The relation between bibliometrics and science policy remains underdeveloped. Relevance of new methods to produce indicators is easily claimed, but often without real insight in the policy processes. Drawing on experiences with the use of S&T indicators in science policy in the Netherlands and on principal-agent theory, I develop an analytical perspective which enbles to assess the role of S&T indicators in science policy. It is argue that the use of S&T indicators can only be understood well if one takes the socio-political context with its specific dynamics and rationalities into account.