Abstract: Focusing on an integrated production-distribution system with multi-products, multi-manufacturers and multi-retailers, such problems as the production cycle of products, transportation frequencies in a cycle and the economic allocation among the manufacturers were investigated. Then, a decision model was developed, where the productive capacity of every manufacturer and that per unit products are taken as constraints with the minimum average total cost of a logistic network within a unit time as the objective to be optimized. In view of the nonlinear programming characteristics of the decision model, i. e., neither the convex nor the concave, a heuristic algorithm was proposed to approximately solve the model. Five instances at different sizes were calculated by the algorithm and the results were compared with the results by the quasi-Newton method. The comparison verified the correctness of the decision model and validity of the algorithm proposed.