It is important to first note that the objective of the data analysis was to develop an integrative framework which would explain the findings associated with this study’s exploratory research. As such, the data required the valuable yet time consuming process of re-reading and reflection to cultivate a deep familiarization which aided in identifying underlying patterns, concepts, and key factors occurring across the interviewee’s responses (Maylor and Blackmon 2005). The goal of which was to enable a conceptual framework to emerge from the analysis (2005); the focus of which being defined by the fit of the data itself and not forced within the confines of any one existing theoretical model (Yin 2003). The emergent framework integrated knowledge regarding the phenomenon from multiple fields of study encapsulated
within contextual quotes from the data collected within this study to achieve research quality.
Given the richness and volume of data collected from the interviews, multiple data analysis techniques were employed simultaneously to arrive at the findings, discussion, and conclusions. Throughout the process of data collection and analysis, Kolb’s learning cycle was applied for its ability to extract qualitative data to be analyzed from semi-structured interviews in studies related to knowledge sharing (Sniazhko 2011). This cycle begins with concrete experiences during the collection of the data followed by a period of reflective observation to become familiar with the data and assist in spotting patterns. This leads to abstract conceptualization to extract key concepts which are then actively experimented with to identify occurrences of the concepts and the emergence of patters linked back to the original data collected (Maylor and Blackmon 2005). Within this cycle, content analysis was central for identifying and categorizing the emergent themes, aided by the development of a coding matrix using Miles and Huberman’s (1994) three stage data analysis process.
The first step of which was transcribing the interviews to aid in familiarizing the researcher with the data.
The interviews were recorded with two devices to provide a back-up in the event of unforeseen recording issues as well as to help with the transcribing process.
Regarding the video-call interviews, the software was also used to record the conversation for additional clarity. During each interview, supplementary handwritten notes were made which aided in the process of capturing non-verbal cues and ensuring all questions were answered. Directly after each interview, additional detailed notes were made while the conversation was still fresh in the researchers mind. All interviews were transcribed verbatim within 24hrs to ensure maximum clarity and aid in the initial stages of analysis through early recognition and documentation of important themes, patterns, and relationships concurrently with the data collection process. The transcription process helped the researcher greatly with internalizing the topics discussed, as well as assisted in the content analysis process through simultaneously looking for and colour coding concepts and theories which
could explain the interviewee’s responses. Throughout the interview process, a list was made of commonly occurring concepts to help identify similarities and differences across the interviewees for each of the tools. This was helpful in sorting out the valuable nuggets of data from the 273 pages of transcripts and identifying applicable aspects of existing theories.
Miles and Huberman’s (1994: 12) three stages of data analysis is a concurrent process of data reduction, display, and conclusion drawing forming the iterative “interactive cyclical process” of analyzing the data collected. This required the interpretations of the researcher in analyzing the content objectively in order to articulate the central phenomenon by determining the factors influencing a knowledge worker’s adoption and usage of the case company’s ESSP tools (Wurtz 2014:19). Once all the interviews had been conducted, a matrix was developed whereby all interviewee responses could be easily compared to identify similarities of recurring usage patterns and illuminate the factors that differed between interviewees which could explain their usage behaviours.
Therefore, a coding matrix was selected as the ideal form of data display wherein the interviewee data was first reduced to codes which were associated with categories allowing for a visual cross-comparison to identify evident patterns and form conclusions that could be verified (Miles and Huberman 1994: 239). Reduction is the continual process whereby the data is selected, focused, simplified, abstracted, and transformed into patterns with the aim of organizing the data to prepare it for conclusions to be drawn (1994: 12). The difficulty of data analysis in this qualitative research was in uncovering and presenting the meaningfulness of the data without stripping it of the context which provides its value (1994). The data was first reduced by highlighting key themes in the transcripts with different colours so that future re-reading allowed the recognized themes to be illuminated without removing them from their context.
Categories were then developed as they appeared in the data assisted by combining overlapping existing theoretical frameworks from each of the literature sections
reviewed. For example, the category perceived valued outcomes (PVO) was first construed from numerous interviewee’s describing their perception of a specific tool’s value, then combined with: individual motivation theories (e.g. achieving valuable benefits), social organizational theories (e.g. outcome expectations: benefits and rewards), and technology adoption theories (e.g. performance expectancy and perceived usefulness).
As it is the words meanings that matters in qualitative interview research, codes were then formed from ‘unitizing’ chunks of data including words, phrases and direct verbatim quotes (Saunders et al. 2009); allowing for labels which assigned units of meaning to descriptive information (Miles and Huberman 1994: 56). An example of coding used in this analysis was “PVO-MOT-POS” indicating that the data chunk referred to a positive motivator associated with an interviewee’s perceived valued outcome derived from a specific tool. This indicated a potential determinant which was influencing an interviewee’s behavioural intention towards the tool; which was then analyzed in conjunction with their self-reported use of the tool.
Once codes were developed within each category and applied to all interviews in the first matrix, a second matrix was then realigned into three factor groups (re: individual, social, and technological) with two categories each containing the determinant factors based on the codes. Prior coding allowed for data-retrieval where the number of interviewees mentioning a specific factor could be easily grouped together with their verbatim quotes (Miles and Huberman 1994). Conclusions in the form of relationships were then drawn from patterns in the number of interviewees mentioning each determinant and their associated intention and actual usage of each tool. This resulted in the findings and discussion sections which focus on verifying the inter-related six categories of determinants containing both motivational drivers and inhibiting barriers within the three groups of: individual factors, social factors, and technological factors.
As the research was exploratory with a small sample size, individual factors were not cross analyzed with the UTAUT moderators of gender, age, or experience: as the nine interviewees were mostly male (8/9), within an older age range (6/9 over 40), and most
had the same experience duration with the tools (8/9 started in 2006 or earlier) (Venkatesh et al. 2003). However, there was a split between those who use the system voluntarily compared to mandatorily (5/9 voluntary) and as such the findings used this as a contextual identifier in the quotes.
In the findings section to follow, quotes have been provided with a contextual descriptor of the interviewee (Manger- or Operations-level; Mandatory or Voluntary user of the tool) and social software tool or subject being described to provide credibility to the quote as well as ensure the interviewee’s confidentiality is retained. This brings the discussion to the final section of the research methodology, achieving research quality.