A thematic interview design uses a semi-structured interview protocol based around key themes and provides the researcher with more flexibility for
dealing with novelty. Instead of a formalised, pre-decided, structure, the content of semi-structured interviews is shaped by what the respondents tell you.
The researcher guides the respondent into particular areas, but what path is actually followed is usually decided by the person doing the talking.
This is a powerful research technique when not much is already known about the topic being researched, or where that topic is particularly complex
(Tolich & Davidson, 1999) . Typically, these kinds of interviews vary in length depending on both the amount of time a participant has available and their
knowledge of the research topic. However, it is anticipated that these semi-structured in-depth interviews will last approximately 60 minutes (and will be held
at a time most convenient for the participants). The approach to interviewing, coding, and analysis of qualitative transcripts will follow that outlined in Tolich
and Davidson (1999).
Balancing quality, timeliness and cost in this evaluation hinges on the careful choice of key informants, and the use of ‘theoretical saturation’ in data gathering.
In line with all small qualitative projects, it is critical that the sample be drawn by the deliberate selection of theoretically important units (here, different kinds
of informants). The researcher, working in close conjunction with the client, decides on analytical grounds what data to collect next and where to find them. Generalisations
from the research, then, are not based on random theory but on ‘typical cases’. Sample size is not determined by probability theory but by achieving saturation (that is,
when additional analysis no longer contributes anything new about an analytical category). Sample size is evaluated in the context of the study (and the theory which emerges).
In all cases, the key role of research judgement needs to be balanced with an awareness of the potential absence of consensus and interpretations of interests within an evaluation.
To ensure this is done here, the research team will draw on the model of ‘pluralistic evaluation’ (cf. Thompson, Atkin & Lunt, 1999).