Master Data Management Implementation Styles

To this point we have defined master data, described the different types of data in an organization and discussed the processes necessary to support Master data management (MDM). MDM programs differ considerably across organizations. The MDM implementation style depends greatly on an organization's core business, corporate structure, data integration, and goals. At a high level, there are four distinct styles of MDM, which require different best practices and technology tools. In practice, these styles are not mutually exclusive in that an organization might have different sets of master data (MD) that are implemented using each of the different styles. This hybrid is often more a matter of an organization’s de facto status rather than their future strategy.

Coexistence Style

In this style, MD is stored in various locations. This style requires the selection of a physical golden record instance that is synchronized with source systems. This style can contain some master data where practical. However, where this is not practical, this style can contain links to other master data sources. This style, thereby, is often a practical choice rather than a strategic vision. It does allow for incremental growth. It is less elegant than some other solutions and will present data synchronization issues. It is sometimes called synchronized MDM (master data management).

Transactional Hub Style

One system becomes the one and only source of master data. A given set of data is selected as a system of record and updates happen directly to this system of record. As updates happen, MD is cleansed, matched and enhanced. After updates are accepted, they are distributed to using systems. This is a good solution but costly and difficult to implement. It is sometimes called Repository MDM.

Registry Style

The registry system contains pointers to where master data lives in operational systems. This method provides read-only source of MD for downstream systems. The registry MD system holds minimal identifier and simply provides cross-reference to data managed in other systems. Data quality is still defined at the source. The registry selects “best version” of data dynamically via matching at the hub. Since there is change to existing data, it is easy to implement. It is also a useful starting point for implementation of MDM. Because of all the connections, this solution can suffer from some performance issues

Consolidation Style

The MD system contains master data as a copy. Data from various sources is merged into a single managed database. En route, the data is cleansed, matched and integrated to provide a single golden record. Source data is left as is. Source data quality issues not necessarily addressed but may be addressed separately from the consolidation. It is suitable for enterprise-wide view of data for either operation or analytical purposes. For operational this style can be exemplified by an Operational Data Store (ODS). For analytical data it is exemplified by a data warehouse. 

Our review of master data concepts in the past several blogs is fairly well rounded for now. In the next issue we will re-focus on data modeling and address considerations for practical use of data modeling