The ongoing declines in Atlantic salmon populations across its range underscore the need for co-ordinated scientific-based knowledge to support management and decisions for their conservation. Current salmon management actions remain largely focused on addressing bottlenecks to production in the freshwater phase of the life-cycle, whereas the continued declines observed in the recent decades are thought to be driven primarily by constraints on the marine phase. The challenges brought by global warming and other emerging stressors require immediate actions, requiring us to re-think the methods behind stock assessment and forge stronger linkages between data, models and policies to promote more effective management actions. We outline a scientific framework that takes a wider ecosystem view, designed to evaluate holistically a suite of indicators and potential drivers of salmon mortality at key phases of the life cycle. The aims of the proposed “Likely Suspects Framework” are to enhance cross-fertilisation of ideas between assessment processes at the stock-complex scale and stock-specific focused management activities, and to develop new decision support tools to improve management efficiencies and scenario testing. Adopting such an approach provides a new way to catalyse the acquisition and deployment of both existing and new data and models that are urgently needed for assisting the conservation and future stewardship of salmon stocks on both sides of the Atlantic.