Retail data analysis has been identified as a crucial component in the pursuit of E-commerce 5.0. The recent rapid development of Information and Communication Technology (ICT) has revolutionized retail data analysis by providing advanced technologies such as big data analysis and machine learning. However, the privacy of customers has become a significant concern, making retail institutions hesitant to share their customer data. This reluctance forms isolated data islands, hindering the realization of comprehensive retail data analysis. I propose a federated learning-based retail data analysis framework, FEDRETAIL, to address this challenge. This framework allows retail data analysis federations to be formed by several institutions. None of these institutions need to exchange their customer data with each other directly, and they always keep the data in their place to ensure their customers’ privacy. We apply our FEDRETAIL framework to analyze a retail dataset via different federated learning paradigms. The experimental results show that our framework not only guarantees the customers’ privacy but also effectively breaks the borders of data islands by achieving higher analysis quality. FEDRETAIL framework closely approaches the performance of centralized analysis, which requires data collection in a commonplace, posing a risk of privacy exposure. In conclusion, the FEDRETAIL framework presents a promising solution for privacy-preserving retail data analysis in the era of E-commerce 5.0, paving the way for more effective and secure data-driven decision-making in the retail industry.