Modelling Churn as a Long-Term Profit Optimization Problem

In typical recommender systems, the main goal would be to maximize the probability that customers would purchase the product/s that the system recommends. However, this method may be a double edged sword to some businesses as the products these recommender systems recommend to customers, while suitable for the customers, may not be very profitable for the company. Therein lies the delicate trade-off where the recommended products that are very profitable to a company may not be suitable for the customers which may cause them to switch to a competitor and vice versa. Therefore the products recommended must be both sufficiently suitable to the customer and sufficiently profitable to the company. This paper considers a recommender system with the goal of maximizing the long-term profit for a company (i.e., the sum of immediate profit made by the recommended product and the expected profit of future product purchases made by the customer). In this approach, the customer gains from obtaining favourable recommendations and the company benefits from acceptable profits from a long-term client. We compare long-term profits gained from a customer using the proposed approach with that of Recommender Systems that optimizes customer suitability as well as the company's immediate profit.