Visual research methods are becoming increasingly important for qualitative studies. Within this dynamically expanding field, methods for analysing ‘natural’ video recordings have developed considerably over the past decades. In this article we discuss methodological aspects of general importance for any analysis of this type of video data. Being a fundamentally interpretive method, our first argument is that sequential analysis is always a hermeneutic endeavour, which requires methodical understanding. The second refers to data collection. We stress that, in addition to sequential analysis, the ethnographic dimension of video analysis should be taken into account methodologically. Video analysis requires, thirdly, a systematic account of the subjectivity, both of the actors analysed as well as of the analysts. Our arguments are grounded in extensive data from several studies, including the communicative genre of powerpoint presentations, commemoration rituals and public events. Selected data fragments are presented here to support our claims. Building upon this expertise, we propose further improvement of video analysis methodology by reflecting on our own practice of analysing video in data sessions (i.e. the ‘video analysis of video analysis’).