3. METHODOLOGY
3.6 D ATA C OLLECTION AND A NALYSIS
3.6.1 Primary Data collection
Aligned with the qualitative, explorative approach of this master’s thesis, primary data has been collected through qualitative interviews with different potential complementors as well as the platform owner. The following sub-section will further elaborate on the collection of primary data.
3.6.1.1 In-depth Semi-structured Interviews
To further examine the research question, in-depth interviews with potential complementors on the platform MILLA were conducted. It was attempted to cover all distinct types of complementors within the research of this study. Hence, the complementors were divided into three different groups based on their business model and their profession. In particular, interviews were conducted
Plan
• Identifying research question
• Discussing topics with case company
• Evaluating case study approach
Design
• Defining unit of analysis
• Literature review
• Case study design (single, multiple, or holistic)
Prepare
• Creating the interview guide
• Preparing for data collection by doing mock interviews
• Gaining relevant approvals
Collect
• Conducting interviews and creating a database
• Using multiple sources
• Maintaining a chain of evidence
Analyze • Analyzing the data by applying suitable coding techniques
Share • Preparing the study in order to fulfil expectations of distinct audience groups
39
with complementors of the following three groups: Companies, Educational platform providers and Amateurs.
Companies are described as organizations that do not follow the primary purpose of producing and offering educational content to a broad audience within their business model. However, these companies often produce online content to educate their own employees for certain topics. It is of further interest of this study to examine the potential way these companies consider participating on MILLA. Three interviews with representatives of companies of different industries could be conducted.
Educational platform providers are organizations that are already specialized and experienced in the online educational sector. These organizations provide their own educational online platforms, on which courses are offered to a broad audience. Thus, it can be noted that they follow a similar business model to the one of MILLA by mainly focusing on educating people through their own online channels. For the purpose of this study, three interviews could be conducted with educational platform providers.
Amateurs are described as individual persons, who do not engage in the educational sector as their primary occupation, but are specialized and skilled in a certain area, which makes them capable to offer online courses on platforms such as but not limited to MILLA. Two amateurs, who are mainly full-time university students, have been interviewed for the purpose of this study. Besides their university studies they already engage in other educational platforms by producing and publishing content of the topic of their own interests. MILLA might become a further platform on which they publish content on and hence, one could relate to them as multihoming complementors.
A further interview has been conducted with a representative of MILLA. By considering the perspectives of all stakeholders, it is attempted to provide a complete view of the research area and therefore, it is of equal importance to include the perspective of the platform owner besides the views of complementors. The table 2 illustrates a list of all interviewees, which, among other things,
40
includes a description of their job level, role on MILLA as well as the type of data that is available for further analyses.
The qualitative interviews were designed as semi-structured interviews to align with the explorative approach of this master’s thesis since the interview structure admits more flexibility to better understand the perspective of the interviewee during the interview. Flexibility in the interview process was of high importance since the preferences of individual complementors were fully unknown prior to the interviews. This enabled the interviewer to refocus certain questions or ask follow-up questions to discover new topics, which have not been considered when designing the interview guide (Baškarada, 2014). An initial structure has been established through an interview guide, which can be found in the appendix. To approach the areas of interest, the interview guide included five open questions, which all interviewees have been asked about. The further course of the interviews was based on the areas the interviewees focused their preferences on or new topics that emerged during the interview. The questions were phrased as open questions using how and why, so interviewees could give in-depth answers and elaborate on their statements.
However, since why questions could cause defensive behaviour by the interviewee, it was attempted to rather phrase how questions (Yin, 2009).
3.6.1.2 Transcription and Coding of interviews
The interviews were recorded with a mobile device after receiving interviewees’ consent.
Subsequently, the interviews were transcribed using the online tool happyscribe8, which has its own speech recognition algorithm to transcribe audio files. Each interview has been uploaded and minor changes were made to the output file if any discrepancies between the original file and the transcribed file could be recognized. The interviews were held and transcribed in English, however, there was one exemption in which the interview had to be conducted in German. Consequently, the interview has been fully transcribed in German and partly translated into English, if needed for analysis purposes. The author of this master’s thesis is a native German speaker as well as fluent in English and thus, qualified to translate German text passages into English. Another interview could not be fully transcribed since the audio recording stopped after a few minutes for unknown reasons.
8 More information: https://www.happyscribe.co/
41
However, field notes were made during the interview and were used as the source of evidence. The table 2 illustrates a summary of all the interview partners.
Table 2: List and description of interviewees
For analysis purposes, the transcribed interviews have been imported and coded in NVivo. Two different coding methods have been used throughout the coding process since the analysis has been divided into two coding cycles. Provisional coding has been applied during the first coding cycle, while Pattern coding has been selected as the most appropriate technique in order to conclude the analysis of data in the second and final coding cycle (Saldaña, 2009).
According to Miles & Hubermann (1994), provisional coding is characterized by a “start list of codes prior to fieldwork” (p.58). Since this is aligned with the abductive approach of this master’s thesis, a provisional list of codes has been created, based on the insights and knowledge gained through the literature review prior to conducting and analysing the interviews. Table 19 illustrates the codes that have been identified as relevant for the purpose of this study. The codes have been divided into two distinct groups, which classify economic as well as technical incentives for complementors in
9 The code list can be found in chapter 3.1.1 Case study in the light of current research. The chapter also provides explanations on how the topics derived from academic literature has been applied to case study of this research
# Job level Type of
stakeholder Company size/Number of
followers Business/Profession Available data 1 Head of Employee Training
Company
< 250.000 employees Mobility
Transcription of interview. Interview has been held and transcribed in German. For the purpose of the analysis, selected parts have been translated into English
2 Government and Regulatory Affairs
Executive < 250.000 employees
Technology Transcription of interview 3 Public Policy and Government Relations
Senior Analyst < 250.000 employees
Technology Transcription of interview
4 Co-founder Educational
platform provider
< 15 employees (Start-Up) Digital education Transcription of interview
5 Entrepreneur Entrepreneur and 5
employees Mathematics Transcription of interview
6 Director B2B and Director Didactics < 500 employees Language education Field notes and transcription of the first minutes of the interview 7 Student and course provider on Udemy Amateur < 300 course attendees on
Udemy Mathematics Transcription of interview
8 Student and educational content provider
on YouTube < 150.000 followers on
YouTube Programming Transcription of interview
9 Representative of MILLA
Platform owner
(MILLA) unknown Politcs Transcription of interview
42
platforms ecosystems. While the codes derived from academic literature in the economic domain describe revenue, non-monetary incentives as well as competition with the platform owner and other complementors, the technical domain focuses on platform ecosystem complexity and openness as potential variables to incentivize complementors.
It has been noted that defining a list of codes prior to collecting and analysing data could negatively affect the results and do not represent the reality that this study, in fact, seeks to explore (Saldaña, 2009). In order to prevent such an effect, the interviews were of semi-structured nature including open questions, which every interviewee has been asked about (further explanations about the research quality of this study can be found in chapter 3.5). This approach also permits discovering additional variables that could not be identified in academic literature yet.
To align with the pragmatic world-view of this master’s thesis and since this study follows the overarching goal of creating a platform governance framework that includes propositions of how to effectively govern complementors in platform ecosystems, pattern coding has been perceived as the most appropriate coding technique to understand the extent to which certain variables affect complementors’ behaviour in platform ecosystems. According to Miles & Huberman (1994), “first-level coding is a device for summarizing segments of data. Pattern coding is a way of grouping those summaries into a smaller number of sets, themes, or constructs.” (p.69). For the purpose of this study, the set of codes identified in the first coding cycle has been broken down into a smaller set of pattern codes that describe mutual behaviour of complementors on certain circumstances. Thus, it will be possible to evaluate coding passages and find patterns in the data.