Summary: Many researchers believe that it is inappropriate to rely on routing decisions when locating depots. Such a conception has unfortunately led to a shortage of published work in this area. In this study we briefly review the recent work on location-routing heuristics and show using robustness analysis that these models are as reliable as location-first routing-second methods. We employ simulation to show that combined models consistently produce solutions of higher quality than sequential ones. Selection criteria based on simulation are developed for choosing the most appropriate locational solution. Computational results based on problem sets of 400 customers and 15 to 25 depots are reported.