While attempts are being made to improve health promotion by following a linear Evidence-Based (EB) approach, the actors involved are aware that the quality of health promotion is not just a matter of supplying 'evidence-based' interventions to local practitioners, but the result of a situated coproduction process that depends on many factors. This paper explores what constitutes an intervention that works from the perspective of health promotion professionals (HPP), and how, according to them, the development and implementation of interventions should be improved. We interviewed 81 HPPs about the use of 10 health promotion interventions at 30 Municipality Health Services in The Netherlands. The HPPs described an intervention that works as something that produces its intended effects after being realized in a local situation. Interventions are realized by combining elements of a supplied intervention (e.g. a theory, artefacts) with elements that are situated in the local context (e.g. funding, local network). Interventions that are transferred contain implicit assumptions about local contexts, but it is often unclear what precisely constitutes an intervention and what is assumed of local contexts. An intervention that works is a situated configuration of aligned elements. A linear EB approach depends on the realization of the local circumstances in which 'evidence based' interventions can work. Various strategies are possible for approximating such circumstances, but the core assumption that the configuration that is realized in practice is similar to the 'evidence based' intervention seems unrealistic for most health promotion in the Netherlands. Under such circumstances, attention should shift from central quality assurance to the system of actors and the distributed actions and heterogeneous learning processes that together add up to interventions that work. © 2012 Elsevier Ltd.