Response surface methodology (RSM) has been widely used in practice, which can optimize single response versus several factors. Naturally people are not only interested in single response optimization, but also multiple responses optimization. In this paper we propose a general framework for multiple responses optimization using Bayesian posterior predictive method. This method can account for the effects of variances, the correlation among the responses, and the model parameter uncertainty. We develop our approach as a guideline for the practitioners, and give an example to illustrate it.