Deriving a Shortlist of Brand Names: A Comparison of Measures Based on Censored Preference Data

Abstract

The task of deriving a shortlist from a large set of potential brand names for a new product has been rarely discussed in the literature. We reject popular ratings-based methods as unreliable and ask respondents to give us their "top five". Three alternative methods of analysing these data are compared. Each involves different levels of complexity of computation and each makes different assumptions about the scale properties of the data. We learn that the simplest method works as well as the more sophisticated method, even though it makes assumptions that severely violate the data properties.