The performance characteristics of a mixture composed of multiple components are largely dependent on the mixing ratio. For example, the electrolyte composition ratio should be carefully selected to improve the cycle life of a secondary battery. The cleansing power of natural oil are affected by the mixing ratio of vegetable oil. However, a problem arises when multiple performance characteristics must be considered simultaneously. The optimal mixing ratio may vary depending on the performance characteristics being considered, but from the perspective of the product development, a single mixing ratio is finally required. This paper deals with a fuzzy logic approach for determining the optimal mixing ratio of a multi-characteristic mixture. Since the mixing ratio has a constraint that its sum is 1, the experimental space is characterized by being composed of a simplex rather than a hypercube. This paper uses Extreme Vertex Design to reflect such ratio constraint in the simplex. Moreover, in this paper, a fuzzy logic is adopted into the process of optimizing the mixture characteristics. The respective responses are fuzzified by the fuzzy membership function and then defuzzified into a single output response through the fuzzy logic base. The proposed approach is illustrated using a cleansing oil experiment dataset. Main characteristics of cleansing oil are emulsifying power, cleansing power, and finish feeling. The 3 components that make up cleansing oil are Natural Oil, Olive Liquid, and Essence Oil. The optimal mixing ratios for each characteristic are obtained as 0.75:0.04:0.21, 0.77:0.20:0.03, and 0.80:0.09:0.11, which do not match each other. Meanwhile, when the proposed fuzzy logic method is applied, the simultaneous optimal mixing ratio is found as 0.84:0.04:0.12. It is noteworthy that this new ratio cannot be found by any interpolation of the three mixing ratios. In this way, the present method can be a useful choice for finding the optimal mixing ratio of mixture products. However, the numerical experiments suggest that the final result could vary depending on the fuzzy logic base and the membership function parameter.