Survival data mining in the telecommunications industries: usefulness and complications

Decision makers in business industries have seen the change from the old economy to a new economy. The old economy is goods-centred, transaction-based in nature, focused on customer attraction, and product-based thinking whereas the new economy is service-centred, subscription-based in its nature, focused on customer retention, and customer-based thinking. The firm of today has to evaluate its performance by taking the customers into consideration. It is important to study customer lifetime value, which is the net present value of customers’ profit over a time period. The two main components used to build the customer lifetime value model are customer length of service (which is represented by customers’ survival curve) and customer monthly margin. While the customer monthly margin can be obtained from an accounting model, the major problem is customer survival time. In this study we take into consideration the telecommunications industry which represents a good example of subscription-based business where customer and customer relation is the vital factor in success. The nature of customer survival time in such an industry has brought many complications in customer survival modelling. Some of these complications are non-smooth survival functions, non-smooth spiky hazard functions, the possibility that both customer and company can initiate the churn, multi-churns, and multi-reactivations. In this study we review survival data mining and we discuss how survival data mining approaches are eligible to represent complications involved and how they are beneficial from both a practical and methodological point of view. A model for customer survival time is suggested and discussed and challenges to apply this model in practice are raised.