CAPÍTULO IV. ANÁLISIS E INTERPRETACIÓN DE LOS RESULTADOS
6.6 DESARROLLO DE LOS CONTENIDOS
In this section, the researcher discusses the analysis of the interview transcripts; however, as emphasised by Fletcher (2017), few critical realist scholars show how the ontology and epistemology assumptions of critical realism can affect the analysis of collected data. Moreover, the researcher could not identify a specific data analysis approach related to critical realism. However, in many cases, critical realist researchers adopt a data analysis approach in line with the adopted research strategy, either qualitative or quantitative. Accordingly, as a qualitative strategy is used in this study, the researcher adopted a thematic analysis (coding and content analysis) approach, which has been widely employed in qualitative critical realist research (Fletcher 2017). Moreover, the adopted data analysis approach should not only explain the observed experiences and events but also identify some of the causal mechanisms of both social structures and agency within the real domain and their effects on these observed experiences and events (Fairclough 2005; Volkoff et al. 2007). Therefore, the analyses of interview transcripts was completed in the following three stages (Fletcher 2017) (see Figure 11): first, an analysis of the interviewees’ instructions to the double, which was used to identify the experiences and events from the empirical and actual domains, respectively (see Figure 12); second, a inductive thematic analysis of both the interviews to the double ( only the first prat) and in-depth interviews with the practice observers, which was used to look for evidence related to the causal mechanisms from the domain of the real; and third, a retroductive backward reasoning from the experiences and events from the empirical and actual domains, respectively (observed from the first stage of the analyses) to some of the mechanisms that helped cause them, or to some aspects of these causal mechanisms made manifest by the methods used (from the second stage of the analyses) (see Figure 12).
Figure 11 The three stages of data analysis
Figure 12 The three stages of data analysis in relation to the three domains of reality
4.5.2 Stage 1: The Analysis of the Interview to the Double
In this first stage of the data analysis (which was partially adopted from the work of Gorli et al. 2012), the researcher analysed the second part of the interview with the practice participants (see sub-section 4.3.1), where the interviewees instructed the double on three different scenarios of dealing with unsatisfied customers. The interviewees’ instructions to the double were analysed daily and, to speed the process of analysis, completed manually before conducting the in-depth interview with the practice observers. During this stage of analysis, the focus was on identifying the following:
(1) The primary action taken by the front-line service employees while trying to satisfy the customer.
(2) The actors involved that affect the front-line service employees while trying to satisfy the customer.
(3) The organisational aspects that affect the front-line service employees while trying to satisfy the customer.
The outcome of this analysis stage was used to (1) identify events that characterise service recovery and also represent the normal day-to-day conduct of service recovery from the front-line service employees’ point of view in the domain of the actual to learn about the respondents’ experiences (the empirical); (2) identify further questions that could be asked during the in-depth interview with the practice observers; and (3) identify statements/experiences that could be used as part of the in-depth interview with the practice observers to allow immersion in the field of study and look for evidence related to the causal mechanisms within the real domain.
4.5.3 Stage 2: Thematic Analysis of both the Interview to the Double and In-depth Interview with Practice Observers
In the second stage, a thematic analysis was conducted of the interview transcripts from both the first part of the interview, with the practice participants (see sub-section 4.3.1), and the in-depth interview, with the practice observers (see sub-section 4.3.2). The researcher used Atlas.it computer-aided qualitative data analysis software (CAQDAS), which has similar features to Nvivo, including coding, analysing and the ability to upload a variety of file formats. However, Atlas.it is the only CAQDAS that offers full Arabic language support. This software provides a powerful thematic analysis of the interview transcripts (Cassell and Symon 2004), which, as stated by DiCicco-Bloom and Crabtree (2006), also introduces a more efficient, flexible systemic analysis. Atlas.it provides the researcher with a variety of tools for examining, coding, relating, searching and classifying the interview transcripts (Cassell and Symon 2004). The coding tool in Atlas.it was used by the researcher during the interview transcripts thematic analysis to identify the Priori codes (which are the themes identified from the literature review) and the Nvivo codes (the new themes emerging from examination and quotation of the interview transcripts). As part of the systematic thematic analysis procedure, the researcher specified the Atlas.it tools used during the interview transcripts analysis.
There are numerous procedures to thematically analyse data, which vary in their similarity to each other (Cassell and Symon 2004; Hycner 1985). A systematic thematic analysis process reduces the effect of the researcher's own preconceptions and helps extract the researcher’s knowledge about
reality (Cassell and Symon 2004; Hycner 1985). The researcher adopted the analysis procedure from Braun and Clarke (2006) comprising the following six phases (see Table 13):
Table 13 Phases of thematic analysis (Braun and Clarke 2006, p. 87) ‘Permission to reproduce this table has not been granted’
Phase 1: Familiarising yourself with your data: As all interviews were conducted in Arabic, the interview recordings were first transcribed in Arabic. Then the Arabic transcription, Arabic interview recordings, notes and comments were imported to Atlas.it, where they were classified into sources. The researcher read the interview transcripts several times to further familiarise himself with the interviewee responses.
Phase 2: Generating initial codes: At this point, the researcher had read through the transcript many times. This allowed the researcher to identify themes and highlight them using the Atlas.it quotations tool.
Phase 3: Searching for themes: The quotations identified throughout the previous steps were clustered together under the themes previously identified (from the literature review) using the coding tool on Atlas.it. The researcher not only looked for the themes identified from the literature review but also any themes emerging from the transcripts. Moreover, the researcher clustered quotations that could not be grouped in the previous step under themes that emerged from the interview analyses using the Atlas.it coding tool.
Phase 4: Reviewing themes: The researcher reviewed every theme to make sure they related to the quotations linked to them.
Phase 5: Defining and naming themes: The researcher named the themes emerging from the interview analysis.
Phase 6: Producing the report: Using the Atlas.it report tool, the researcher reviewed reports based on the themes and interview participants.
4.5.4 Stage 3: Retroductive Analysis
In the third stage and final stage of the data analysis influenced by critical realism, the researcher used a retroductive backward reasoning of the experiences and events from the empirical and actual domains, respectively (see Figure 12) (observed during the first stage of the analysis), to
some of the mechanisms that helped cause them, or to some aspects of these causal mechanisms made manifest by the methods used (themes from the second stage of the analysis).
This was done by entering experiences and events, identified through the first stage by analysing the interviewees’ instructions to the double, into Atlas.it as codes (this was done manually in the first stage of the analysis and before conducting the in-depth interview with the practice observers). The coded experiences and events were linked to some of the causal mechanisms that helped cause them, or to some aspects of these mechanisms made manifest by the methods used (themes from the second stage of the analysis). From there, the researcher used the Atlas.it network feature to perform retroductive backward reasoning for the experiences and events from the empirical and actual domains, respectively (observed from the first stage of the analysis), to the to some of the mechanisms that helped cause them, or to some aspects of these causal mechanisms made manifest by the methods used (themes from the second stage of the analysis) (see Appendix I).