Emerging crowd-based phenomena epitomize new modes of coordination made possible by digital technologies. Extant literature provides rich insights into the ways digital technologies may create and transform the conditions underlying the coordination of complex work. Yet, crowd-based interactions do not rely on inter-personal familiarity, which previous studies highlight as crucial to achieve coordination, and are episodic, interaction-based, rather than focused on managing complex task interdependencies over time. Important questions are thus raised about how crowd-based coordination is accomplished and how digital technologies are implicated in the creation of conditions for coordination. In this paper, we argue that conventions are the main mechanisms of crowd-based coordination. In particular, conventions help frame interactional interdependencies, create a stable structure for expectations to converge, and instill a shared obligation to read, organize and respond in certain ways in particular episodes of interactions. Moreover, the enactment of conventions is dynamically shaped by crowd generativity and deliberately designed algorithmic processes. By developing a theoretical account of conventions in the digital age, our study contributes to the literature on technology-enabled coordination and advances understanding of the workings of digital platforms.