Modelling consumers' preferences of coffees: Evaluation of different methods

When developing a new product, it is useful to map our consumer preferences as a function of product descriptors. Classical external preference mapping methods perform a principal component analysis of an experts' product description, then model the consumers' preferences as a function of the two components. These two steps make the data easier to interpret, but result in some loss of information. Using an investigation into coffee, this paper compares the performances of different models, showing some of the disadvantages of this method.