Evidence‐based policy making has become popular in political circles, as it promises an indisputable, ‘true’ knowledge base for policies. The call for more ‘evidence’ is especially loud in environmental and sustainability politics. This is in sharp contrast with the much lower level of certainty that social sciences consider realistic with regards to the politics of complex, disputed, so‐called ‘wicked’ problems of the sustainability agenda. At the same time, sustainability scientists have understood the reality of political marketing and strive/pretend to produce more certainty than is scientifically sound, in order to get policy makers to act. The question is why many governments seem to favour ‘evidence‐based’ policymaking for sustainability challenges, whereas they know that this can lead to taking decisions which do not take into account uncertainty and unpredictability. In the same way predictions of economists are used as evidence although they often turn out to be wrong. Both can lead to large societal costs. What are the reasons for such collective cognitive dissonance? Is it because the costs of unwise decisions will be often later and elsewhere? An answer can be found in those strands of governance theory that acknowledge the normative dimension of governance practices and propose mechanisms for dealing with normative and cognitive tensions, such as metagovernance and transgovernance theory. I argue that the popularity of using the metaphor ‘evidence’ for knowledge is a function of the culture and traditions of administrative organisations and their political leaders. The culture and traditions are expressed in the predominant application of a governance approach with a specific appreciation of what usable knowledge or ‘evidence’ is. The paper concludes with first recommendations for improving the evidence base for political decision‐making with regard to sustainable development, and questions for further research.