To address the demand for the collaborative-decision optimization between linehaul and delivery transport of e-commerce platforms, this study investigates a novel variant of two-echelon vehicle routing problem by jointly considering three important practical factors: multi-commodity, unpaired supply-demand and soft time windows. This study formulates a mixed-integer programming model, and constructs a simple heuristic based on the cost-saving principle to quickly obtain an feasible initial solution, which is improved by incorporating three removal operators and one repair operator, two insertion operators and four exchange operators into an iteral local search framework. The experimental results show that: ① Our iterative local search algorithm can quickly obtain the optimal solutions for small scale problems, and has obvious advantages over CPLEX in solution quality and efficiency for medium and large scale problems; ② Compared with the manual decision of enterprises practice, our collaborative decision can save 16.65% of costs; ③ Compared with single-commodity transportation pattern, multi-commodity combination transportation can save 69.83% of costs; ④ The relaxation of hard time windows constraints can save 4.92% of costs.