This paper presents the defect reduction in a surface mounting process using a combination of statistical process control and design of experiments techniques. Such process is used to produce the flexible printed circuit. It consists of three sub-processes; solder printing, mounting and reflowing process. The potential defects, including component missed position and non-component, were selected using the pareto diagram. Then, the main effects to these defects were introduced. For quickly response to these defects, the potential effects and their possible levels were firstly determined from the production data in the past. It was found that there were four potential factors which were the vision parameters, the number of fiducial marks, the reflowing profile and the number of push up pins. The improvement was performed using the main effect determination and found that the defect rate was reduced from 17,234 PPM to 8,141 PPM. After that the influential factors affecting to these defects were analyzed by applying 2k factorial experiments and the conventional statistical analysis. It was found that there were three potential factors which were the design aperture, the reflowing profile and the number of push up pins. It included an interaction effect between the number of push up pins and the number of fiducial marks. On the constrained response surface optimization a multiple linear regression model in a concept of the path of steepest descent, then, was applied to determine the suitable level of each factor. With the suitable level found, it was found that the variation of process was reduced providing the reduction of the defect rate to 6,265 PPM and seemed to better than the conventional statistical analysis.