Data Model

In all three models, the design process must include a common data model for the hub database. In the repository model, the MDM data model becomes the hub-database data model. The model includes mapping from the application data models to the MDM data model, but these mappings are used only to create the hub database and define the application changes required to modify the application to use the hub database as the source of their master data.
The other two hub models also require an MDM data model and mappings from the current applications, but they are used differently. In the registry model, the data model is used to define queries and views, and the mapping is used to do the necessary transformations to map the application data to the MDM data model in each query. In the hybrid model, the common attributes are replicated to the hub database and the non-common attributes are transformed as part of queries, so both kinds of mapping are used. Almost by definition, there will be alternate mappings for some attributes, and rules must be defined for which mapping to use. For example, a customer address is generally stored in several databases so rules must be defined to control which address to use first and which alternate to use if the preferred address isn't available. (These business rules can get to be pretty complex if many databases are integrated in the MDM hub, so I will talk about business rules later in Workflow and Business Rules.) The data models and business rules are documented in the MDM metadata and should be used as required to implement data-driven processing for populating, maintaining, and querying the MDM hub data.