‘Evaluation clauses’ are a legal basis requiring mandatory evaluation of public policies impact. Evaluation becomes a procedural institutionalization but, although they are widespread, no phrasing standards have been yet established. Without clarification, the questions raised by origins and effects of mandatory evaluation remained impossible to treat. Past studies dealing with this topic dismissed the legal feature of the evaluation process to focus only on organizations. This paper aims to apply a methodology based on the numerical taxonomy to construct a conceptual scheme for classifying evaluation clauses. A hierarchical cluster analysis was run on 322 cases with eight variables related to their normative density. From such heterogeneous legal object, two families of clauses have emerged, and three significant templates are identified. This paper opens a research track which considers evaluation clauses to investigate both the reason why a specific type of phrasing is chosen and the consequences of these institution types.