Considerable analytic attention has focused on interviewees’ talk as ‘accounts’, or ‘versions’, rather than as direct reports of attitudes or perceptions. However, despite recognition of the ‘co-construction of accounts’, little analytic attention has been given to the interviewer as central in the production of the talk. With this attention, we are left with an analysis of some decontextualized-features-of-talk (or discourse/identities/narrative/repertoires/rhetoric). By contrast, this article argues that interviews are inherently social encounters, dependent on the local interactional contingencies in which the speakers draw from, and co-construct, broader social norms. Through a close reading of one published paper and extracts from an interview, I document how interviewees and interviewers work to construct themselves as certain types-of-people in relation to the topic of the interview and reflexively the interview itself. I then show how a prescription from some interview methods texts document an ideal about interview practices. The analysis of the interview extracts shows this ideal-in-practice (cf. Hester and Francis, 1994). I argue that whatever ideals are practised, no single practice will gain ‘better data’ than the other practices. The ‘data’ obtained are highly dependent on and emerge from the specific local interactional context which is produced in and through the talk (and concomitant identity work) of the interviewee and interviewer. An awareness of this local context of data production is central to analysing interview data whatever analytic stance is taken when analysing the data.