Statistical literature is being more and more concerned with debates about hypothesis testing and p-values supporting the significance of a given variable specification. Specifically, if on one hand statistical foundations about significance are not arguable, scholars should be able to distinguish between significance and variable importance. This is a matter of serious concern in questionnaire analysis to derive respondents’ profiles and develop targeted marketing strategies, for instance. To this aim, this contribution proposes a hypothesis system that considers the normalized dissimilarity measure to assess the importance of explanatory variables in the setting of mixture models for ordinal data to account for uncertainty of choice.