3.3 Hypotheses development
4.3.2.1 Population and sampling
Recruitment of participants started at the survey stage. The retailers gave an expression of interest (EOI) by providing an email and a contact person name. The selection of the retail organisations was based on the fact that they have deployed IoT to capture data in logistics flow or their plans to do so. The selected retail organisations were approached via e-mail and invited to be a voluntary participant for this phase of the study. Twelve (12) retail industry participants were interviewed via purposive sampling, where the most suitable respondents from an identified organisation were approached to participate in the interview (Bryman 2015; Creswell & Poth 2017; Yin 2009). The participant selection aimed to include participants from retail firms across various sectors. The respondents represented medium and large companies and also both bricks-and-mortar, e-tail and multi-model organisations. Overall, the 12 participants held management positions and represented various retail subdivisions or sectors used in the survey study (ABS Retail industry analysis
129 2014; ABS_ANZSIC 2013). The interviewees profiles and their retail industry subdivisions are displayed in Table 4.6. Given that the purpose of interview data was to use as a secondary method to interpret/validate/explain the survey findings, interview data believed adequate for that purpose representing early and late IoT adopters, various retails forms and sizes, and all retail sub-divisions of the ANZSIC classification were collected (Flint et al. 2012). Target respondents were recruited from supply chain management and IT management departments of retail firms who possess technology-enabled supply chain management knowledge. As most of the retail respondents indicated the important role of third party logistics service providers in their respective supply chains fulfilling inbound and outbound delivery operations, and the extent IoT has transformed those logistics functions, a manager from a third-party logistics (3PL) service provider who services the retail industry was also interviewed to collect supplementary evidence on what IoT technology they have deployed and how that benefits their customers.
Table 4.6 The interviewees profile and retail subdivisions
Retail category ANZSIC
Subdivision
No. of participated firms
Restaurant, café, takeaway Subdivision 41 1
Supermarkets, grocery Subdivision 42 2
Household goods (e.g. hardware, furniture) Subdivision 42 1 Clothing, footwear and personal accessories Subdivision 42 1
Electrical, electronic, computer Subdivision 42 2
Pharmaceutical, cosmetic, toiletry Subdivision 42 1
Motor vehicles & parts Subdivision 39 1
Fuel and convenience stores Subdivision 40 1
Department stores Subdivision 42 1
Other Subdivision 42 1
Source: ABS (Australian and New Zealand standard industrial classification (ANZSIC) 2006)
4.3.2.2 Interview protocol and pilot testing
This study utilised semi-structured interviewing to help gain rich insights into the core themes. A list of pre-determined questions (interview schedule) guided the semi-structured interview process, with the added flexibility to accommodate questions conforming to the interview context (Brinkmann 2014; Kallio et al. 2016). Semi-structured interviews enable flexibility for the respondent’s spontaneous descriptions and narratives, while also offering
130 structure to capture respondents insights in a systematic manner (Denzin & Lincoln 1994; Yin 2009). The use of open-ended questions enabled the participant to engage in an open manner within the set framework, to bring up new facts and ideas throughout the interview process. It allowed the researcher to pose “how” “why” questions, but also to probe and explore knowledge areas that had not been anticipated. The questions were designed based on hypothesised relationships in the conceptual framework (Figure 3.1). The open-ended questions were focused on verifying the relationship between IoT capability, SCI and performance and on the specific enabling IoT technologies and practices.
The interview schedule (see Appendix H) had 19 questions under two sections which focused on general information about the respondents and IoT enabled supply chain practices respectively. Section 1 was mainly on information about the representative firm and the respondent’s background in the focal retail firm. Section 2 asked about when the respondent organisations first adopted IoT and their motives and drivers to do so. Following this, participants were asked to reflect on the IoT technologies deployed in-house, supplier- related and customer-related operations and their respective performance outcomes. Simultaneously, their future IoT deployment plans in their in-house, supplier-related and customer-related operations and expected outcomes were also examined. Then participants were directed to reflect on how IoT deployment in internal processes affected external relationships, and in turn, how IoT deployment in all three supply chain dimensions influence the performance of the supply chain and in turn the focal firm. IoT enabled supply chain performance outcomes were discussed in terms of traditional cost, quality, delivery, and flexibility dimensions and firm performance outcomes were probed under prevailing triple bottom lines in economic, environmental and social sustainability. The respondents were then asked about their exploitation of captured IoT data followed by what they think are the obstacles to IoT deployment and obtaining best outcomes from it. Finally, an open- ended question was asked on whether they had anything important to say that was missed during the interview.
The wordings of the questions were cautiously examined to reduce social desirability bias that may lead respondents to answer the questions favourably (Nederhof 1985). The draft interview schedule was first reviewed by the supervisors. Then a discussion was held to evaluate the appropriateness of the questionnaire, where the length, the breadth and the word expressions were examined. The revised interview schedule was further tested for appropriateness through pilot interviews ahead of the target respondents. Pilot interviews conducted face-to-face with three managers in the retail industry verified the relevance of
131 the questions, and the reliability and validity. In order to minimise bias during the pilot interviews, commenting on responses and providing clarification unless otherwise asked was avoided. The respondents were not interrupted unless the answers were exceedingly prolonged or far deviating from the focus. The same consistent approach was adopted for the final interviews. The pilot interviews lasted for between 45 minutes to 1 hour and audio recorded. Subsequently, feedback on the clarity of questions and suggested improvements were provided by the subjects of the pilot study. Minor improvements to the interview schedule were made through the feedback. The feedback was used to revise and improve the final questionnaire. The interview schedule finally employed in this study is attached as Appendix H.
4.3.2.3 Arranging and conducting interviews
Recruitment of participants was conducted at the survey stage where the interested organisations completing the anonymous survey were asked to provide an email and a contact person for further investigation if interested. The selection of the retail organisations was based on the fact that they have deployed IoT to capture data in logistics flow or their interest and plans to do so. The selected retail organisations were approached via the provided e-mail to invite to be a voluntary participant for this phase of the study. In most instances, the initial contacts themselves volunteered to be interviewed or another volunteer was nominated (after passing gatekeepers). Negotiation with gatekeepers such as CEOs, Heads of HR and Directors was carried out in some cases, when an intervention for organisational permission process was required, but was mainly left to the discretion of the voluntary participants who are top level employees of the firm (Bryman 2015; Denzin & Lincoln 1994; Yin 2009). Once consent was granted, the formal information sheet for individual interview participant organisations (Appendix C)was sent. The information sheet included the research objectives, purpose of the interview, how long it would take, potential research contributions and anonymity guarantees. Management scholars have reasoned that, to get a higher chance of acceptance from potential participants, the time and resource requirement should be kept to the minimum feasible (Easterby-Smith, Thorpe & Jackson 2012). Hence, the interviews were limited to approximately an hour and only a single voluntary informant was invited from each organisation.
The time and place for the interviews were scheduled at the informants’ convenience. All interviewees were provided with a hard copy of the information sheet for individuals
132 (Appendix C). All informants confirmed consent by signing the consent form (Appendix G) before conducting the interviews. The nature of research, potential benefits and related risks were outlined to the informants in an explanatory statement, who were also advised to gain consent from managers. They were advised of the choice (informed consent) of participation and withdrawal (Corbin & Morse 2003). Privacy and confidentiality of respondents were assured.
With permission, the interviews were recorded using a digital audio application on a smart phone and uploaded to a secure drive. The recordings enabled the researcher to corroborate the reliability of the interview data. However, if the informants expressed any reluctance about being recorded, a contingency plan was to make extensive notes of the interviews. To make the interviewees comfortable, to develop rapport and initiate the discussion, the first question centred on their organisational role. Leading questions were avoided to allow free flow of discussion (Denzin & Lincoln 1994; Richards 2014). Field notes were taken on the emerging themes. Interview process lasted from mid to end 2017.
4.3.1.4 Qualitative data analysis
Following the transcription of the 13 interviews, the software application Nvivo 11 was used to help the thematic analysis process. The field notes, websites, and in some cases, follow- up email and phone conversations supplemented the interview data.
A data reduction process was conducted, using an inductive approach to transform data into orderly and simplified themes to develop a meaningful outcome. Interviews were analysed using a typical open-coding process used in qualitative research (Creswell & Poth 2017; Strauss 1987). This process involved reducing and categorising the text into meaningful segments and labelling them with an appropriate title to best define the material (Creswell & Poth 2017; Yin 2009).
The code formation is contingent on the understanding and interpretation of the data (Richards 2014). Coding assists in understanding patterns through data abstraction. Three types of coding are found in typical qualitative analysis: descriptive, topic, and analytic coding (Richards 2014). Information on characteristics or attributes are denoted in descriptive codes and used to accrue demographic information as they were important to investigate patterns. Here “in Nvivo” coding (where the exact words of the respondents are used to label the code) was used frequently. Topic coding is classifying as per the key theme. Analytic coding is used to reinforce an emerging theory or to confirm theory and denote the
133 reflections on the subject (Richards 2014). The analyses involved all three key coding types. A new code was created in any case where the content slightly deviated from the existing code.
Subsequent to the primary coding process, transcripts of each were examined to identify new themes, revise existing codes and establish sub-categories (Richards 2014). The codes were reviewed and, in some cases, consolidated to avoid repetition. The identification of key themes involved consolidating interrelated codes into broader groupings. Patterns and relationships were identified across the data (axial coding). The interview transcripts were combined with field notes at this stage and, when it was feasible, coded data were verified by web information (e.g. firm size).
The findings from the interviews helped gain an improved detailed understanding of the relationship between the study variables and extent of IoT deployment in the Australian retail industry to improve SCI for greater performance and other vital information on IoT deployment. The findings of the qualitative phase of the study are detailed in Chapter 6. The discussion of findings in Chapter 7 integrates the analysed outcomes of the survey with the insights gained from the case studies (Flint et al. 2012).
4.3.1.5 Trustworthiness
Trustworthiness covers characteristics such as credibility, transferability, dependability and conformability. Therefore, it is fundamental to be thorough with the methods and the results (Bryman 2015). Credibility of the study was confirmed at a later point through triangulation of quantitative and qualitative findings. Dependability was ensured by keeping complete records of data collection and analysis. Conformability was assured through good faith by dismissing personal values and theories. A high-quality digital recorder on a modern smartphone was used for audio-recording and transcribing the interviews, to ensure the data quality.
Semi-structured interviews could be subjected to data quality issues such as biases by the respondents (Saunders 2011). Care was taken to avoid bias at the design stage . Leading questions were avoided while conducting interviews. The innovative technology related interview responses can result in social desirability bias, as respondents may like to report that they are progressive. Therefore, data triangulation was utilised to reduce the effect of these biases and to improve the data accuracy (Denzin & Lincoln 2008). Triangulation is
134 explained as the use of multiple information sources to balance out subjective influences to understand a phenomenon (Creswell & Poth 2017; Flick 2004).
The reliability of coding is vital as the coding process can be extremely subjective and dependent on the researcher’s comprehension and interpretation. Therefore, inter-rater reliability was tested (Richards 2014). Inter-rater reliability tests are a way of comparing the uniformity of the codes created by two different individuals on the same transcript (Richards 2014). A second coder with no direct role in this study was used to verify the codes. A researcher with extensive qualitative coding experience coded transcripts for 4 interviews, and compared codes for discrepancies, to confirm a general agreement in coding consistency.
4.4 Chapter summary
The methodology applied in this study is explained in this chapter. The study philosophies, ontologies, epistemologies and methodologies were rationalised. The employed mixed methods approach was justified and the two phases adopted in this study were explained in detail. The steps followed to collect quantitative data via a survey of Australian retail organisations, in terms of scale development, questionnaire preparation and instrument validation techniques were discussed. The on-line questionnaire was sent to a random sample of retail organisations, and 231 complete responses were received. After initial screening, the data set of 227 cases were tested using the two-step approach of structural equation modeling, where the measurement model was tested before the structural model. The chapter detailed the SEM methodology, the model fit, reliability and validity tests and hypothesis testing. A quantitative survey study was followed by a series of semi-structured interviews in a qualitative case study phase. 12 retail industry managers and a manager from a 3PL service provider was interviewed. The qualitative data was first transcribed and then thematically analysed with the help of Nvivo 11 qualitative analysis program, where data reduction process transformed data into orderly and simplified themes. Data was analysed simultaneously, hence the next two chapters discuss the findings of the two phases together in an overall findings and discussion.
135
Chapter 5
Quantitative research findings
5.1 Introduction
The finding of the quantitative survey data is discussed in this chapter by explaining the stage by stage analysis procedure. It focuses on the measurement and structural models in SEM to validate the hypotheses developed in Chapter 3.
The preliminary analysis in section 5.2 discusses the details of the screening of the survey data for inadequate, unengaged, mechanical or unanswered responses, followed by missing value analysis. The assessment of univariate and multivariate distribution normality and outliers for each measurement item is followed. Then, in an attempt to justify the suitability of the study sample, the demographic evidence is scrutinised in section 5.3, under the characteristics of the organisation, their deployment of IoT in supply chain operations and the respondent’s suitability for the study. The descriptive statistics for the measurement items are examined in section 5.4. Subsequently, the procedure for testing the measurement model is discussed in section 5.5. The measurement models for each construct were validated disjointedly by applying exploratory factor analysis (EFA) and confirmatory factor analysis (CFA), prior to testing the full measurement model. The model fit of measurement models was assessed by applying CFA and using goodness-of-fit indices to guide towards model-fit. The reliability and validity of the final measurement model were confirmed in section 5.6 prior to structural analysis of the theoretical model in section 5.7. The structural model was also validated by the goodness-of-fit indices ahead of verifying the study hypotheses. To further validate the theoretical model, the effect of control variables on the structural model was tested to exclude any confounding effects and post hoc analysis was conducted to conclude that no model re-specification was possible.
The chapter concludes with a summary of the key survey findings.