This paper presents an optimization analysis of the geometrical accuracy of metal binder jetting products. An orthogonal array L9 was designed using Taguchi method to study four process parameters (printhead speed, dark body, powder applicator speed and shell thickness) at three levels. Printed parts were measured by a coordinate measuring machine, and parallelism error was evaluated, both in planes aligned with binder injection and with powder spreading direction. Samples geometry was characterized by a squared through hole to evaluate differences in external and internal features. In the nine experiments, the parallelism varied in a range between 0.01 and 0.04 mm. No significant difference was observed in relation to plane direction and feature type, while parallelism tended to decrease with the increase in the sample’s distance from the printing platform. A grey relational analysis was employed to optimize process parameters, in order to improve the geometrical accuracy. Through the analysis of means of grey relational grade, it was found that powder applicator speed has the highest effect on response variables, while printhead speed has the lowest contribution. The optimal parameters combination was determined to effectively improve the quality of binder jetting products.