Nascent research on data work has recently criticized prevailing data science narratives that emphasize front-stage data analytics and assume data to be objective and readily-available resources waiting to be fed to data models. Data work studies have shed light on what happens at the back rooms of data science, uncovering practices of finding, producing, and curating digital representations of physical phenomena. Yet, existing theorizing efforts remain silent regarding the importance of embodiment in data work. Drawing on insights from an ethnographic study on dairy farms, we shed light on an overlooked, yet instrumental, aspect of data production: the bodies of practitioners and their embodied interactions not only with the physical structure producing digital representations, but also with the physical referents of those representations. Our preliminary analysis points to three distinct yet interrelated data work practices—automated, procedural, and indwelling—which have important implications for the representational fidelity of digital representations.