This paper proposes a principled approach to multicriteria decision making where the worth of decisions along attributes is not supposed to be quantified, as in multiattribute utility theory. In the latter, alternatives are ranked after merging utility functions defined along each attribute. In constrast, our approach presupposes an aggregation of local preference relations, which includes information about the relative importance of coalitions of criteria. It is shown that the requirement for ordinal representation and fusion of preference leads to a very specific family of decision rules, called concordance rules that generalize well-known voting methods.