An Internet of Things (IoT)-Enabled Multi-Source Sensor and Systems Data Fusion Architecture for Smart In-Store Retail Demand Forecasting and Inventory Replenishment 

• Ada John

  ladoke Akintola university of technology

  Author

This study addresses the critical challenge of demand forecasting and inventory replenishment in physical retail stores, where reliance on historical point-of-sale (POS) data alone fails to capture real-time customer behavior, shelf-level dynamics, and environmental factors that drive purchasing decisions. Despite advances in AI-based demand forecasting, no validated framework exists that integrates heterogeneous IoT sensor data streams—including in-store beacons, RFID tags, environmental sensors, and traffic counters—with enterprise systems data for granular, store-level replenishment decisions. This research develops and evaluates an IoT-enabled multi-source data fusion architecture that combines sensor-derived behavioral signals with traditional transactional and inventory data. Using a design-based research methodology, the study constructs a three-layer fusion pipeline (sensing, integration, and prediction layers) and validates it against historical sales data and simulated sensor inputs. The proposed architecture achieves 89.4% forecasting accuracy, significantly outperforming baseline time-series and machine learning models. Feature importance analysis identifies dwell time, shelf-level traffic, and ambient temperature as the strongest non-transactional predictors. The framework offers a replicable approach for retailers seeking to operationalize IoT infrastructure for demand forecasting and demonstrates that sensor fusion meaningfully improves replenishment timing and reduces stockout risk. Practical implications include a phased implementation roadmap for store-level IoT deployment.

• pdf

Copyright (c) 2026 Ada John (Author)

This work is licensed under a Creative Commons Attribution 4.0 International License.