Abstract: According to the characteristics of errand purchasing, such as order sparsity, the long delivery distance, and the dependence of the delivery distance on the return of the deliverymen. In order to improve the delivery efficiency of the orders and the income of the deliverymen, a real-time order allocation and routing model is established. A strategy of long-distance direct delivery and short-distance combined based on the rolling horizon is designed to solve this problem. A numerical analysis is used to verify the effectiveness of this strategy. And a comparative study with the ordinary crowdsourcing combined order strategy is conducted. Its applicability is further studied through a sensitivity analysis of the parameters such as vehicle capacity, limited delivery distance and total order quantity. The results show that, the established mathematical model and the designed strategy have advantages in delivery timeliness and deliverymen utilization rate. This strategy is more suitable for the errand purchasing. Increasing the limited delivery distance can reduce customers' waiting time and improve their satisfaction; reducing the total number of deliverymen or increasing the order quantity can improve the income and enthusiasm of deliverymen; the average completion time per order can be reduced by increasing the limited delivery distance, and the average return of deliverymen increased by reducing the total number of deliverymen.