Machine Learning for Predicting Customer Lifetime Value in Retail

Abstract

Customer Lifetime Value (CLV) has emerged as a cornerstone metric for evaluating long-term customer profitability and guiding strategic marketing decisions in modern retail. Traditional statistical approaches to CLV estimation often struggle to capture non-linear customer behavior, high-dimensional data, and dynamic purchasing patterns. The proliferation of machine learning (ML) techniques has revolutionized predictive analytics by enabling more accurate, adaptive, and scalable CLV prediction models. This study investigates the application of machine learning algorithms—namely Linear Regression, Random Forest, Gradient Boosting, and Artificial Neural Networks—for predicting CLV in a retail environment.

Grounded in Customer Equity Theory, Predictive Analytics Theory, and Data-Driven Marketing Frameworks, this research develops a comparative predictive framework and empirically evaluates model performance using transactional retail data from 6,200 customers. The results demonstrate that ensemble and deep learning models significantly outperform traditional regression methods in CLV prediction accuracy. The study contributes to both marketing science and machine learning literature while offering actionable insights for retail managers seeking to optimize customer acquisition, segmentation, and retention strategies.