A machine learning approach for direct markrting in digital channels

Abstract

Targeting the right customers to offer specific products is a critical and a chal lenging problem in direct marketing. Creating an effective communication plan for the customers in a multi-campaign environment is the core problem of direct market ing strategies. In this study, a machine learning approach is proposed to make direct marketing communication more effective in the telecommunication sector. The methodology is designed for a multi-campaign environment, where the aim is to target the right customers and to contact them in specific digital channels under certain business constraints such as budget and contact frequency. The solution ap proach is structured in two stages integrating machine learning modeling and optimal decision making. In the first stage, a random forest algorithm is employed. The aim is to estimate the buying probabilities of the customers based on their historical data. The main model outputs are the customers’ expected response probabilities on the spe cific products. In the second stage, the outputs of random forest model are fed into the optimal decision-making process. The goal is to achieve the optimal direct marketing plan. A mathematical optimization model is formulated by considering probabilities of buying, and business constraints with the objective of maximizing revenue. In order to illustrate this methodology, three different products of a telecom company are selected. The final results are compared with model free approach based on expected revenue. This analytical study can be easily extended to solve similar problems in other sectors.