This paper studies the design of unemployment benefits within a Crawford–Knoer framework. First, we provide a new characterization of the worker-optimal stable allocation and show how an increase in the benefits of an unemployed worker propagates within the market. To capture this propagation, we construct a directed acyclic graph that identifies workers whose targeting maximizes aggregate welfare. Building on this network structure, we design the optimal policy that maximizes the gains in aggregate welfare using a modified version of the eating algorithm of Bogomolnaia and Moulin (2001). Finally, we analyze how the introduction of unemployment benefits affects the set of stable allocations and derive design recommendations.