Techniques and procedures

The sample technique was non-probability, purposive heterogeneous, since it allowed for key themes to be described and with a small but varied sample, emerging patterns could carry more weight. Non-probability was appropriate since no statistical inferences had to be made and purposive heterogeneous was fitting due to the focus on key themes, the phenomenon at hand, and since it enabled a variety of perspectives on the key themes, making patterns more valuable (Saunders et al., 2007). For heterogeneity, sample selection criteria were established. The sample was split into managers and diversity consultants, coming from each business sector (public, private with public shares [PLC], private with private shares [LTD] and charitable).

Nine participants took part in semi-structured ethnographic interviews, just above McCracken’s (1988) threshold of eight; allowing me to reach saturation of information. The number of interviewees may depend on the specific research and methodology (Baker & Edwards, 2012) and in this case, nine was applicable. More details on the interviewees is provided later, but it can be mentioned the interviewee range was somewhat limited, due to circumstances, but each candidate was relevant for participation. All interviewees had expertise with at least one of the concepts being covered, if not all. It was ensured that the interviewees selected could still, as a collective, provide a range of perspectives, which in turn allowed for richer data to be collected.

4.7.2 Data collection instruments

Secondary data established a theoretical base, through documentary data as journal articles, across the concepts of the glass ceiling, change agents and managers’ roles in it all (Saunders et al., 2007). This bank of understanding on the phenomenon provided further direction for primary research, for which, semi-structured interviews were employed. A list of interview questions was drawn up and, in some cases, shared with interviewees before the interview. The interview was more conversational where possible, in line with the ethnographic aspect. The question list consisted of open-ended questions for a full answer range and was used as an interview guide rather than the exact order being followed (Ayres, 2008a). The full guide is in Appendix A. Semi-structured interviews allowed for concepts to be properly explored, with a range of experts’ views providing a thorough comprehension of phenomena and relevant areas.

4.7.3 Data collection procedures

The secondary data collected was existing theory, found through online journal article databases. Once selected, Mendeley software stored and organised the articles, necessary for the volume. The primary data collection procedure was semi-structured interviews, chosen on the basis of ethnographic, exploratory and inductive research. Semi-structured interviews allowed for rich qualitative data collection, as desired. The interviews were carried out from 04.04.2020 to 01.06.2020, over telephone and video calls via Skype, Zoom and Microsoft Teams, due to travel restrictions at the time of data collection. Audio, and where applicable video, of the interviews was recorded for transcribing, to collect all information for analysis.

On the selection criteria, managers were interviewed for the perspective from a management position and since they could be a change agent. Diversity consultants were interviewed for the specialist perspective on changes regarding discrimination, diversity, equality and inclusion.

The list of interviewees included below (see Table 4.1) covers the interviewee titles, company they work at the time of research and where the company is in terms of selection criteria. All interviewees were kept anonymous based on the requests of a few.

Table 4. 1 Interviewee Details. Source: author

Interviewee number

Interviewee title Employed at Selection criteria (Role/Company type)

1 Portfolio manager An asset management

company

A coaching company Diversity consultant / LTD

4 Consultant pathologist A healthcare system Manager / Public sector

5 Leadership and

neurodiversity consultant

A diversity consultancy Diversity consultant / LTD

6 Partner An auditing and

consulting company

Manager / LTD

7 Finance director An investment

management company

A diversity consultancy Diversity consultant / CIC (akin to

charitable sector)

9 CEO/diversity and

inclusion consultant

A diversity consultancy Diversity consultant / LTD

4.7.4 Data analysis

Once secondary data was gathered, it was discussed and more closely analysed. The literature review included main points from the articles covered and visualisations of processes found to be most relevant or providing the best base for recommendations. After primary data was collected through interviews, the audio/video recordings were transcribed; and the transcripts used for analysis. Thematic transcript coding was done on the computer program Atlas.ti and was appropriate since it allowed for important concepts within the data to emerge (Ayres, 2008b). The initial list of themes or codes was based on literature findings and evolved as further themes were identified in the data. I started with nine themes, including: glass ceiling definition, change agent definition, British business, glass ceiling experience, eliminating the glass ceiling, managers, change management, best practices and timeline. A tenth theme emerged: worst practices. Key points within the themes were determined, commonalities found and then coded. After coding the transcripts, a code list emerged and was cleaned. A code network, found in Appendix B, was created, to visualise connections between the concepts, which then lead to the conclusions that were drawn.

4.7.5 Reliability and validity

While this methodology follows Saunders et al.’s (2007) format, I did not find it appropriate to consider reliability since it concerns whether data collection techniques will produce consistent findings (Saunders et al., 2007). Due to the ethnographic nature of this research and the topic, consistent findings would be impossible to produce. After initially trying to ensure reliability, it became apparent it was unrealistic and I did not see sense in attempting the impossible.

Validity, however, was relevant and could be divided into internal and external. Threats to internal validity begins with history and maturation (LeCompte & Goetz, 1982). This was not a big threat, being cross sectional with a two-month primary data collection period and a phenomenon that changes extremely slowly according to literature. Secondly, I did not see observer effects as too much of an issue. The threat of selection and regression was mitigated through the subject sample. LeCompte and Goetz (1982) also mention mortality, referring to the changes to groups over time, which again was not an issue due to the timing.

Spurious conclusions were also considered; keeping the research logical and not jumping to conclusions (Saunders et al., 2007). This could be achieved through four steps: identify the research population, collect the data, interpret the data and then develop conclusions. Then, external validity, which is concerned with generalisability, (Saunders et al., 2007) has four factors that can impact group comparability; the main basis for generalising in ethnography (LeCompt & Goetz, 1982). First, selection refers to the subjects involved and can be compensated for by acknowledging the specifics of the research, with recommendations not necessarily being applicable to the entire population. On setting, I recognised that the contexts in which data was collected could impact it; here the limitations were recognised, considered in the introduction and expanded on in the conclusion. Historical effects did not have much significance, much like the cohort effect. Finally, construct validity was partially accounted for through establishment of definitions of the concepts.