A SYSTEMATIC REVIEW OF PREDICTIVE ANALYTICS IN MARKETING DECISION-MAKING EXPLORING AI-DRIVEN CONSUMER SEGMENTATION AND AB TESTING

Abstract

This systematic review examines how predictive analytics supports marketing decision-making by synthesizing research on AI-driven consumer segmentation and A/B testing strategies as complementary components of evidence-based marketing systems. Using a PRISMA-guided screening and selection workflow, the review analyzed a total of 57 papers that met predefined eligibility criteria focused on marketing-relevant predictive modeling, segmentation methods used for decision support, and experimentation designs applied to validate marketing interventions and personalization policies. Findings indicate that predictive analytics is most often operationalized as an end-to-end decision pipeline in which diverse data ecosystems—spanning first-party CRM and transaction histories, digital behavioral traces, campaign exposure logs, contextual signals, and content-based attributes are transformed through feature engineering into model-ready representations that drive targeting, retention, personalization, and pricing or promotion decisions. The reviewed studies show that segmentation has expanded beyond traditional strategic planning roles to include tactical campaign governance, analytical feature construction, and high-granularity personalization through clustering, probabilistic membership models, representation learning, and deep latent-space approaches, with interpretability and segment stability repeatedly emphasized as prerequisites for operational use. The experimentation literature positions A/B testing as the central causal validation mechanism for selecting among creative variants, interface designs, targeting rules, and algorithmic policies, while also highlighting the importance of metric governance, attribution windows, statistical power, multiple-comparison control, and operational safeguards such as ramping, logging, and auditability. Across the evidence base, the strongest contributions link prediction, segmentation, and experimentation into integrated decision cycles where model outputs prioritize audiences, segment’s structure heterogeneity for analysis and governance, tests estimate incremental effects, and results update thresholds and rollout rules under real constraints. Overall, this review consolidates a fragmented literature into a coherent synthesis of methods, evaluation standards, and operational patterns that characterize how predictive analytics, AI segmentation, and A/B testing jointly function as a scalable decision architecture in contemporary marketing environments.