While it is generally assumed that the aim of teacher evaluation is to "formatively" support teachers' professional development, research finds that teacher evaluation practices are predominantly "summative". This paper describes a Norwegian governmental policy experiment aiming to overcome this fallacy through a bargaining process, where experience-based knowledge was combined with research evidence. When preparing to introduce teacher evaluation, the Ministry of Education and Research commissioned a group of researchers and a group representing practitioners to identify teacher evaluation practices that are conducive for educational quality. Drawing on experiences from the policy experiment, the article discusses three approaches to teacher evaluation: the political, the administrative and the professional. The analysis indicates that successful implementation of interventions needs a new educational infrastructure and professional school leadership. One conclusion is that teacher evaluation cannot be successfully implemented through traditional linear approaches. A more productive approach is to treat it as a wicked problem.