A Soft Computing Based Customer Lifetime Value Classifier for Digital Retail Businesses

Abstract

This era of Big Data and digital marketing has created varied scopes and alternatives to spend investments, in order to maximize profit in business. However, identifying the right target group (TG) to invest the right resources emerges as a critical problem, for both acquisition and retention. Customer lifetime value (CLV) is a metric that may aid in such decisions, the calculation of which seems quite generic in traditional literature. This paper recognizes the potential of a large, user-specific transaction dataset from a retail business operating online and proposes an AI-powered CLV classifier. For this, we exploratorily mine customers’ buying tendencies from their geographic, monetary, chronological, categorical data using PostgreSQL and then, use those as input features to a TensorFlow-coded neural network (NN) in order to predict their profitability in different ranges. We further validate the features using statistical inference, the optimized feature set resulting from which is again passed through another NN, providing more crisp metrics in lesser computation. The training data is prepared by iterative thresholding on customers’ number of days ordered. An SQL-backed recommender, working on maximum buyers’ tendencies, is also proposed to cross-sell and up-sell customers in order to maximize the said CLV. The research indicates ways to filter out factors influencing CLV and endorses the applicability of modern soft computing in devising business-specific solutions by exploiting commonly available users’ databases.