In manufacturing, powder mixing processes are vital for ensuring product quality. The mixing progress and efficiency are determined based on the fundamental convection and diffusion mechanisms. While mixers are believed to have a unique primary mixing mechanism, recent findings from our group have verified that the main mechanism can change as the mixing progresses. The transitions were successfully captured using a new method incorporating proper orthogonal decomposition (POD) into the discrete element method simulation, proving POD as a valuable tool for mechanism identification. Nevertheless, the existing POD method cannot quantitatively evaluate these mechanisms, hindering a comprehensive analysis of their magnitudes and transitions. This study combines analysis of variance (ANOVA) with POD to solve the problem, establishing a POD-ANOVA framework to quantify the degree of contribution of the mechanisms. The capability of POD-ANOVA is assessed in the transverse mixing of a rolling drum. For a quantitative evaluation of the mechanism magnitudes, POD-ANOVA is performed over the entire mixing process (denoted as Standard POD-ANOVA). The convection and diffusion rates are then derived from the overall mixing rate. Validations show that the two rates corroborate well with common indicators of mechanism intensities. Furthermore, Standard POD-ANOVA is applied over sequential time domains to track mechanism transitions; however, it is found to be insufficiently precise. Thus, a new time-windowing POD is implemented, leading to Windowed POD-ANOVA. Over short time windows, the improved method can effectively quantify the transitions. Consequently, the proposed methods enable a quantitative evaluation of powder mixing mechanisms scientifically for the first time.