Yin (2014) distinguishes two types of strategy for the analysis of qualitative case study research data. The first approach involves a pre-formulated theoretical proposition and a coding system. The other option is to develop a case description that would lead to emerging research questions and case study frameworks instead of them being pre-defined. As is often the case, the distinction is not as easy to make as is the case with this thesis. The interviews followed a semi-strict structure with all the participants being asked the same question only with slight alterations based on the participant's expertise and company area of business.
Therefore, emphasis was on pre-formulated prepositions, but the research framework and research questions were also refined during the process. There is a wide selection of tools available to analyze business-related case studies, but the most relevant techniques for this thesis were explanation building and cross-case analysis. Analysis of multiple case-study data usually begins with analyzing each case separately, followed by a cross-case analysis that aims to achieve a comparison through which similarities and differences are contrasted to the existing theory background. The single cases include a general description of the case arranged either thematically or in chronological order.
Explanation building is a technique that consists of a repeating search process to find causal links in the empirical data that are then presented in a narrative form. (Eriksson & Kovalainen 2008) According to Kähkönen (2011), one of the most critical phases in case research is the data analysis portion. In this research, the data was gathered by semi-structured interviews conducted during the winter of 2021. The discussions were recorded using Microsoft-Teams as a platform, and personal notes were also taken during the interviews. The recorded interviews were all later transcribed non-literally in the Finnish language. The transcription of the interviews produced 75 pages averaging 12,5 pages per interview. Initial intelligent verbatim transcription of the data acted as the first line of analysis. The research focused not on the participant's behavior and reactions, nor the research participant and researcher's interaction. The non-literal transcription also acted to simplify the process and enable a quicker analysis of the relevant data, thus acting as the research case record. The case record serves as the primary resource package for the empirical data where the data has been thematically assembled for better readability (Eriksson & Kovalainen 2008). The simplified transcripts were read numerous times. As a result, relevant notes and categorization of the data according to their relevance to the interview questions and themes had been made utilizing Microsoft Excel. Regarding themes from 1 to 3 and 5, the answers were categorized along the themes and questions presented in the interview form (see appendix 1). This categorization acted as the coding for the qualitative data. Thematic coding often begins at the data gathering process, as was the case here; thematic coding refers to features, instances, issues, and themes classified by assigning a label to them, often derived from the research theory (Eriksson & Kovalainen 2008). The transcriptions were analyzed per sentence, and answers were arranged under labels based on their relation to a specific theme and performed question. Unrelated explanations and answers were omitted to clarify the data further.
Regarding themes 3 and 4, the answers regarding or heavily implying relevant supply chain vulnerabilities and resilience capabilities were identified from the transcribed interview utilizing color-coding (red for vulnerability and green for resilience capability relation) and summarized into a table and arranged based on their prevalence (see appendix 3 and 4).
Any other relevant narrative was categorized and marked according to their relation to a specific theme and question. Important themes were pandemic disruptions, initial response measures, quantified pandemic effects, resilience significance, cost vs. risk exposure compromises, observed resilience gaps, and changes brought on by the pandemic. Observed vulnerability and resilience capability sub-factors were arranged under the most relevant categories and scored by assigning the value of “1”. This assignment required a sub-factor explicitly named in the interview or heavily implied to be a significant vulnerability source or a part of a resilience capability (see appendix 3 and 4). The details and answers were further analyzed based on their relation to the resilience capacity elements (readiness, response, and recovery) and resilience capacity approaches (control, coherence, and connectedness) (see appendix 5). The scores for the seven vulnerability categories and 14 resilience capability factors were calculated by adding the sub-element scores together (see appendix 3 and 4).
The participants were asked to rank respective vulnerabilities and resilience capabilities from important to non-important and highlight the most critical or important ones for the company's sake. This information was utilized to create a weighting system where the most critical element named by the participant was multiplied by the value of two, and the other highlighted factors were multiplied by:
2 − 1
𝑇ℎ𝑒 𝑛𝑢𝑚𝑏𝑒𝑟 𝑜𝑓 ℎ𝑖𝑔ℎ𝑙𝑖𝑔ℎ𝑡𝑒𝑑 𝑒𝑙𝑒𝑚𝑒𝑛𝑡𝑠.
The unhighlighted elements were multiplied by the factor of 1, meaning that their score represented the number of relevant sub-elements deciphered from the interview transcripts.
The weighting was utilized to ensure that the vulnerability and resilience capability elements were more reflective of the participants' perception and expertise than relying solely on the admittedly subjective scoring based on the mentioned or implied critical sub-factors for both vulnerabilities’ resilience capabilities. The scored and weighted vulnerabilities and capabilities were analyzed by utilizing the Spearman correlation coefficient (See appendix 3 and 4 for results) to find possible connections between experienced vulnerability and resilience capability factors. One of the research's key aspects was finding possible connections between supply chain vulnerabilities and resilience capabilities.
The comparison was achieved by combining vulnerability and resilience capability sub-factors under 7 and 14 main vulnerability and resilience factors, thus effectively scoring elements based on their qualitative research data prevalence. Vulnerability (V) sub-factors were spread under seven main categories (see appendix 4), while supply chain resilience capability (C) sub-factors were spread under 14 main categories (see appendix 5). The possible connections were hypothesized as follows:
Ho: There is little to no meaningful correlation between the vulnerability factor (VX) and the resilience capability factor (Cx) (−0,5 < 𝑝 < 0,5)
H1: The presence of the capability factor (Cx) lowers the presence of vulnerability factor (VX) (−1 ≤ 𝑝 ≤ −0,5)
H2: The presence of the capability factor (Cx) lowers the presence of vulnerability factor (VX) (0,5 ≤ 𝑝 ≤ 1)
However, it is essential to note that a seeming correlation does not mean any causality or a relevant link in either case. Explanation building based on the interview narrative was used to assess the observed capacities, capabilities, and vulnerabilities. The numeric data is purely based on qualitative data observation; thus, a high level of subjectivity is at play. To use the Spearman correlation coefficient, we must assume that the variables tend to change together.
The variables (Vulnerabilities and Capabilities) form a monotonic relationship as we cannot assume linearity between variables. The resilience capacity and the approach bias for the said capacity were evaluated both individually and as a group. The identified vulnerability and resilience capability sub-factors were placed in a matrix similar to Ponomarov and Holocomb (2009) (see table 2). This matrix was used to identify if the resilience bias was proactive, reactive, or recovery capabilities. The resilience approach was either through the supply chain -control, -coherence, or -connectedness approaches.
5 EMPIRICAL PORTION: SUMMARIZED FINDINGS
In this portion of the thesis, the empirical data is presented by first giving brief introductions of the participant companies and describing the summarized empirical data arranged by the interview form's themes.
5.1 Empirical Findings