Benchmarking has a long history in the management world. It first emerged in the private sector as an engineering tool and having passed through different stages of development, is now also a policy-making tool in the public sector. The paper argues that although typologies of benchmarks can be developed and a generic methodological approach can be formulated, these cannot be unconditionally used in the sector of public policy. A number of problems can be identified: the lack of agreement on what public policy is, the contradiction between learning and copying in public sector organizations, the dualism between top-down and bottom-up approaches, and the conflict between accountability and public trust. These problems are also reflected in the specific case of innovation policy.