Resin colorant, which is a resin compounding product, is evaluated by whether customers can use it as raw material. Therefore, the resin colorant manufacturer provides a sample for the customer to perform a quality evaluation. This quality evaluation depends on the characteristics and the conditions of manufacture of the resin which the customer uses. Because this information is not disclosed to the colorant manufacturer, who sees only the evaluation results, the manufacturer has to design a recipe for resin colorant ( blending ratio of raw materials) in a very uncertain situation. In this situation, the design work depends on the intuition and experience of skilled technical staff. In this study, we analyzed customers' evaluation results of past experimental recipes and built a model to predict the incidence of defective samples from the mixing ratio, using impact analysis based on independent component analysis. We also developed a design method by which to improve recipe formulation ingredients efficiently in order to reduce the incidence of defective samples at the design stage. In addition, we verified the usefulness of this recipe design method using information on resin colorant recipes produced by Tokyo Ink Corporation. The method was found to agree with the knowledge of skilled personnel and thus to have the potential to contribute significantly to the efficiency of recipe design work.