An Order Dispatch Approach in Large-scale Telemarketing System
Shi Qundong, Zhao Jiangjiang, Xiaokun Zhang, Dongdong Li, Xiangyi Meng · 2019
In traditional order dispatch rule in large scale telemarketing, employee selects one of hundreds orders competing with others during limited time. However, the selected order maybe is not suitable which mean employee finish the order with a poor success rate, while other employee would get a better success rate on this order. The most important thing is that it leads to a lower global success rate which reduces employee income. In this paper, we propose a novel approach used multi-sources information to help employee select order and improve the global success rate. Generally, we model this as assignment problem instead of recommend problem: 1) predict the employee-order success rate by historical data, employee information and order information 2) base the values in 1), dispatch order by combinatorial optimization algorithm. In order to balance tradeoff between present and future, we filter some assigned results out under operational conditions, and dispatch them with new orders in next turn. The proposed approach significantly improve employee income through online A/B test and play an important part in ChinaMobile's telemarketing system.