Author(s)

C Vercellis

Abstract

The process of marketing campaign optimization takes as input a set of offers, a set of customer segments and a set of communication channels, and determines the most profitable combinations by which offers should go to segments over channels, taking into account a set of constraints for the campaign. In this paper, we argue that the combination of data mining techniques with optimization models can lead to more effective approaches to campaign management, and to an overall improved support for marketing decision makers. Given a specific marketing task, such as customer retention or acquisition, a class of multivariate splitting rules, in which an optimization problem is solved at each node in the tree, is proposed in the first stage to derive a set of interesting segments, by scoring the customers or the prospects. Then, in the second stage of our procedure, a mixed integer optimization model is formulated and solved for the overall campaign optimization, taking as input the customer segmentation derived in the frost stage, together with the set of offers defined by the marketing managers, and the constraints on the limited resources available for the whole campaign. 1 Introduction Data mining techniques are emerging as a key tool for implementing advanced marketing approaches, such as one-to-one marketing [7] or customer relationship management (CRM) [2]. There are usually a wide variety of ways in which a company may interact with its customers and prospects. Whenever a new campaign is planned, marketing is faced with many possible choices along three main dimensions: campaign targeting, that is to whom the offer should be made;

Keywords