Reference models are a reification of a very attractive vision: They promise higher quality of information systems at less cost. This vision goes along with two pivotal claims. On the one hand, reference models are intended to provide appropriate descriptions of an application domain. On the other hand, reference models are aimed at delivering blueprints for a distinctively good design of information systems and related organizational settings. Thus, they are descriptive and prescriptive at the same time. While many conceptual models include descriptive and prescriptive elements, reference models should fit the specific needs of a whole range of organizations. Since the idea of reference modelling is emphasizing the improvement of quality, evaluating them is a core issue: From the perspective of prospective users it can hardly be taken for granted that a particular reference model is of superior quality. However, evaluating reference models is a major challenge. This is for various reasons. Not only that reference models inherit the problems well known from the evaluation of conceptual models in general. Furthermore, their claim for general (re-) usability implies to take into account the possible variety of requirements and specific constraints within the set of potential applications. Another source of complexity is the variety of objectives related to the use of reference models. They include economic goals, such as increase of productivity, or goals related to specific analysis, design or implementation tasks. In addition to that, testing the claim for excellence faces deep and subtle epistemological problems. Against this background, the paper proposes a method for evaluating reference models. It consists of a conceptual framework that serves to structure the overall evaluation problem, which is supplemented by a prototypical process model that demonstrates how to organize a specific evaluation project.