We propose a new method for modelling users' preferences on attributes that contain more than one trait. Starting with a data set the users have to enter a sort of order on the attributes in form of formulas corresponding to their preferences. Based on this order they only receive the relevant formal concepts, i.e.; "object-attribute clusters", where relevant corresponds to the users' point of view. The preference modelling is done within the framework of Formal Fuzzy Concept Analysis. This has numerous advantages. First, the relevant information is contained in a complete lattice, the concept lattice, that allows the users to browse among their preferences. This lattice may be used for further data analysis by applying different methods from Formal Concept Analysis. Second, we can investigate the computation of non-redundant bases for the entered formulas. Since the users are allowed to enter the formulas, these may be redundant. The base offers a better overview of the preferences and thus the formulas can be altered more easily.