Innovation is said to be dependent upon collaboration and networks. The innovation system thinking empha-sizes networks but also their supporting informal institutions, learning processes, and the relations between actors in the system. Despite the importance of networks, evaluation studies have been sparse on investigat-ing effects of these as well as on the broader criteria to evaluate the innovation networks and functioning of the system. This paper discusses what are such criteria for evaluating innovation policies that rely on enhanc-ing system connectivity and repair system failures? By way of illustration, and as a mean to be specific on these criteria, the paper discusses the possible rationale for governments to support business angel networks (BAN) and what criteria to apply when evaluating such networks. <br/>It is found that applying traditional evaluation criteria for assessing BANs may provide only a partial picture. The broader benefits of an angel network include both direct and indirect effects. In a story of the evolution of Danish Business Angel Network it is illustrated that it was closed down because of lack of understanding of the important effects that go beyond the immediate effectiveness. Other evaluation perspectives that includes also effects at a system level would probably have resulted in a different decision.