Quantitative three-dimensional analysis of orthodontic tooth movement (OTM) is possible by superposition of digital jaw models made at different times during treatment. Conventional methods rely on surface alignment at palatal soft-tissue areas, which is applicable to the maxilla only. We introduce two novel numerical methods applicable to both maxilla and mandible. The OTM from the initial phase of multi-bracket appliance treatment of ten pairs of maxillary models were evaluated and compared with four conventional methods. The median range of deviation of OTM for three users was 13-72% smaller for the novel methods than for the conventional methods, indicating greater inter-observer agreement. Total tooth translation and rotation were significantly different (ANOVA, p < 0.01) for OTM determined by use of the two numerical and four conventional methods. Directional decomposition of OTM from the novel methods showed clinically acceptable agreement with reference results except for vertical translations (deviations of medians greater than 0.6 mm). The difference in vertical translational OTM can be explained by maxillary vertical growth during the observation period, which is additionally recorded by conventional methods. The novel approaches are, thus, particularly suitable for evaluation of pure treatment effects, because growth-related changes are ignored.