Digital Twins are commonly used as virtual representations of physical objects in manufacturing industries. Information for Digital Twins may be collected from multiple sources and stored in a distributed manner, leading to a Federated Digital Twin. Since decisions about such a federation are crucial for the system and its architecture, they should be guided by reliable and well-evaluated methods. However, current research is focused on distributed data sources but is missing decisions about the distribution of the digital twin itself. We present an approach to partition Federated Digital Twins by classifying information types, computing resources, and concerns of data suppliers. Furthermore, we show how decisions are made based on the Decision Model and Notation standard and evaluate the approach using an industrial case study.