Design, implement, and evaluate AI-powered customer analytics for SMEs

Thesis Abstract

The rapid digitization of consumer marketplaces has intensified competition for small and medium-sized enterprises (SMEs), yet many SMEs struggle to extract actionable insights from customer data due to limited analytics capabilities, fragmented data sources, and resource constraints. This study addresses the problem by designing, implementing, and evaluating an AI-powered customer analytics solution tailored for SME contexts, aiming to democratize advanced analytics and improve marketing, sales, and service decisions. The objective is to develop a scalable analytics framework that integrates internal and external data, deploy machine learning models for customer segmentation, propensity scoring, and churn prediction, and assess the solution’s impact on decision effectiveness and business outcomes. The research adopts a design, implement, and evaluate approach grounded in the Information Systems Success Model (DeLone and McLean) and the Technology Acceptance Model (TAM) to examine adoption, use, and value realization. The study is conducted in three phases. In Phase 1, 280 SMEs across manufacturing, retail, and services sectors within the mid-sized urban regional cluster are surveyed to assess data readiness, analytical needs, and perceived usefulness. In Phase 2, a prototypical AI-powered analytics platform is designed and implemented in collaboration with 12 SMEs, featuring data integration from transactional systems, customer Relationship Management (CRM) data, and publicly available market signals, withMachine Learning (ML) modules for RFM-based segmentation, gradient boosting for churn prediction, collaborative filtering for personalized recommendations, and explainable AI components to ensure transparency. The implementation employs a mixed-methods data collection plan quantitative logs from the platform capturing model performance (precision, recall, AUC-ROC, lift) over six months, and qualitative interviews with 24 SME executives to explore usability and decision-making impact. Phase 3 executes a quasi-experimental evaluation over nine months, comparing SMEs using the analytics platform (treatment group, n=12) with a matched control group (n=12) on key business outcomes, including customer lifetime value (CLV), average order value (AOV), basket size, and marketing ROI, while controlling for sector and firm size. Data analysis leverages regression analysis to quantify relationships between analytics use and performance metrics, time-series analysis to observe trend changes, and thematic analysis of interview transcripts to illuminate adoption dynamics and perceived value. Model performance will be validated through cross-validation and backtesting, while robustness checks address potential confounders and data quality issues. Expected findings indicate that the AI-powered analytics platform enhances segmentation accuracy and churn prediction, leading to more targeted marketing and retention strategies. Quantitatively, treated SMEs are anticipated to show statistically significant improvements in CLV (effect size medium, p<0.05), AOV (p<0.10), and marketing ROI (p<0.05) after six to nine months of usage. Explainable AI modules are expected to facilitate managerial trust and informed decision-making, while adoption-related factors such as perceived usefulness and data readiness are predicted to mediate the relationship between platform use and performance outcomes. The study also anticipates sector-specific nuances in analytics benefits, with higher gains in retail and services due to greater customer touchpoints and transaction frequency. The contribution to knowledge includes (1) a practical, scalable design blueprint for AI-driven customer analytics tailored to SME constraints, (2) empirical evidence on the effectiveness and mechanisms through which AI analytics improve business performance in SMEs, and (3) an operationalized framework linking data readiness, user adoption, and value realization within SME contexts. Theoretical contributions extend the application of the DeLone–McLean Information System Success Model and TAM in the specific setting of AI-enabled customer analytics for SMEs, and provide insight into the role of explainable AI in fostering managerial buy-in. The study offers actionable recommendations for SME owners and policymakers on data governance, platform configuration, change management, and investment case construction. The main conclusion is that carefully designed AI-powered customer analytics, aligned with SME capabilities and governance practices, can meaningfully enhance customer-centric decision-making and financial performance, with adoption and data readiness identified as critical enablers. Recommendations include phased implementation with iterative value demonstrations, investment in data quality and integration, training focused on interpretability, and ongoing monitoring of ethical and privacy considerations.