Learn about a transformer-based attention model for customer lifetime value prediction that incorporates marketing touchpoints in the customer journey, including emails, promotions, and engagement signals. Benchmarked against RFM-based stochastic models and HMMs, the model then leverages attention weights to investigate whether strategies that drive short-term conversion undermine long-term value. Current CLV models treat customer behavior as independent of marketing actions, missing predictive signal in journey sequences. This approach achieves higher recall in identifying high-value customers, with implications for acquisition efficiency and customer equity accuracy. Attention interpretability reveals that the link between promotional acquisition and lifetime value is heterogeneous, offering actionable paths to identify high-potential customers.