The adaptive engagement framework: enhancing banking customer experience through AI-powered invisible marketing

Area of Science:

• Artificial Intelligence
• Machine Learning
• Digital Banking

Background:

• AI-driven personalization offers opportunities for targeted financial recommendations in digital banking.
• Customer resistance to marketing can be mitigated by invisible marketing strategies.
• Understanding customer interaction propensity is key to effective engagement.

Purpose of the Study:

• To propose an Adaptive Engagement Framework for invisible marketing in digital banking.
• To classify customers based on interaction propensity and behavioral receptivity signals.
• To align engagement timing, channel, and content for optimized marketing.

Main Methods:

• Analysis of 45,211 customer records from Bank Mellat.
• Application of Random OverSampling for class imbalance.
• Utilized Mutual Information for feature importance and dimensionality reduction.
• Evaluated Random Forest, Decision Tree, Support Vector Machine, and Deep Neural Network classifiers.
• Developed a feature-to-image transformation pipeline for CNN evaluation, including a novel Residual Dual-Attention Depthwise-Separable CNN (RDAD-CNN).

Main Results:

• Call duration and account balance were dominant predictors of customer interaction class.
• Random Forest achieved high performance (accuracy = 0.9698).
• The proposed RDAD-CNN significantly outperformed other models across nine metrics (accuracy = 0.9906).

Conclusions:

• The Adaptive Engagement Framework provides a rigorous foundation for AI-driven invisible marketing in banking.
• Behavioral and financial signals are primary for engagement targeting, not demographics.
• Future research should focus on longitudinal validation and cross-institutional replication.