Mold qualification of new plastics parts is commonly done by trial-and-error. Although a suitable working point may be found, resulting in quality features within their tolerance range, e.g. part dimensions, warpage, and weight, no one can prove if this is the best centered and robust process setting. Thus Design of Experiments (DoE) methods are used both, to determine process influences (input factors) on the part quality (output factors) systematically and to calculate mathematical polynomial process functions, which predict the part quality within the examined test window continuously. The process optimum, where every tested part quality parameter is as close to its nominal value as possible, is defined by using multivariate regression and adequate optimization algorithms. To reduce test time and costs a two-level D-optimal screening DoE combined with replicated center points was applied to optimize an injection molded mirror drive component made of PET. The experiments were done on an ENGEL injection molding machine, systematically varying seven machine settings. Two diameters and part weight with known nominal (target) values and tolerances were chosen as quality parameters. These were measured by GAGE-proved gauges. Statistical analysis, model building and optimization were done in VARIMOS® and CQC® software. Most important on the both diameters was the holding pressure (~42%), followed by cooling temperature, hot runner temperature, injection rate and cooling time (16 to 10%). Part weight was affected by holding pressure (73 %), holding pressure time (19%), and injection rate (8%). The DoE optimization suggests major changes of the existing trial-and-error optimized machine settings to minimize the target value deviation. Moreover, cycle time and part weight will be reduced by 2.6 s (-14%) and 0.22 g. The validated results will be presented.