A PCI (Process Capability Index)-based desirability function approach for multi-characteristics parameter design problems is presented. Three types of PCI's, including the one newly developed for the-nominal-the-best characteristic, are employed depending on the type of performance characteristics, and a new desirability function (i.e. a logistic function) is proposed for the PCI's. The weighted geometric mean of the desirabilities for PCI's is then used as an aggregated performance measure for compromising the conflicts among the levels of a design parameter. Using PCI's and a logistic desirability function for solving a multi-characteristics parameter design problem is relatively simple and easy to implement. In addition, the proposed approach yields dispersion-sensitive results regardless of the type of performance characteristics. The developed approach is illustrated with an example from a drawing process of optical fiber, and compared with the existing desirability function approaches.