24,281 1622,844 ,000 ,988 ,000 Interaction – probing -19,638 2565,513 ,000 ,994 ,000 Interaction – non-idea related info -12,074 2549,635 ,000 ,996 ,000 Interaction – answering questions -20,515 3969,267 ,000 ,996 ,000 Interaction – suggested solution -15,155 1086,598 ,000 ,989 ,000
Interaction – idea related
info 2,779 1,071 6,732 ,009*** 16,104 Constant -9,957 2,322 18,392 ,000 ,000 Status – launched Interaction – community moderation -,430 ,688 ,390 ,532 ,651 Interaction – probing ,049 1,267 ,001 ,969 1,050 Interaction – non-idea related info ,082 1,363 ,004 ,952 1,086 Interaction – answering questions -1,311 2,242 ,342 ,559 ,269 Interaction – suggested solution ,929 ,543 2,924 ,087* 2,532
Interaction – idea related
info 1,251 ,248 25,524 ,000*** 3,494
Constant -6,682 ,472 200,233 ,000 ,001
* p=< 0.10, ** p=<0.05, *** p=<0.01
Unstandardized Coefficients Standardized Coefficients
B Std. Error Beta t Sig.
(Constant) -144,151 14,720 -9,793 ,000
Top Commenter -1494,962 91,795 -,172 -16,286 ,000***
Idea Launched 527,388 73,678 ,076 7,158 ,000***
Number of comments 79,220 ,698 ,929 113,492 ,000***
Number of idea planner comments 227,713 90,719 ,061 2,510 ,012**
Interaction – community moderation -582,317 77,121 -,098 -7,551 ,000***
Interaction – probing -1067,801 129,579 -,076 -8,241 ,000***
Interaction – non-idea related info -375,634 100,925 -,036 -3,722 ,000***
Interaction – answering questions -481,138 153,711 -,029 -3,130 ,002***
Interaction – suggested solution -56,730 44,289 -,009 -1,281 ,200
Interaction – idea related info 328,817 64,318 ,079 5,112 ,000***
Change Statistics R R Square Adjusted R Square Std. Error of the Estimate R Square Change
F Change df1 df2 Sig. F Change
,909 ,827 ,827 771,905 ,827 1579,610 10 3300 ,000
Unstandardized Coefficients Standardized Coefficients
B Std. Error Beta t Sig.
(Constant) ,999 ,003 349,604 ,000
Interaction – community moderation ,005 ,011 ,011 ,484 ,629
Interaction – probing -,024 ,023 -,020 -1,055 ,291
Interaction – non-idea related info -,006 ,016 -,007 -,401 ,689
Interaction – answering questions -,038 ,029 -,026 -1,304 ,192
Interaction – suggested solution ,016 ,009 ,031 1,814 ,070*
Interaction – idea related info ,100 ,007 ,274 14,590 ,000***
Change Statistics R R Square Adjusted R Square Std. Error of the Estimate R Square Change
F Change df1 df2 Sig. F Change
,268a ,072 ,070 ,156 ,072 42,739 6 3303 ,000
* p=< 0.10, ** p=<0.05, *** p=<0.01
Table 4 – Hypotheses testing: DV = Number of Votes
* p=< 0.10, ** p=<0.05, *** p=<0.01
Table 6 – Coefficients: DV = Idea Acceptance
Table 5 – Model summary of linear regression analysis: DV = Number of Votes
Discussion
Main Findings and Theoretical Contributions
Open innovation is an ever growing phenomenon. Especially since the rise of social networks, organizations have been increasingly using social media technology to open their innovation processes to external idea generation. For organizations this brings the opportunity to gather thousands of ideas from their customers and for customers it provides a unique opportunity to articulate their needs and to participate in the innovation generation and selection process of organizations they are passionate about. For organizations it of major importance to effectively manage their OUICs, which is the reason most of the OUICs have employees of the sponsoring firm working on the website. However, there is a lack of research into the work these employees do. Specifically, there is a lack of knowledge about firm-user interaction from the firm’s perspective. To my knowledge, I am the first to study the firm-user interaction from the firm’s perspective on OUICs. I set out to gain a deeper understanding on firm-user interaction types by studying 453 comments of Starbucks’ employees. 19 different categories based on the actual content of the comments were found. These categories however were to numerous for useful theory building and statistical analysis. I further abstracted the 19 categories into six higher level constructs. These constructs were developed on the basis of the purpose of comments. Each construct thus holds all categories and comments that have a specific higher level purpose as intended by the employees of Starbucks. To answer the first research question, what are the types of firm-user interaction firm employees use in
OUICs?, I found that there are six firm-user interaction types from a firm’s perspective used in an
OUIC. The six main constructs and their respective general purposes can be found in table 8.
Construct Purpose
Idea related information To provide information and feedback on ideas to the community. This will enable users to come up with new, improved ideas or better tailor ideas to Starbucks’ needs.
Community Moderation To motivate the community to participate, grow a sense of community and make sure the community’s behavior is in line with the norms and values of the website.
Suggested solution To provide users with a solution to their problem or need and ensure the satisfaction of customers of Starbucks.
Probing To gain additional information on an idea to improve idea planners’ ability to value and understand the idea
Non-idea related information To provide customers with general information to better understand the situation or position Starbucks and to enable customers to better understand the rules or procedures followed.
Answering questions To answer the questions of users of the MSI website.
To answer the second research question, How do different types of firm-user interaction influence
idea quality?, I ran multiple statistical analyses to study the effects of the six different firm-user
interaction types on idea quality. Idea quality in OUICs can be measured in two ways. For a firm the idea quality can be measured be seeing how far an idea progresses in the innovation value chain which can be measured by the official status an idea has (Hansen and Birkinshaw, 2007). In addition,
the quality of an idea is rated by the community expressed in the number of votes an idea has; an idea with a lot of votes is deemed of higher quality by the community than an idea with less votes. To examine the effect on the idea quality determined by the firm, several statistical analyses were run. This study found that the use of idea related information interactions has a strong positive effect on the chance of an idea getting the under review, coming soon and launched status.
Contrary to the hypothesis, the suggested solution interaction is positively related to the likelihood of an idea having the launched status. An explanation for the positive relation might be that suggested solution interactions are mainly meant to encourage the proactivity of users on a local level while idea planners would handle the idea on a corporate level.
Non-idea related information interactions were found to have a positive effect on the chance an idea having the reviewed status, which was contrary to the hypothesis. I found that idea planners might assign the reviewed status of ideas to disable further commenting on specific ideas, it is possible that a references to either the terms and conditions or a statement about the function of the website was made by idea planners in these cases. This can explain why non-idea related information interactions are positively related to the likelihood of an idea having the reviewed status.
Regarding the number of votes as a measure of idea quality from a community standpoint, idea related information interactions have a strong positive effect on the number of votes an idea receives. Of the five other interaction types, all except suggested solution interactions had a significant and negative relation to the number of votes an idea receives. Suggested solution interactions had no significant effect on the number of votes. In addition, I have found that the number of votes an idea receives is negatively related to the idea poster being a Top Commenter. The reverse is true for users who have had their idea launched. Users with the Idea Launched badge receive more votes on the ideas they post compared to users who don’t have the Idea Launched badge. Possible explanations for Top Commenter’s ideas receiving less votes include the community being more critical of these users or that the community does not like users who are Top Commenters. All the findings of this study are summarized in figure 2.
Answering Questions Idea Related information
This paper provides three contributions to the crowdsourcing literature. Firstly, this study contributes to the crowdsourcing literature (e.g. Afuah and Tucci, 2012; Leimeister et al., 2009; Nambisan, 2002) by developing six firm-user interactions types from the firm’s perspective. Employee interaction with users on OUICs has, as far as I know, never been researched. A unique look into the workings of an Online User Innovation Community seen from the organizations’ side is provided. These six interaction types enrich the understanding of the functioning of crowdsourcing platforms and the way firms involve themselves in the idea generation process of their customers on crowdsourcing platforms. Secondly, I provide a contribution to the crowdsourcing management literature (e.g. Di Gangi et al., 2010; Franke and Piller, 2004; Jeppesen and Molin, 2003). I have found six different types of firm-user interaction and also studied the effects of the different interactions types. These effects clearly show that one type of interaction is superior when it comes to idea quality. For the community and the firm, the quality of an idea was positively influenced when idea planners used idea related information interactions. The other types of interactions did not have a significant effect. With these results I help enrich the understanding of idea generation in crowdsourcing platforms. Thirdly, I contribute to the crowdsourcing community literature (e.g. Bayus, 2013; Ebner et al., 2009; Fleming and Waguespack, 2007). The community clearly has an important function in OUICs. The community’s main responsibility is the creation of high quality ideas; in addition the community’s votes are being used by idea planners as a filter to select potentially high quality ideas. (Blohm et al., 2011; Riedl et al., 2010) All of the interaction types except idea related information interactions have a negative effect on the idea quality from the community’s perspective. I enrich the literature on communities by being the first to research the effect between firm interaction and the community’s perception of ideas.
Managerial Implications
This study brings forward some important managerial implications. I have found that organizations should watch out with using too many social network platforms at the same time. In the case of Starbucks, I found that the promotion of the V2V website attracted a lot of interest from the idea planners. Idea planners started posting a copied post about V2V on a lot of ideas posted on the MSI website, even when the idea had nothing to do V2V. This promotion of the V2V platform takes away the attention from the goal of the MSI website; generating ideas.
A strong positive effect of the idea related information interaction type on the chance of having the launched status was found. For organizations, simply collecting ideas is not enough. For an idea to have value it needs to be launched and the chance of an idea having the status launched is higher when idea planners interact with users by using idea relation information interactions. Organizations should instruct and train their employees to engage in idea related information interactions to increase the quality of ideas posted on the OUIC.
Limitations and Further Research
The study has some limitations. Firstly, I used only one specific source to collect the data. Since Starbucks is one of the most popular OUICs on the internet it provides me with a very large and accessible data base, however it also decreases the generalizability of this study. Therefore, future studies should try to validate the results in different OUICs from companies in both similar and different industries.
Secondly, this study combines both qualitative and quantitative methods. By constructing and testing firm-user interaction constructs in one study I have made a large contribution to the understanding of firm-user interactions in OUICs. However, instead of having concentrated on one method, I needed to divide my attention over two distinctly different methods creating the risk of losing some of the intricacies of the data. In addition, it means that I analyze the constructs I developed myself while the constructs have not been verified by independent researchers. Therefore I see that future research can complement this study by validating the constructs developed.
Thirdly, I gathered the data from a website which contains only text and analyzed this data for the meaning behind the text. It is very difficult to know what the process was by which a comments came to be and it is therefore possible that the analysis of some comments is wrong or that the actual purpose of some comments has not been found. Future research can complement this study by diving deeper into the organization behind the OUIC. By speaking to the actual idea planners, their motivations for the use of different interaction types as well as the effect of user comments on the forming of the idea planners’ interactions can be studied. This would provide even more insight in the workings of OUICs on the firm side.
Other venues for further research include the verification and extension of the developed theory. I am the first to research the firm-user interaction types from a firm’s perspective on OUICs and other researcher need to follow to allow for further theoretical exploration of the subject. Another interesting direction for future research would be to look at the development of the number of votes an idea has over time. Does the number of votes from the community really react to the employee comments or did the community already have its mind made up about the quality of an idea from the very beginning? Research into the blog posts of the idea planners provides yet another direction for research into the firm’s management of OUICs.
Acknowledgements
