Impact assessment of R&D is currently facing important challenges. First, the nature of innovation as well as the scope of innovation policy are transforming in a way that has not been taken into account in evaluation practices. While linear model of innovation has been replaced by a systemic view of the innovation process, current evaluation methods are still largely based on the linear model. At the same time, innovation policy is transforming into an expanding, horizontal and network-based policy field which calls for broader approaches in evaluation. Second, impact assessment has traditionally been largely based on the idea of legitimisation of innovation policy programmes. As a consequence, evaluation has not been informed by theories of learning and strategic management nor aimed at producing knowledge to support the continuous improvement of the activity or organisation in question. Due to emphasis on legitimation, information based on quantitative indicators has been in the forefront instead of more diversified approaches to data collection. Third, innovation policy evaluation has been backward-looking and non-continuous in nature and there has been a lack of longer-term, foresight-oriented element to support strategic thinking. This article reviews the current problems of innovation policy evaluation and introduces a new, systemic and future-oriented evaluation model. The novelty of the model is that it responds to the challenges that innovation policy evaluation is facing by combining foresight methods to the traditional impact assessment methodologies under single evaluation framework. On this basis the article considers what kinds of new competences the future-oriented impact assessment model requires from the evaluators, objects of evaluation as well as strategic decision-makers.