We present the practical application of the Digital Shadow (DS) in injection molding production to manage the resulting heterogeneous volumes of data. A Digital Shadow of a system is a task- and context-dependent, purpose-oriented, aggregated data set that forms the basis for informed decision-making. We examine two use cases located at different levels of the automation pyramid: Geometry-dependent process configuration (Control and operational level): The DS is used at runtime to optimize the injection profile in order to achieve a more uniform melt front velocity and thus higher part quality. Production planning and control (Planning level): The DS is used to calculate optimal machine allocation plans based on objectives such as minimum total setup time or minimum total lateness of orders. A reference model of the Digital Shadow was designed and refined within the German Cluster of Excellence "Internet of Production" (IoP) at RWTH Aachen University. For both use cases, we create a concrete data model and demonstrate how the different positioning within the automation pyramid impacts the specific data structure, the system models used, and the real-time requirements. These findings serve as a guideline for the development of future Digital Shadows in further application scenarios.