The empirical data was analyzed using thematic analysis method, in which the objective is to identify, analyze, and report themes that are recurring in data by using the following steps: getting familiar with the empirical data, creating pre-liminary codes, searching for themes, evaluating the found themes, defining and naming themes, and writing the report (Braun & Clarke, 2006). The first step in-cludes the transcription of the interviews and writing down the preliminary view.
Verbal description works as a foundation for data analysis in which the purpose is to discover regularities and patterns (Darke et al., 1998). Darke et al. stated that accurate transcriptions of interviews are important especially when the research is thesis.
In order to utilize the empirical data, data must be refined into more under-standable form. Qualitative data is rich and holistic by its nature, and it can be complex because qualitative data can provide expressive descriptions that are connected to real contexts (Miles & Huberman, 1994, p. 10). To make sense of the collected data, it was necessary to use methods that can refine the data. The col-lected data was quite rich, and many different topics emerged from each inter-view. Data reduction thrives to simplify the original data into more abstract di-rection (Darke et al., 1998). Data reduction is an important part of data analysis process where data, for example, transcription data, is simplified and focused into more abstract form from which conclusions can be drawn, and this can be achieved by concentrating, categorizing, and constructing data, and as well dis-carding excess data (Miles & Huberman, 1994, p. 10). Miles and Huberman con-tinued and specified that the actual data reduction during the data analysis phase includes many activities, for example, writing summaries, coding, forming themes, writing memos, and creating clusters and partitions.
The coding activity was an important part of the data analysis. Coding the data is a thematical analysis method for simplifying large amounts of data by defining codes for different parts of the data (Braun & Clarke, 2006). The tran-scription data was processed and coded using a qualitative analysis software At-las.Ti as Darke et al. (1998) suggested that special software facilitates data analy-sis. With the help of the coding, different themes started to emerge from the data as well as from the literature. The themes were identified using both of the exist-ing strategies: inductive strategy and theory-based strategy. In inductive strategy, themes are formed using the empirical data, whereas in theory-base strategy, themes are based on literature (Braun & Clarke, 2006).
The coding was not the only important part of the data analysis. Other anal-ysis methods were used as well. The case subject can be an organization and a single case study can have subcases (Miles & Huberman, 1994, p. 26). As the case company being the case subject, every customer interview was defined as a sub-case and the three sub-case company interviews formed one subsub-case. This enabled the use of two different case analysis strategies during different phases of the thematic analysis. Comparing cases using within-case and between-case
methods enable gathering knowledge on constructs in general and how these constructs relate to each other (Miles & Huberman, 1994, p. 27). Within-case anal-ysis facilitates processing one case and large amount of data, and enables the emergence of unique themes because part of the within-case analysis is to de-scribe a specific case and build a narrative for it (Eisenhardt, 1989). The cases should be compared to each other to reveal how the cases are similar or how they differ. The strategy for between-case analysis is to use categories and levels, in which similarities and differences are searched, which will form, for example, different categories (Eisenhardt, 1989). Categories and levels are similar concepts than themes. Between-case analysis increases the validity of a research and im-proves the chance to discover new findings from the empirical data (Eisenhardt, 1989).
Miles and Huberman (1994) describes the data reduction as an ongoing pro-cess throughout the qualitative research where the understanding of the study subject is formulated continuously (p. 10). This idea seemed to be valid after con-ducting the research. The data reduction was an ongoing activity throughout the data analysis phase. Another important part of data analysis is data display in which data is presented in an accessible and compact form, for example, graphs or charts (Miles & Huberman, 1994, p. 11). They added that data display supports the understanding of a reader towards text.
The data was analyzed iteratively, and during the first iteration the objec-tive was to familiarize with the data. The data analysis phase began with going through the data and coding different sections of the data using preliminary codes that mainly described the content. Also, the constructs from the conceptual framework were created as codes and coded to the data. Because the rich data was described quite precisely, over 500 codes were created. The use of the At-las.Ti was necessary because coding the data without a special software seemed to be impossible. After the first iteration, every main question and the conversa-tion after each quesconversa-tion were marked as one data block or one quotaconversa-tion as these were called in the Atlas.Ti. The relevant main constructs were assigned to each quotation, as well as the descriptive codes that were created. At this point, there were codes from three different levels. The first level was the top level of the main constructs which consisted of customer needs, service provider needs, transition, solutions, and benefits. The second level included the constructs that belonged to each first level main constructs. The third level included all the descriptive codes that were created.
The next iteration included different activities that were designed for form-ing a better understandform-ing regardform-ing the whole data. The data was reviewed through once again, and the coding was under assessment as well. Every quota-tion was shortened by leaving the quesquota-tions out. Each descriptive code received a prefix that indicated to which main construct the code belonged to. Codes that had similar topics were grouped together. Similar codes were merged together, and unessential codes were deleted. For each subcase, a short narrative was cre-ated to visualize each case verbally instead of the data and codes. These activities refined the data to some extent.
The third iteration included some of the same activities as the second itera-tion. The data was reviewed once again, and some codes were merged, some de-leted, and some renamed. Hierarchical networks were created to link every code to another. Before this, the hierarchy was indicated with simpler methods, for example, quotations included codes from the three different levels. Quotations were cluttered with too many codes, and this was solved with parting every quo-tation into smaller and more specific quoquo-tations. Only the essential information were left. There were still too many codes for each quotation, so the main con-structs were left out. At this point, every quotation only had the descriptive codes.
These measures simplified and clarified the data. There was a demand for some changes to the structure of the main constructs. The structure was changed, and new constructs were created to clarify the structure, hierarchy, and the content.
At this point, the main constructs were changed from the original conceptual framework. Networks were built for each subcase which made it easy to observe each case visually.
The fourth iteration included the same activities that had become basic at this point, for example, reviewing the data, merging, deleting and renaming the codes. There were still nearly 350 codes at the start of this iteration. The created networks were used to organize some of the codes more appropriately. Different themes started to emerge due to the previously done activities. A new level was created to the hierarchy. A theme level was created between the second level which consisted of the constructs and the third level which consisted of the de-scriptive codes. Networks were used to organize every fourth level code under an appropriate theme. The hierarchy and the links between the codes were up-dated. A new network view was created to present the first three levels. This view was the abstract visual result of the data analysis, which helped to understand the whole better.
The fifth iteration included reviewing the data and the networks, and mak-ing some adjustments. When everythmak-ing seemed to be in place, every fourth level code were merged into the appropriate third level code. This was done because the data was needed to be accessed through the themes which were the third level codes. Before this, every quotation had only level four codes. After this merging activity, there were left five main construct codes, 29 construct codes, and 78 theme codes. This was the end of the actual data reduction phase. By using Atlas.Ti analysis tools, a table was created that included every theme code occur-rence for each subcase. This helped to outline the prevalence and the significance of each theme. The data was ready for the reporting.
After the data analysis, the findings chapter was written using different themes that were created during the data analysis. Different sections of the data was easily accessed through these themes. When the data analysis was done properly, it facilitated the writing process of the findings. A new conceptual model was created to summarize and visualize the findings of this study as ear-lier suggested to use data display.
6 FINDINGS
This chapter outlines the most prominent findings of this case study. These find-ings are the results of analyzing all the eleven customer cases which are sup-ported by different views from the service provider interviews. The cases were compared to each other to find similarities and contradictions. In addition, the cases were compared to the service provider’s point of view to find out if the customers and the service provider have similar understanding regarding differ-ent matters that were gathered through interviews.
Next, I will present the findings in more detail using the framework’s five stages as a classification method: customer needs, service provider needs, transi-tion, solutions, and benefits. These five stages form the main chapters of the find-ings. The first chapter addresses different customer needs before purchasing a product and during the purchase process. Customer needs are important cata-lysts for customers to start searching for something that would fulfill their needs.
These needs also form the requirements for the purchase that must be realized by the service provider in order to enable the purchase. In the next chapter, these customer needs create needs for the service provider as they pursue to fulfill these different customer needs. These customer needs direct the needs of the ser-vice provider as the serser-vice provider is required to adapt for them to succeed in fulfilling different customer needs. In other words, the service provider needs to develop different aspects of their company to be able to fulfill these different cus-tomer needs. The third chapter addresses the adaptation and development that the service provider must perform to meet various customer needs and the needs of their own. The fourth chapter introduces different solutions that have been enabled through the process of need based adaptation and development. These solutions thrive to fulfill the various needs of the customers and the service pro-vider. The last chapter presents different benefits that the customers and the ser-vice provider can achieve through the various solutions.
In the findings chapter, the viewpoints and citations of the customers and the service provider will be presented using anonymity. It was agreed with the service provider that every interviewee in this research will be anonymous be-cause some of the interviewees requested anonymity, and only some of the
customers of the service provider are public, therefore, the service provider pre-fers not to disclose their customers.