Classification relies on the rules expressed by domain experts, or on the labeled attribute explaining the output classes. However, such information is not always available. In this work, we explore classification according to aggregation functions of mixed behaviour by the variability in ordinal sums of conjunctive and disjunctive functions. By this approach, entities are classified into three classes: yes, no and maybe including inclination to the classes yes and no. By the proposed approach, domain experts (or ordinal users) explain classification linguistically, without stating IF-THEN rules and labeled output. The applicability is illustrated by two examples. The discussion of the results and further research activities conclude the paper.