Developing Causal Uplift Algorithm for US Omnichannel Personalization Optimizing Lifetime Value Predictions

This study proposes a novel causal uplift learning framework for omnichannel personalization in the United States retail and digital commerce ecosystem, with a specific focus on optimizing customer lifetime value (CLV) through incremental treatment effect estimation. Traditional personalization systems rely on predictive models such as gradient boosting, deep neural networks, or collaborative filtering to forecast purchase propensity; however, these models fail to isolate causal impact and often optimize correlation rather than true incremental revenue. To address this limitation, we introduce the Hierarchical Counterfactual Lifetime Optimization Network (H-CLOuN), a hybrid causal meta-learning architecture that integrates doubly robust estimation, temporal uplift modeling, and reinforcement-based policy optimization. The proposed algorithm combines (i) a representation-learning encoder for cross-channel behavioral embeddings, (ii) a multi-head T-learner structure for heterogeneous treatment effect (HTE) estimation across email, mobile push, paid media, and in-app promotions, and (iii) a dynamic policy gradient module that updates action policies using incremental CLV feedback. The causal objective function is defined as: (max⁡)┬(π(a∣x)) E[τ(x,a)⋅ΔCLV(x,a)] where τ(x,a)represents estimated uplift and ΔCLV(x,a) represents discounted incremental lifetime value. Empirical evaluation is conducted using simulated and real-world omnichannel transaction datasets comprising 2.4 million U.S. consumers across 12 months. Comparative benchmarking against leading causal frameworks including X-Learner, Causal Forest, DragonNet, and Doubly Robust Learners demonstrates superior performance of H-CLOuN. The proposed method achieves a 14.7% improvement in Qini coefficient, 11.2% uplift in incremental CLV, and 9.5% reduction in treatment misallocation relative to the best-performing baseline. Graphical evaluations include Qini curves, cumulative uplift gain charts, and incremental CLV growth trajectories across treatment intensities. Results confirm that causal uplift optimization significantly outperforms traditional propensity-based personalization, particularly in high-heterogeneity consumer segments. The framework provides a scalable solution for enterprise marketing systems seeking evidence-grounded personalization strategies aligned with measurable revenue impact.