Surface deflections have a great effect on the external appearance of automobiles. Usually, they are occurred on large flat panels containing sudden shape changes and of very small size about $pm$30$sim$300$mu$m. Since the current numerical method is not sufficient for predicting these defects, the correction of these defects still depends on trial and error, which requires a great deal of time and expense. Consequently, developing the numerical method to predict and prevent these defects is very important far improving cosmetic surface qualities. In this study, an evaluation system that can analyze surface deflections using numerical simulation and a visualization system are reported. To calculate the surface deflections numerically, robust algorithms and simulation methodologies are suggested and to visualize them quantitatively, the curvature variation algorithm is proposed. To verify the developed systems, the experimental die of the handle portion of exterior door is analyzed. The results showed that the experimental and simulational visualization are in good agreement. Compensation methods to correct the surface deflections are also tested. The evaluation system proposed in this paper could be used to predict and minimize the occurrence of surface deflections in die manufacturing.