A question often asked of video-based research concerns how representative it is, and what unacknowledged constraints operate to limit the validity of video analyses. Questions can be asked in relation to selectivity including where the camera is pointing, the episodes analyzed, and theoretical stance. In this paper we describe research utilizing multiple tilt and zoom cameras and radio microphones in a specially set up classroom with sophisticated control apparatus. Six classes of students took single inquiry lessons with their teachers, on separate topics, with tasks challenging them to reason through object exploration, modeling and representing through drawings and iPad productions. The research team was faced with the challenge of making sense of the enormous data set of group-teacher-object interactions, and the methodological problem of making justifiable analytical decisions. In this paper we will articulate how we progressively reduced the data set through collaborative analysis and discussion, resulting in the generation of new research questions. We will describe the process by which the role of the teacher circulating around groups, intervening to clarify and challenge, was identified as a focus of interest, and how we identified and tracked non-trivial interventions of a single teacher, analysing video sequences for each of the groups to include one minute before, through, and after the intervention. This process enabled us to identify the varied nature and purposes of the interventions and their effects on students. Productive teacher interventions included: • Providing suggestions or offering pathways out of confusion• Pursuing multiple agendas including the nature of scientific processes• Focusing on conceptual clarification, task clarification and timely completion• Selecting interventions appropriate to the resources of the students.This resulted in descriptions of the complex, responsive and highly contextual nature of expert teacher practice, in contrast to contemporary trends which seek to standardize and codify ‘best practice’ in teaching. We argue that while the context and process of selection of the teacher and of the interventions renders this analysis inevitably ‘unrepresentative’ in the traditional sense, the findings yield legitimate (and useful) insights into teacher-group intervention processes. We make the case that idiosyncratic examples overcome the banal and provide opportunities for building new theoretical and conceptual insights. This raises some important points of discussion about ‘case study’ as a method and how we might claim the legitimacy of research decisions such as what constitutes ‘data’ and how we decide on the appropriate unit of analysis.