When reconstructing bone geometry to calculate joint kinematics, shape-model scaling can be more accurate and repeatable than linear scaling given the same anatomical landmarks. This study perturbed anatomical landmarks from optical motion capture and determined the robustness of shape-model scaling to misplaced markers compared to a traditional approach of linear scaling. We hypothesised that shape-model scaling would be less susceptible to variance in marker positions compared to linear scaling. The positions of hip joint centres and femoral/tibial segment lengths across perturbations were compared to determine each scaling method’s range of geometric variation. The standard deviation (SD) of the hip joint centre location from the shape model had a maximum of 1.4 mm, compared to 4.2 mm for linear scaling. Femoral and tibial segments displayed SD’s of 5.4 mm and 5.2 mm when shape-model scaled, compared to 9.2 mm and 9.5 mm with linear scaling, respectively, thus supporting our hypothesis. Geometric constraints within a shape model provide robustness to marker misplacement providing potential improvements in repeatability and data exchange.