A new customer churn management model: Retail sector application

One of the biggest problems brought about by the increasing competition in the field of marketing is the churn of customer portfolio and the tendency of these customers leaving to the competitors. For this reason, businesses in the retail sector have started to use information technology tools to prevent these customer churn and transitions and to create loyal customers. In particular, by means of statistical programs that contain computers, databases and mathematical models, companies try to identify their loyal customers as well as those who may be lost. In this study, using the data obtained from a supermarket chain operating in the retail sector, a model proposal is presented, which includes the use of classification and forecasting techniques for which customers in the customer portfolio are churn customers and which customers may be a churn customer in the future by using this information. According to the previous studies in the literature, it is stated that the costs of all processes including the detection of customers who may be churn, preventing them from being churn and bringing them back into the scope of loyal customers will be at least 5 or 7 times less than the cost of the process of acquiring new customers with the same profitability and loyalty, in case of loss of these customers. For this reason, focusing on customers who are likely be churn is both economically beneficial and time loss, and risky situations that may be encountered in the process of acquiring new customers will be eliminated in advance. In the models presented in the study, the effects of the opinions of the experts experienced in the retail sector and the additional variables other than RFM on the performances of the classification and estimation models were examined. Expert opinions were added to the RFM-based forecasting models using the AHP method. As a result of the study, it has been revealed that the models in which expert opinion is used have accurate and positive results according to the classification and estimation accuracy performance criteria.