The aim of this work was to perform a real and virtual optimization of the serial production of an injection molded automotive part. The part, a mirror drive component, was produced by Magna Auteca AG, Weiz, Austria. The cycle time, part weight and various defined dimensions (eight diameters, eight planarities, six average diameters) had to be optimized using DoE (design of experiments) software to find robust process parameter settings. The mold temperature, the melt temperature, the back pressure, the injection speed, the holding pressure, the holding time and the cooling time were varied on the production machine and in injection molding simulation according to a D-optimal experimental design. For planning and interpretation of the experiments the software package CQC®, part of Varimos®, was used. With CQC® it is possible to conduct statistical analysis on the results and to build mathematical models by multiple linear regression. These mathematical models relate the part qualities to the parameter settings. For the component weight a very accurate mathematical model was obtained (R²=99.40%). For measuring the dimensions a coordinate measuring machine was used. For seven diameters and three planarities mathematical models with coefficients of determination R² lower than 75% were achieved due to the high deviations in the measurements and poor measuring system capability. With these unsatisfactory models, the process optimization in real experiments was not possible. Therefore, every part, that was produced in the DoE, was remeasured four months later by a GAGE-proved gauge, which is used by Magna Auteca in the SPC. Using this data, CQC® calculated R² of 84.81% and 86.56% for two definied diameters. The maximum relative deviation between the diameter's model predictions at the serial production parameter settings and the real part measurements was 0.005%. By the optimization the cycle time could be reduced by more than 2s (13%) compared to the serial production. In the injection molding simulation software Cadmould® a detailed FE-model of the mold, including hot runner and cooling system, was created. Moreover, it was necessary to remeasure the material data of the used PET (viscosity, pvT-data, thermal conductivity and specific heat capacity) for the simulations. Regarding the two diameters, the deviations between simulation results for the center point parameter setting and real measurement results were less than 0.21%. However, the virtual optimum differed clearly from the real one. But even in the simulated optimum, the cycle time could be reduced by 1s compared to the serial production.