Smart manufacturing uses data generated within manufacturing systems to provide intelligent and flexible control of production processes. To realize the vision of smart manufacturing, digital twins play a key role. As a virtual representation of any real-world counterpart, the digital twin enhances various use cases by providing analytical, simulation and replication capabilities. Most research works focus on potential application scenarios for digital twins within manufacturing. Despite its great potential, few works address the generic creation of digital twins within an industrial environment. To fill this gap, we introduce a data-driven framework for digital twin creation in industrial environments. The core creation process of our framework consists of four vital parts, and the respective data required, to build a digital twin. Before data is even acquired, we argue that individual conditions must be set to determine the overall scope. We further address existing interactions and dependencies between components. To validate the framework, several semi-structured expert interviews are carried out. Furthermore, related works are identified using a systematic literature review - followed by a comparison to our proposed framework.