This thesis examines how computer modelling matters for policy-making by looking at two case studies of European fisheries management. Based on documentary analysis and ethnographic interviews and observations, the main case is located within the European Union (EU) and centred around the flatfish fishery in the North Sea with a supplementary one from outside the EU and focused on the North East Arctic cod fishery in the Barents Sea. As in other much-contested areas of public policy, fisheries officials in the EU and neighbouring countries seek to develop a universalistic and objective ground by which to depoliticise management decisions. In this sense, modelling has long become their preferred approach to produce policy relevant representations of the otherwise hidden dynamics of a fishery. Social constructivists in the field of Science and Technology Studies (STS) studying the modelling used in areas of policy-making such as, for instance, climate change have questioned whether models are the right tools for this job given that the modelling may conceal large uncertainties about their accuracy and relevance to policy-making. Some of these scholars argue for producing ‘good’ models for policy-making, and thus more robust policies, by constructively engaging the non-modellers or non-specialists in the quality assurance of the modelling. ‘Fisheries Studies’ literature suggests, however, that modelling can contribute to policy resilience despite its well-known limitations to produce accurate fish counting. It follows that models are doing something else than providing policy-salient real-looking representations. How may modelling count differently for policy-making in fisheries and beyond? Drawing on the ‘co-production’ of science and social order framework from STS, the thesis puts forward three related arguments. First, that the technologies designed to depoliticise decision-making, including modelling, become spaces for political work by policy-makers, stakeholders and scientists. Second, that the role of computer modelling for policy stems from how representational validity and political usefulness are produced together. Third, that the role of modelling for policy is mediated by virtue of being assessed together with other technologies for depoliticising as part of a whole sociotechnical infrastructure to allow evidence-based decisions. As a distinctive contribution, this thesis thus questions the presumption in many social constructivist accounts that modelling alone becomes central to the policy process and its outcomes. The significance of modelling for policy-making should be understood in terms of its contribution to processes of sociotechnical framing. Narratives that foreground the former and background the latter show an analytical bias that needs turning around.