In the presence of transmission outages, uncertain demand and variable renewable supply, network operators keep a reliability margin to avoid interruptions and black-outs. The reliability margin is presently determined by the N-1 reliability criterion. Our analytical model defines the optimal reliability margin by balancing congestion costs and interruption costs. This leads to more efficient use of transmission capacity and to smaller investment needs than with the N-1 criterion. A numerical illustration shows the net benefits of the new reliability criterion.