Advances in computer-based simulation technologies have made the process of digital twining (creating a digital copy of an analogue system) more economical and more robust. Digital twins are used to make cheap and quick predictions about the behavior of the analogue system under varying permutations. Yet there is substantial variation in how digital twins are built and the effects they have on planning processes. In this talk, I focus on the way that the organization of work shapes how simulation models are built and I demonstrate – with data collected through a year-long ethnography of urban planning in two major metropolitan areas – how differences in organizing practices produce digital twins with different degrees of granularity and immersiveness. I then show how different combinations of these two qualities can lead to dramatic differences in the way predictions are made and how the consequences of those predictions are enacted.