In production networks, production must run efficiently across company boundaries. Companies must be able to react quickly as a single unit. Two trends are influencing this situation: On the one hand, the progressing servitization leads to the increased offering of digital services in the field of manufacturing. From the literature, it is known that digital services let manufacturers, suppliers, and industrial customers interact more closely and frequently in a production network. On the other hand, the concept of the digital twin is trending. It promises the real-time prognosis and control of production systems. Although the concept of the digital twin itself can be vague there are some technologies trying to implement the digital twin of production. The asset administration shell (AAS) is an example of such a technology that draws growing attention. Picking up the initial situation these two trends could be used to create a feedback loop between the production system and network and thus improve the overall efficiency in production networks. Based on this idea, the paper first presents an approach to model systematically a possibility for a feedback loop orienting to the business model concept. Second, a reference architecture is derived from the RAMI 4.0 standard. The specified reference architecture is the basis for the specific implementation. Third, a procedure is developed to implement a specific architecture. For implementing an architecture, the usage of the asset administration shell is assumed. Finally, the approach is validated in a use case from the high precision weight industry.