This article presents a data warehouse used for risk assessment in food products. The experimental data stored in this warehouse are heterogeneous, they may be imprecise; the data warehouse itself is incomplete by nature. The MIEL++ system – which is partially commercialized – is composed of three databases which are queried simultaneously, and which are expressed in three different data models: the relational model, the Conceptual Graph model and XML. Those models have been extended in order to allow the representation of fuzzy values. In the MIEL++ language, used to query the data warehouse, the end-users can express preferences in their queries by means of fuzzy sets. Fuzzy pattern matching techniques are used in order to compare preferences and imprecise values.