A clustering approach for knowledge discovery in database marketing

From a Marketing perspective, the Customer Relationship Management (CRM) can be viewed as a process, known as Database Marketing (DBM), for establishing a profitable interaction with clients. Currently DBM is mainly approached by classical statistical inference, which may fail when complex, multi-dimensional, and incomplete data is available. An alternative is to use Knowledge Discovery from Databases (KDD), which aims at automatic pattern extraction using Data Mining (DM) techniques. This paper exploits a clustering approach for DBM, with the intention of finding a set of simple rules which explain clusters of clients with homogeneous behaviours. This strategy was applied in a domestic distribution database taken from a multinational organization. The dataset was created after a direct marketing project, where discount vouchers have been offered to thousands of potential clients, through an own branded magazine. Each coupon included a personal inquiry, with a total of five questions. The final aim was to find the adequate customer profile for each product. In this study, a specific hygienic product was targeted, since it presented high sales. The aim was to correctly classify which clients use (or not) the voucher, using the five answers as inputs. The work involved validation and elimination of irrelevant data, extensive data pre-processing, data visualization, exploratory clustering using a Self-Organizing Map (SOM), and finally the application of a decision tree in order to achieve the set of classification rules. Regarding the results, in 60% of the data, a predictive accuracy greater than or equal to 75% was achieved. Furthermore, the readiness of the rules favoured the interpretation of the behaviour of the clients.