Abstract: In order to know the basic information of cigarette retailers and analyze their basic characters,on the basis of the nature of tobacco industry,a model was developed with an index combination method by integrating the business status,market type,sale behavior and number of cigarette brand sold.Retailers were clustered with SOM(Self-organizing Feature Maps) method,the results were further verified from the sides of minimum inter-cluster distance and the sum of intra-cluster square dispersion,then the number of clusters for retailers were finally determined.The results showed that the best results were obtained when 8713 retailer samples were grouped into 7 clusters.The corresponding service and management suggestions were provided for the retailers in each group in terms of marketing service,market management,customer relationship management,etc.