Innovation policy is in need of a rationale which allows for the design and evaluation of policy instruments. In economic policy, focus has traditionally been placed on market failures, and efficiency measures have been used to decide whether policy should intervene and which instrument should be applied. In innovation policy, this rationale cannot be meaningfully applied because of the uncertain and open character of innovation processes. Uncertainty is not a market failure and cannot be repaired. Inevitably, policy makers are subject to failure and their goals cannot pragmatically be represented by a social optimum. In eschewing the concept of ‘optimal innovation’, avoiding evolutionary inefficiencies becomes central to analysis and to innovation policy making. Superimposed on the several sources of evolutionary inefficiencies are so-called ‘network inefficiencies’. Because of the widespread organization of innovation into innovation networks, network structures and dynamics give useful hints for where and when innovation policy should intervene.