Background: The canonical view of expert legitimacy in policymaking links it to objectivity and autonomy from politics. Yet, in practice such 'epistemic gains' stemming from the separation of facts and values are problematic, as expert advice inherently combines political and technical considerations. Aims and objectives: This article addresses the puzzle of double --- technocratic and political --- legitimacy of experts by proposing a framework for understanding expert legitimacy as an interplay of three analytical levels: epistemic, individual actor and institutional. The paper explores this problem in the case study of global poverty measurement as a field located at the interface of science and policy. Methods: This is a comparative case study of poverty measurement in the World Bank and UNICEF. Data was collected through semi-structured interviews with 40 experts employed by the two organisations. Findings: The analysis posits expert legitimacy as constructed via navigation between specific practices of knowledge production, such as the production of policy-relevant and methodologically robust knowledge, a strategic distance between the research and the political setting aimed at extending or shortening the distance between experts and policymakers, and institutionalised cultures of evidence of the organisations through which expert advice is given. Discussion and conclusion: The paper offers a theorisation of expert legitimacy as symbiotic negotiation between technocratic and political modes of accountability which are irrevocably linked while remaining strategically separated.