Implant migration can be accurately quantified by model-based Roentgen stereophotogrammetric analysis (RSA), using an implant surface model to locate the implant relative to the bone. In a clinical situation, a single reverse engineering (RE) model for each implant type and size is used. It is unclear to what extent the accuracy and precision of migration measurement is affected by implant manufacturing variability unaccounted for by a single representative model. Individual RE models were generated for five short-stem hip implants of the same type and size. Two phantom analyses and one clinical analysis were performed: “Accuracy-matched models”: one stem was assessed, and the results from the original RE model were compared with randomly selected models. “Accuracy-random model”: each of the five stems was assessed and analyzed using one randomly selected RE model. “Precision-clinical setting”: implant migration was calculated for eight patients, and all five available RE models were applied to each case. For the two phantom experiments, the 95%CI of the bias ranged from −0.28 mm to 0.30 mm for translation and −2.3° to 2.5° for rotation. In the clinical setting, precision is less than 0.5 mm and 1.2° for translation and rotation, respectively, except for rotations about the proximodistal axis (<4.1°). High accuracy and precision of model-based RSA can be achieved and are not biased by using a single representative RE model. At least for implants similar in shape to the investigated short-stem, individual models are not necessary. © 2015 Orthopaedic Research Society. Published by Wiley Periodicals, Inc. J Orthop Res 34:903–910, 2016.