The cost saving percentage is important to enterprises that make decision to participate in an online logistics distribution alliance. However, computing the information needs to optimize 2N-1 (N is the number of enterprise) complexity NP-hard problems being similar to multi-depot vehicle routing problem, and to allows very limited computation time. In this paper, a fast method is proposed to estimate the result of collaborative vehicle routing problem. Firstly, the feasibility of adopting the estimation method in allocating cost is proved. Secondly, based on Beardwood's research conclusions the method for estimating distribution distance of collaborative vehicle routing problem is developed, according to the number of customers, demand quantity and position of customers and depots of all enterprises in alliance. At last, some benchmark instances are designed based on actual investigation data. Then the method proposed in this paper and an three-phase optimization algorithm CSV (Cluster+Savings+Variable Neighborhood Search) used to solve the instances. The results demonstrate that the computational time of the estimation method is almost negligible compared with the CSV, which could meet the requirement of on-line real-time calculation, and the cost savings percentage deviation is within 10% for all instances. What's more, the bigger the problem, the smaller the deviation.