What are the essentials when using multicriterion aggregation with non-cardinal or mixed data?

Very often in multicriterion analysis the actions to be compared are evaluated according to a set of conflictual and heterogeneous criteria. Also these evaluations can contain elements difficult to quantify. In this case, the dominant information is often noncardinal or mixed. Several procedures of multicriterion aggregation (MCAP) are implemented in such a context. After recalling the main difficulties which a MCAP can encounter in such a context, from the theory of measuring viewpoint, we propose a set of requirements which should be checked by a MCAP so it can be described as acceptable for such evaluations.