The desirability function approach to multiple response optimization is a useful technique for the analysis of experiments in which several responses are optimized simultaneously. But the existing methods have some defects, and have to be modified to some extent. This paper proposes a new method to combine the individual desirabilities.