Abstract:

This paper outlines the conventional techniques of testing preferences; it suggests that, quite apart from technical discussions about the advantages of one or other preference testing method, when more than one preference dimension is brought into regard, the results of well-proved methods take on a radically different aspect. The problem dealt with is how the preferences for different informants, who may be different types of informant, for different products or services, under possibly different conditions or assumptions on the part of informants, can be represented. At first a conventional representation of a possible preference test is shown, and then a principal component analysis of the same data is mapped in an original way that is described in detail. The results are dramatically different from the more conventional representations of preference test findings the product with the most consistent and relatively high score is when one takes all the preferences for all the products into account simultaneously - the worst possible bet for the market.