The automotive industry is a highly competitive sector. Manufacturers must effectively control highly complex production processes in order to fulfil all customer orders for customized cars on time, on budget and to the required quality. In this paper, the authors focus on improving the flow time of asynchronous automotive assembly lines by reducing the buffer time. A simulation-search heuristic procedure was developed and confirmed in a 5 workstations asynchronous assembly line installed in an automotive company. The proposed procedure identifies optimal performing buffer profiles for each storage level which guarantees lowest flow time while keeping the same throughput level. Experiments results show that our new algorithm significantly outperforms existing results, especially for large scale problems.