The rapid growth of digital retail platforms, changing customer preferences, and increasing market uncertainty have created significant challenges in managing modern retail supply chains. Supply chain performance optimisation, real-time decision-making, effective inventory management, and accurate demand forecasting are crucial for staying competitive. By facilitating data-driven prediction, pattern recognition, and automated decision support across a variety of supply chain processes, machine learning (ML) approaches provide encouraging possibilities. With an emphasis on optimisation of logistics, supplier selection, demand forecasting, inventory management, warehouse automation, risk management, and ML, this article offers a thorough examination of machine learning's uses in retail supply chain optimisation. Existing ML approaches, including supervised learning, ensemble methods, deep learning models, and hybrid optimization techniques, are analyzed based on their capabilities, benefits, and limitations. The review highlights that ML-driven solutions improve forecasting accuracy, reduce operational costs, enhance resource utilization, and increase supply chain resilience. There are still major obstacles, though, and they have to do with data quality, the interpretability of models, scalability, deployment in real-time, and interface with current systems. Intelligent, explicable, and scalable ML frameworks are necessary to enable decision-making throughout the retail supply chain, according to a comprehensive review of current studies that found research gaps. The findings of this review provide insights into current advancements and future research directions for developing adaptive and transparent machine learning-based retail supply chain optimization systems.