Data analysis

After the collection of empirical data, the data analysis, interpretation and careful drawing of conclusions can be seen as the most important and critical stages of the research (Hirsjärvi et al. 2009: 221). The data analysis refers to careful reading, organizing, classifying, outlining and deliberating of the empirical data, and aims to make sense of the content or structure of the data while considering the research problem of the study. The analysis is conducted by interpreting the empirical data and discussing and reflecting it with the prior theory and researcher’s own thinking. Thus, it involves consideration of the studied phenomenon and the research questions from a specific viewpoint. (Saaranen-Kauppinen & Puusniekka 2006.)

Hirsjärvi et al. (2009: 223) suggest that the processing and analysis of data should start as soon as possible after the data collection, whereas Saunders, Lewis & Tornhill (2009:

485) highlight that the process of data analysis generally starts simultaneously as collecting the data and continues later on. In this research, the analysis of empirical data was initiated by transcribing the recorded interview data into a written format almost immediately after each interview. Transcribing the recorded data into a written format is suggested to facilitate the organization and analysis of the data (Saaranen-Kauppinen &

Puusniekka 2006). Transcription was first conducted as word for word and included the entire recorded data gathered through the interviews. Saaranen-Kauppinen &

Puusniekka (2006) note that the exactness of the transcription is affected by the chosen type of analysis. Regarding this thesis, the interest lies in the content of the empirical data to explain the phenomenon of sustainable supply management in Finnish SMEs in detail, rather than in the expressions or used language. Thus, the transcribed data was later cleaned up and for instance unnecessary expletives were removed to make the data

more readable and easier to organize. These transcribed interviews were then sent to the respondents to ensure the factual accuracy of the data. Preliminary analysis was initiated as transcribing the data by listening, writing and reading the interviews several times as well as by outlining the written material considering what is relevant regarding the research problem and questions of the study.

The data analysis approaches can be divided into data-driven analysis, theory-driven analysis and theory-bonded analysis (Tuomi & Sarajärvi 2009: 95–100; Saaranen-Kauppinen & Puusniekka 2006; Eriksson & Kovalainen 2008: 128–129). In data-driven analysis, units of analysis are chosen based on the empirical data considering the aim of the research, and the prior observations, knowledge or theories should not influence on the execution of the analysis. Theory-driven analysis, on the other hand, relies on a specific prior theory or model that guides the analysis of the data, and the aim is to test prior knowledge in a new context. Theory-bonded analysis can be placed between these two extremes, and is characterized by some theoretical linkages. In this approach, the units of analysis are chosen from the data but the prior theory may assist in the progress of the analysis, and as the data is categorized and conceptualized it is linked with the theoretical concepts. The theory-bonded analysis often relates to the abovementioned abductive logic, in which the researcher aims to combine the data and the prior theoretical models. (Tuomi & Sarajärvi 2009: 95–100.) This research applies the theory-bonded analysis approach, which is in line with the above justification of abductive research logic of the study.

The data analysis in this research is conducted as a qualitative, theory-bonded content analysis, which aims to study the phenomenon systematically and objectively, and produce a general description of it. Content analysis pursues to analyse the textual data and seek meanings of it through interpretation and reasoning. (Tuomi & Sarajärvi 2009:

103–108, 112.) The aim of the content analysis is to describe the studied phenomenon in a condensed form and to link the research findings with the wider context and with findings from previous studies (Saaranen-Kauppinen & Puusniekka 2006). The analysis also seeks to clarify the data so that it is possible to produce explicit and reliable conclusions about the studied phenomenon (Tuomi & Sarajärvi 2009: 108).

The content analysis is initiated by splitting the empirical data into small pieces, which are then conceptualized, grouped and finally restructured into a logical entity (Saaranen-Kauppinen & Puusniekka 2006; Tuomi & Sarajärvi 2009: 108). The transcribed interview data is examined by classifying, seeking of similarities and differences as well

as by compressing of data (Saaranen-Kauppinen & Puusniekka 2006). The data analysis in this research is based on the presentation of Tuomi & Sarajärvi (2009: 108–113) about the process of data-driven content analysis. The authors note that the theory-bonded content analysis proceeds as the data-driven analysis relying on the empirical data, but differs in a way in which the empirical data is combined with the theoretical concepts as the data is abstracted. In theory-bonded analysis, the applied theoretical concepts of the phenomenon are derived from the prior theory. (Tuomi & Sarajärvi 2009: 117.)

The data analysis started by recognizing the issues and phrases in the empirical data that are relevant considering the research question and objectives of the study. These expressions were then simplified through coding, which means splitting the data into smaller pieces (Saaranen-Kauppinen & Puusniekka 2006), and labelling these expressions to facilitate the grouping and organization of the data (Saunders et al. 2009:

492). These codes and simplified expressions were then gathered as lists from which similarities and differences of the codes were observed and analysed. Similar codes were then categorized into groups, which can be considered as subcategories, and were then labelled accordingly. The analysis was continued by combining similar subcategories with the same content, which led to formulation of the main categories.

The abstraction and conceptualization of the original expressions occurred as the analysis proceeded. (Tuomi & Sarajärvi 2009: 101, 108–113.) The subcategories were formed based on the expressions and findings from the empirical data and were then combined with the theoretical concepts deriving from the prior theory by formulating the main categories. These theoretical concepts that already guided the theme interviews with the company representatives, helped to describe and analyse the central features of the empirical data (Eriksson & Kovalainen 2008: 129). Finally, all the main categories were further combined into one connective category that depicts all the abovementioned categories. These categories will eventually assist in answering to the research question and objectives of the study. (Tuomi & Sarajärvi 2009: 101.) The progress of the content analysis is illustrated in the Table 4 below with extracts from the empirical data.

Table 4. The progress of the content analysis (revised from Tuomi & Sarajärvi 2009).