A three-dimensional single container loading problem (3D-CLP) is aimed to load kinds of rectangular boxes into a single container with maximal volume utilization. This paper presents a multi-stage search based simulated annealing algorithm (MSSA) for single container loading problem (3D-CLP). An approach for representation of feasible solution is presented. An empty maximal-space list is used to manage the free spaces and difference process is applied in space decomposition. For simulated annealing algorithm is an individual improve approach which rely on a large number of trials, a multi-stage search process is applied to improve accuracy. This approach is tested on the set of test cases proposed by Bischoff and Ratcliff , and a comparative result is presented.