When conducting a performance evaluation on five-axis machine tools, it has been impossible to overcome the restrictions of cross-matching across different models over the years; complete evaluation reports have been often limited to certain types of cutting movements, cutting angles, and processing of test pieces of a specific geometry. In order to successfully complete the cross-performance evaluation, we propose a statistical approach to solve this long-standing problem. Pyramid part machining test pieces were used in this study as the source of data analysis to obtain a quantized value for the interactions through analysis with the Taguchi method S/N ratio and by using the variables separable model. A comprehensive evaluation of the cutting performance was performed on four completely different five-axis machine tool models. We found that PY-TM had the best results, and that PY-A most needed to redefine its quality improvement program, which the mean standard deviation (σ-Sigma) of each evaluation type of five-axis machine tools also showed similarity with the above results.