Generation of geometric data sets, e.g., reconstructions of process monitoring data, X-ray computed tomography (XCT) volumes, and dimensional inspection point clouds, underpin the digital twin concept which has found purchase in additive manufacturing (AM) applications. Predicting component performance and tracking processing pedigree depends on ‘registration’ of these data sets, i.e., geometric alignment. This challenge is non-trivial due to the complex, multi-modal, and difficult to represent data structures encountered in building digital twins. This work proposes a generalized datum-based approach to the task of geometric registration. Effects such as thresholding, scaling, manufacturing errors are considered. Case studies are presented which show the advantages of the approach and identify best practices in geometric registration for digital twin construction. Results show that the datum-based approach shows potential for intuitive, functional, and automated registration routines.