Effect of Personalized AI-Driven Recommendations on Customer Retention in E-Commerce Platforms

Abstract

This study examines the causal impact of personalized AI-driven recommendation systems on customer retention metrics across e-commerce platforms. Drawing on a longitudinal dataset of 2.4 million anonymized customer transaction records spanning 36 months (January 2022 – December 2024) from four major e-commerce platforms — FlipNest (India), CartBridge (Southeast Asia), NovaShop (Germany), and RetailAxis (United States) — we employ a Difference-in-Differences (DiD) framework augmented with propensity score matching to isolate the effect of AI recommendation exposure from confounding factors. Our findings reveal that customers exposed to personalized AI recommendations demonstrate a statistically significant 34.7% improvement in 12-month retention rates (p < 0.001), a 28.3% increase in average order frequency, and a 19.6% uplift in customer lifetime value (CLV) compared to matched controls. Collaborative filtering-based models outperform content-based systems by 11.2 percentage points on repeat purchase probability. However, the effects are heterogeneous: high-income segments show diminishing marginal returns beyond the third recommendation touchpoint, while first-time buyers exhibit non-linear sensitivity to recommendation relevance scores. We further identify recommendation fatigue as a significant moderating variable that attenuates retention gains by up to 22% when exposure frequency exceeds platform-specific thresholds. This paper contributes a scalable measurement framework for AI recommendation ROI and offers actionable thresholds for deployment across market segments.