Abstract

"The bank Banco Credito e Inversiones (BCI) started its virtual bank in 1996 and its registered customers perform currently more than 10,000 Internet transactions daily, which typically cause less than 10% of traditional transaction costs. Since most of the customers are still not registered for online banking, one of the goals of the virtual bank is to increase the number of registered customers. Objective of the presented work was to identify customers who are likely to perform online banking but still do not use this medium for their transactions. This objective has been reached by determining profiles of registered customers who perform many transactions online. Based on these profiles the bank's Data Ware-house is explored for ""twins"" of these heavy users that are still not registered for online banking. We applied clustering in order to group the registered customers into five classes. One of these classes contained almost 30% of all registered customers and could clearly be identified as class of ""heavy users"". Next a neural network assigned online customers to the previously found five classes. Applying the network trained on online customers to all the bank customers identified twins of heavy users that, however, had not performed online transactions so far. A mailing to these candidates informing about the advantages of online banking doubled the number of registrations compared to previous campaigns."