Forecasting Customer Lifetime Value: A Data-Driven Approach to Optimizing Marketing Budget Allocation

Abstract

In a competitive marketplace, firms need reliable forecasts of Customer Lifetime Value (CLV) to guide marketing spend. This study investigates a data-driven framework that forecasts CLV and links the predictions to budget allocation across loyalty programs, premium offers, and discount strategies. Using transactional, behavioral, and demographic data, we compare established statistical baselines with machine learning methods and a hybrid model that combines probabilistic features, sequence modeling, and learned embeddings. The hybrid approach captures purchase frequency, churn likelihood, and spending patterns while remaining practical for managers. We evaluate the framework on two retail datasets and a combined sample. The hybrid model reduces error and improves ranking quality over BG/NBD and tree-based methods, enabling more consistent identification of high-value customers. We then translate forecasts into action by simulating budget allocation and reporting gains in Return on Marketing Investment (ROMI) when targeting segments defined by predicted CLV. The results show that precise CLV forecasts support better campaign selection, stronger retention, and higher long-term profitability. This work bridges data science and marketing practice by showing how a hybrid CLV model can balance short-term promotions with sustained customer value and inform resource allocation at scale.