The rise of the fourth industrial revolution has propelled research and development in Digital Twins (DTs) in the manufacturing sector. DTs establish a crucial link between physical systems and their digital replicas, enabling real-time monitoring, process optimisation, and predictive analysis. This paper introduces a historical data management framework for DTs in manufacturing, addressing existing research gaps. Our framework includes a data classification scheme based on change frequency, distinguishing static, dynamic, and periodic data. We propose an architecture involving registration, data transformation for unified schema creation, utilisation of queues for asynchronous data storage, and a tailored database selection process. Implementation in a UR5e 6-axis collaborative robot’s digital twin demonstrates the practicality and efficiency of the proposed framework in the aspect of the latency and throughput of the digital twin. The proposed architecture is benchmarked against existing architectures, and the results are discussed.