Social impact of research is difficult to measure. Attribution problems arise because of the often long time-lag between research and a particular impact, and because impacts are the consequences of multiple causes. Furthermore, there is a lack of robust measuring instruments. We aim to overcome these problems through a different approach to evaluation where learning is the prime concern instead of judging. We focus on what goes on between researchers and other actors, and so narrow the gap between research and impact, or at least make it transparent. And by making the process visible, we are able to suggest indicator categories that arguably lead to more robust measuring instruments. We propose three categories of what we refer to as ‘productive interactions’: direct or personal interactions; indirect interactions through texts or artefacts; and financial interactions through money or ‘in kind’ contributions.