Aaron Ping ^*

• Research & Innovation, Nihon L'Oréal, Kawasaki, Japan

Cosmetic researchers recruit consumers to evaluate new formulas as part of the product development process. This screens out poorly performing formulas in favor of better ones for further testing. Trained experts score them on a battery of sensory attributes until a few formulas are selected for more costly, blind-use-tests (BUTs) featuring randomly recruited consumers. Once formulas pass a BUT they are ready for commercialization.Resources would be more efficiently used if BUT results were predicted from earlier rounds of testing. However, predicting the relationship between sensory testing and BUT testing is limited by the lack of data in common between the two methods. Even though hundreds of consumer responses are recorded, only their mean is merged into the set of data used for analysis. This reduces the amount of data available for decision-making and introduces the challenges associated with analyzing small samples. This paper proposes improving on this mean-based approach by adding bootstrapping when combining sensory expert responses with BUT responses. It compares the BUT predictions captured via bootstrapping versus the predictions obtained using only the means from the original data sets.