This chapter deals with the methodological procedures of a CA study by tracking the development of a collection of instances of a multimodal practice and its variants. We describe the development of a study of the use of the German formats darf/kann ich…? (‘may/can I…?’; Deppermann & Gubina, 2021). Requesters use this format to ask if they may/can perform some embodied action while already starting or even fully performing it before the requestee’s confirmation. We first describe the process of sampling candidate cases to create a collection allowing us to identify a certain practice. Second, we describe how we analyzed (i) the time course of embodied action and its relationship to participants’ talk, (ii) the relationship the linguistic turn format, the sequential position and the multimodal context of the turn, and (iii) the relationship between situated action formation, linguistic design, action types, and interactional properties of a practice. Finally, we stress the importance of applying various strategies of comparative analysis and analytic induction to a larger dataset. We also discuss attending to the multimodal formation of social action on the basis of video data and multimodal transcripts is crucial for our understanding and analysis of face-to-face interaction.