Comparing apples to oranges

Abstract

MaxDiff scaling and discrete-choice conjoint methods measure subjects' preference structures relevant to a common referent, thus removing a common scale origin for between-subject comparisons. Subject-level estimates of the partworths from discrete-choice and MaxDiff are not on a common scale across subjects; they do not have the same scale "zero point." This article describes augmenting choice-based tasks with ratings to recover the lost origin. The benefit is to extend the utility of discrete-choice methods to domains such as segmentation and targeting.