The concept of fuzzy signatures might be useful when modeling complex, well structured problems, where one or several components of the structure are determined at a higher level by a sub-tree of other components. The data set belonging to the problem has an arbitrary structure, from which the structure of the data may slightly differ. An aggregation operator is given for each node, for the purpose of modifying the structure, so that data with missing components can be evaluated. Deducing a conclusion from an observation having such a structure is a key issue. In this paper fuzzy signature based rule bases will be introduced, then the generalisation of the well known Mamdani method for signature based rules will be shown step-by-step. Finally, an example of inference on fuzzy signatures will be discussed.