Campaign management design based on segmentation by rank clusters

Abstract

When building campaign management systems, two main aspects are important from an algorithmic point of view. First, many customers need to be classified with respect to their (bank) product portfolios, which means ranking the potential products for advertisement such as direct mailings, e-mails, website or mobile app advertisement. Second, campaigns for different customer segments need to be planned in advance to balance the costs and the success of marketing activities. For these tasks, we propose a process, starting with rule-based classification, including an evolutionary algorithm combined with a nearest-neighbor technique to obtain understandable, optimized rules. The classification results lead to product rankings for each customer. We cluster the rankings with p-means, using a modified Canberra distance, so that we obtain customer segments. Then, we design a new index for measuring the assignments of campaigns with products to the clusters. With this campaign clustering index (CACLUX), we gain the potential to plan the campaign design for customer segments in advance, comparing poor assignments with good ones. Overall, we present a line of action that generates on the basis of rule-based customer classification and product rankings quantified assignments of campaigns to clusters.