Similar policies, if applied to different contexts, can have many different effects and a policy that has proven successful in one context does not automatically have to be as successful in another context. This displays the problem of the limited external validity of impact evaluations. How can policy makers then know in which context a formerly successful policy could work again? Martin J. Williams argues that the transferability of a policy depends on the interactions of the policy’s mechanisms with features of the new context. To analyse whether a formerly successful policy will work out in another context, Williams proposes a new approach called mechanism mapping which aims to help policymakers identify potential mechanism context interactions, and come up with suitable adaptations.