This study aims to validate model-based radiostereometric analysis (RSA) on the glenoid component of reversed total shoulder arthroplasty. We compared two different modalities of model-based RSA, elementary geometrical shapes and reversed engineering. We also explored two different ways to position the patient to obtain different projections of the implant, the hip-position (transversal) and shoulder-position (sagittal). Phantom accuracy was determined by performing nine translations (x, y, z) and five rotations (x, y, z), and expressed as the mean difference between RSA measurements and micrometer values. Precision was measured using 12 double examinations of the phantom and 19 in patients, and expressed as1.96 × standard deviations of the paired differences between double examinations. The accuracy was high for both modalities, but rotation around the symmetrical axis of the implant could not be measured using reversed engineering. Clinical precision ranged from 0.13 to 0.25 mm for translations, and 0.4° to 0.7° for rotations, using reversed engineering. For elementary geometrical shapes, the precision ranged from 0.18 to 0.34 mm for translations, and 0.8° to 1.8° for rotations. The hip-position was abandoned due to poor implant visualization. Model-based RSA on the glenoid component of reversed total shoulder arthroplasty has a high precision and accuracy, comparable to RSA results on hips and knees. Patient positioning is vital for obtaining adequate results. We found that reversed engineering was the more reliable method, and recommend reversed engineering as the method of choice for further clinical RSA investigation of the glenoid component of reversed total shoulder arthroplasty. © 2018 Orthopaedic Research Society. Published by Wiley Periodicals, Inc. J Orthop Res.