In urban evacuations, especially in response to an expected disaster, capacity reversibility (also known as contraflow) has been considered a workable strategy to reduce traffic congestion and to meet evacuation time deadlines. Currently, contraflow strategies are mostly planned by relying on engineering judgment because of the lack of appropriate large-scale decision support tools. A tabu search-based heuristic approach is introduced here that can be applied on realistic-size networks. The approach relies on insights from an analytical formulation of optimal reversibility design that reduces total system travel time. Computational results on a hypothetical and an urban network example network are presented and discussed.