A dataset is considered complete if, in addition to the pure data (e.g. measured values), further information, such as origin and license for use, is also available. This requires not only the complete collection of data and metadata, but also assistance or instructions for collection and use, as well as an infrastructure for storage, publication and unique identification. Such a "value chain" is made up of various processes that should be handled by people in defined roles. In this context, it is also necessary to provide a technical infrastructure which, on the one hand, generates a precise input mask for the user's specific case, but which, on the other hand, must direct general search queries in a targeted manner to the correct data records. And of course, this entire construct cannot be controlled without rules and documentation, and must be equipped with an adapted training offering and an intuitive user interface. The training offer is also aimed at people with roles or assigned tasks in the data management system to ensure that the roles are performed with the necessary quality by defining the necessary qualifications. A plausibility check, which takes into account the expected flexibility of the system, guarantees the consistency of the data records. In one of the highest levels of sophistication, the plausibility check is followed by a kind of self-healing mechanism that suggests a data set to the user according to the specifications or makes the user's re-intervention obsolete altogether. In this contribution, the focus will be on the roles and on the interaction with the data management system. All necessary qualifications and tasks are assigned to the roles. The advantage of the subdivision into roles is that it remains open whether the additional tasks are covered by existing or new personnel or whether one person holds several roles. This means that the specific personnel approach can be adapted to the institution's own needs.