This chapter outlines the decisions taken regarding the research design and the motivations behind those decisions. The research process for the data collection and data analysis are described along with aspects regarding validity and ethics.
The research design of a study explains the means which the study used to get from the research questions to the concluding remarks that answers them (Yin, 2014). In this study, the purpose was to examine how the organisational compliance with ambidextrous elements is perceived at different hierarchical levels in manufacturing SMEs, in order to increase the understanding of organisational ambidexterity in this type of organisations. Based on the nature of the research questions and in order to fulfil the purpose of the study, there was a need to study and investigate various individuals’ perceptions, hence, a qualitative research approach was applied. Agostini et al. (2015) points out that organisational ambidexterity obligates a qualitative research design due to the inherent complexity that could not be captured and understood by any other research design. The empirical data that was needed to answer the research questions was collected through two single case studies, in which both qualitative and quantitative techniques were used when collecting the data. The techniques used were a combination of questionnaire and interview, according to Williamson’s (2002) definition. The usage of both qualitative and quantitative techniques for the data collection was another argument for the usage of case study design as method in this study (Williamson, 2002).
The two single case study design allowed the phenomenon to be studied closely, hence creating an in-depth understanding needed to fulfil the purpose of the study (Bryman & Bell, 2015; Williamson, 2002). By using two single case study design the findings deriving from each of the cases could be compared both within and between the cases (Bryman & Bell, 2015). Using combined questionnaire-interviews resulted in both qualitative and quantitative data. The usage of both quantitative and qualitative data in qualitative research is supported by Bryman and Bell (2015) which states that qualitative research does not completely consist in the absence of numbers. By using two different techniques, it enabled the avoidance of too immense a reliance of one single approach and increased method triangulation (Knights & McCabe, 1997; Williamson, 2002). Method triangulation is defined as the usage of more than one technique or source of data when studying a social phenomenon, thus, the results can be compared and the data can complement each other, resulting in a greater confidence of findings (Bryman & Bell, 2015).
Triangulation is commonly associated with quantitative research, but Bryman and Bell (2015) advocates that triangulation also can take place in qualitative research. The selection of the two cases was based on that both companies were manufacturing SMEs that participated in a research project at Jönköping University, and that both companies had an interest for the research questions and the purpose of the study. The case companies are described further in chapter 4.
The research was conducted with deductive reasoning, starting from studying the theory and then studying the phenomena and collecting the empirical data (Bryman & Bell, 2015). The theory consists of two frameworks: (1) structural ambidexterity presented by O’Reilly and Tushman (2011) and (2) contextual ambidexterity by Birkinshaw and Gibson (2004). Based on these two frameworks a combined analytical framework was developed which guides the research process. The research process is illustrated in Figure 6 and is described in the following two subchapters.
To capture the compliance with the ambidextrous elements at different levels, each element was broken down into one or several statements. By breaking down the elements into statements, ambiguous double-barrelled questions were avoided (Bryman Bell, 2015). This process resulted in 14 statements (see appendix 0). In the theory, the contextual elements refer to the individuals’ capabilities within a company such as initiative taking and multitasking. The statements for the contextual elements were formulated in a way that assess whether the company allows the individuals to execute the capabilities that are defined in the contextual element, i.e. if the company allows the individuals to take initiatives. All the structural elements were already assessing the company and not the individuals’ capabilities, hence those statements were mainly formulated in a way that avoided double-barrelled questions. The interviewees were then asked to indicate their perception of the statement, thus assessing the company’s compliance with the ambidextrous elements on a Likert scale followed by a motivation of their answers. Thus, the data collection techniques for this study were a combination of questionnaire and interview, according to Williamson’s (2002) definition. The reason for using combined questionnaire-interviews was to get quantitative data regarding the compliance of the ambidextrous elements, and simultaneously get nuanced qualitative explanations regarding their indications. The Likert scale included a 7-points scale, as seen in Table 2. A Likert scale is commonly a 5- or 7-point scale, where the interviewees have the option to choose a neutral midpoint in case they feel insecure (Bryman & Bell, 2015; Krosnick & Presser, 2010). Although the validity and reliability do not differ notably between 5- and 7-point scales (Krosnick & Presser, 2010), a 7-points scale provides more alternatives for the interviewees which in cases where they are able to differentiate between several alternatives a bigger scale allows the interviewees to indicate the answer most corresponding to their perception (Krosnick & Presser, 2010).
After indicating their answers to each statement on the Likert scale, the interviewees were asked to give qualitative motivations and give examples about their perception of the compliance with the element. The combined questionnaire-interviews were carried out in a semi-structured manner, they followed the structured order of the questionnaire, but the follow-up questions were unstructured (Williamson, 2002). The questionnaire-interviews with the managers were carried out with one interviewee at the time while the questionnaire-interviews with the workers were carried out in groups of two to make the workers feel comfortable and be able to help each other to associate the statements to their work situation. However, the workers were asked to first indicate their perception regarding the statement individually and then discuss their answers together. The questionnaire-interviews were carried out in Swedish, the native language of the interviewers and most of the interviewees with the purpose to make it easier for the interviewees to understand the statements and express themselves without being hindered by their language skills. Even though this meant that all statements had to be translated from English to Swedish without losing its essence.
Moreover, the key terms exploration and exploitation were anticipated to be difficult to grasp for all interviewees. Hence, they were defined prior the questionnaire-interviews to ensure that all interviewees understand the core of the research and uses the same definitions, which facilitates for comparison between the answers. All definitions and all statement could be read by the interviewees on the questionnaire, in addition to the interviewers reading them aloud. The questionnaire-interviews started with a brief introduction to the study and the definition of the key terms, hence the interviewees answered personal factual questions to give information about their role and background, before finally starting to answer the statements (Bryman & Bell, 2015). To capture as much data as possible during the questionnaire-interviews and in order to have the possibility to go back and control the answers they were recorded. As Bryman and Bell (2015) and Williamson (2002) empathises the importance of during interviews. The questionnaire-interviews were conducted by two interviewers in order to be able to ask follow-up questions (Bryman & Bell, 2015; Williamson, 2002).
The interviewees were chosen based on theoretical sampling, where the suitable candidates were not chosen on a random basis, but selected according to their role and perceived suitability for the study (Bryman & Bell, 2015). Eisenhardt and Graebner (2007) mean that the selection of highly knowledgeable participants, such as actors from different hierarchical levels and units is a key approach to limit bias in interview data. All the interviewees and their hierarchical classification in this study are found in Table 3, the hierarchical classification is made based on the specific context of the case company.Table 3 Interviewees roles and hierarchical classification in this study
The data consisted of quantitative data in terms of the numbers indicated on the Likert scale. The data also consisted of qualitative data in terms of oral descriptions and motivations given by the interviewees for how and why they did the indication they did on the Likert scale regarding each statement. To facilitate the analysis of the quantitative data, the data was visualised in a boxplot, see Figure 7 for example of these. The box in the boxplot visualises the middle 50 per cent of the interviewees and the upper and the lower boundaries of the box visualises the interviewees that indicated the highest and the lowest within the middle 50 per cent (Bryman & Bell, 2015). The line that crosses the box visualises the median and the cross is the mean (Bryman & Bell, 2015). The ends of the vertical line that go downwards and upwards from the box visualise the interviewees who indicated the lowest and the highest of all interviewees (Bryman & Bell, 2015). By using boxplot, the dispersion and the mean of the interviewees for each statement can be visualised (Bryman & Bell, 2015). To make sense of the dispersion of the answers, the data had to be classified in a systematic way. If the answers among the interviewees in different groups only differed by 1 number in a numerical order it was regarded as no difference. E.g., when one interviewee indicated 6 and another one 7 on the Likert scale, as seen in Table 4. This resulted in that the perception of compliance with the ambidextrous elements at different levels in manufacturing SMEs could be analysed, which answered the first and second research questions of this report.
The answers from the interviewees were transcribed and translated from Swedish to English. By transcribing all the data, it was possible to capture not just what the interviewees had said but also the way they said it, which is of interest when studying people’s perceptions (Bryman & Bell, 2015). When the data was transcribed and translated, the analytical framework (see Figure 5) could be used to guide the analysis of the data and the different interviewees divided into its hierarchical levels. Since the elements were broken down into statements, the answers regarding all statements could be clustered into their respective element before the analysis begun. During this analysis it became clear that the manager level that was expected to constitute one level had to be divided into two levels as the answers from these two levels showed different patterns at both case companies. Hence, this level was divided into middle managers and top managers. When the data had been analysed separately for the different hierarchical levels for both Plastic Inc. and Metallic Inc., the differences was first analysed at different levels within each case and then analysed between the cases and the different levels.
To compare the perceptions of structural element and contextual elements in the discussion, an average for each group’s indications to the elements (including all statements for the element) was calculated. Hence, the averages were colour coded to mark the group’s common perception. The Likert scale was divided in three sections and given colours according to Table 5. A coloured line was used to visualise the hierarchical levels’ perception of the compliance with the ambidextrous elements. By analysing and comparing the quantitative data from the questionnaires and the qualitative data from the interviews, method triangulation was conducted. This increased the confidence of the findings since the qualitative data provides a more detailed picture while the quantitative data provides a narrower and more focused picture of the statements (Williamson, 2002). The comparison and analysis of the data between the cases enabled the researchers to identify or cross-check the data and look for consistency of the data between the different cases and levels. During this comparison it became clear that both case companies presented similar patterns which was that the manager levels differentiated from each other, which increased the source triangulation as this result showed on consistency of the data derived at two different case companies (Williamson, 2002).
The validity is regarded as one of the most important criteria of research (Bryman & Bell, 2015). However, there are ongoing discussions regarding the relevance of the different criteria used for assessing the quality of qualitative research (Bryman & Bell, 2015). This is due to that the criteria for assessing the quality of the research mainly are dominated by positivistic/quantitative inspired logics (Halldorsson & Aastrup, 2003). Halldorsson and Aastrup (2003) argue that qualitative research which is commonly dominated by naturalistic/qualitative aspects should be assessed according to a qualitative alternative in order to avoid a misfit. Trustworthiness is a parallel criterion for naturalistic/qualitative research which includes credibility, transferability, dependability and confirmability (Guba & Lincoln, 1989). The usage of both qualitative and qualitative data assures the credibility in this study as it increases triangulation by the avoidance of too immense a reliance of one single approach and the results can be compared which is one technique for ensuring credibility (Bryman & Bell, 2015). Due to the operational focus of this work, this study only investigates the compliance with the ambidextrous elements within the manufacturing unit, excluding other support functions, at manufacturing SMEs. However, by investigating other functions the credibility of this study could have been ensured further.
The second criterion is the transferability and refers to the generalisability (Guba & Lincoln, 1989). However, true generalization is not possible, as both time and space changes both the context and the individuals in it, thus creating constraints in generalising the findings (Erlandson, et al., 1993). Although, the knowledge attained from this report and its findings can be applicable for other manufacturing SMEs that constitutes a similar context as the case companies (Erlandson, et al., 1993). The difficulties with the different definitions was anticipated, hence the definitions and examples of exploration and exploitation was attached on the questionnaire that the interviewees filled in during the questionnaire-interviews. Despite the attempts to mitigate the risk, the interviewees tended to use their own individual interpretations of exploration and what activates that could be related to exploration. Such a context dependent vocabulary may impact the direct transferability of the findings if one is not observant on other contexts’ potential differences in how the concepts are perceived. However, that the interviewees have their own individual interpretation is in itself a finding since it shows a gap between the industry and the theory of the field within organisational ambidexterity that shows that the definition of exploration and exploitation varies. One way of decreasing this could have been to acquaint the interviewees with the different terms and their definitions during a session before the questionnaire-interviews as this could have decreased the risk for individual interpretations. Due to the diverse group of interviewees it was difficult to assure that the terminology would be understandable to all interviewees, and some terms such as strategy and trade-off posed a challenge for some interviewees. This can also have an impact on the direct transferability of the findings. To minimize this, the questions regarding any of the terms was explained in a similar way to the interviewees to avoid bias by the researchers. Another factor that could have an effect on how the interviewees answered is the Hawthorne effect. The Hawthorne effect concerns the effect of being the studied subject in an experiment, which affect the subject into wanting to answer or express themselves in a way that is according to what the subject thinks that the study aims to achieve (Bryman & Bell, 2015).
Dependability is the third criterion for quality assessment of qualitative research and is a parallel to the conventionally term of reliability (Halldorsson & Aastrup, 2003). Dependability concerns the stability of data and that the researcher keeps all records during all phases of the research process (Bryman & Bell, 2015). Dependability was ensured by maintaining all records and documents from the questionnaire-interviews and the data collection during the whole research process. Confirmability is the fourth and final criteria for assessing qualitative research, and concerns the objectivity of the researcher and that the research was conducted in good faith (Bryman & Bell, 2015). In this study, confirmability was ensured by demonstrating how the findings can be confirmed and presenting the sources which all conclusions and interpretations are based on.
It is the responsibility of the researchers to carefully assess the ethical aspects of their research (Bryman & Bell, 2015). Diener and Grandall (1978) have defined four main areas of consideration to asses when conducting business research. These areas are harm of participants, lack of informed consent, invasion of privacy and the involvement of deception. To ensure that this report fulfils and take these ethical considerations in regard different actions have been taken. The first area concerns the harm of participants. To ensure that no harm can happen to any of the participants in this study, the identity of all participants and both companies is confidential. The purpose with omitting the names of the interviewees and the companies is to protect the interviewees and the company’s anonymity. However, since the purpose of this study was to investigate how the compliance with the ambidextrous elements is perceived at different levels in manufacturing SMEs, the hierarchal level of the interviewees within the companies was of interest and could therefore not be left out in this study. Instead, the names of the interviewees were replaced with their official titles to differentiate the different levels of the interviewees. Those interviewees that expressed consent when being asked if he or she allowed the questionnaire-interview to be recorded, were informed that the recording will be accessible to the researchers only. The second area defined by Diener and Grandall (1978) concerns the lack of informed consent. All participants in this study have been informed of the purpose and the nature of the research before being asked if they want to be involved. The third area concerns invasion of privacy. Invasion of privacy is very much linked to the previous area, lack of informed consent (Bryman & Bell, 2015). If the participant is aware of the details of the research and gives his or her consent to participating in the study, the participant also acknowledges that this might intrude to the right to privacy for a limited time, as the participant agrees to be involved in the research. Some topics can be judged as sensitive to everyone on beforehand and therefore be handled sensitively, although it is not always possible to foresee beforehand (Bryman & Bell, 2015). Therefore, each case in this study was treated individually and sensitively which according to Bryman and Bell (2015) is a way to manage invasion of privacy during interviews. The fourth area defined by Diener and Grandall (1978) concerns the involvement of deception. Deception takes place when the researchers present their research as something other than what it is (Bryman & Bell, 2015). Before the interviewees in this study were asked if they gave their consent to be involved in the study, they were given the information about the nature of the research and its purpose.
Table of Contents
1.2 PROBLEM DESCRIPTION
1.3 PURPOSE AND RESEARCH QUESTIONS
1.5 OUTLINE .
2 Theoretical background
2.1 ORGANISATIONAL AMBIDEXTERITY
2.2 THE LEADERSHIP ASPECT OF ORGANISATIONAL AMBIDEXTERITY
2.3 STRUCTURAL AMBIDEXTERITY
2.4 CONTEXTUAL AMBIDEXTERITY
2.5 ANALYTICAL FRAMEWORK.
3 Research methodology
3.1 RESEARCH DESIGN
3.2 RESEARCH PROCESS
3.3 RESEARCH QUALITY
4 Company descriptions
4.1 METALLIC INC.
4.2 PLASTIC INC.
5.1 STRUCTURAL AMBIDEXTERITY
5.2 CONTEXTUAL AMBIDEXTERITY
6.1 RQ1: HOW DO DIFFERENT HIERARCHICAL LEVELS IN MANUFACTURING SMES PERCEIVE THE COMPANY’S COMPLIANCE WITH THE AMBIDEXTROUS ELEMENTS RELATED TO STRUCTURAL AMBIDEXTERITY?
6.2 RQ2: HOW DO DIFFERENT HIERARCHICAL LEVELS IN MANUFACTURING SMES PERCEIVE THE COMPANY’S COMPLIANCE WITH THE AMBIDEXTROUS ELEMENTS RELATED TO CONTEXTUAL AMBIDEXTERITY?
7.1 DISCUSSION OF FINDINGS
GET THE COMPLETE PROJECT
Organisational ambidexterity in manufacturing SMEs An empirical study of managers’ and workers’ perceptions of ambidextrous elements