Master Data Management - Implementation styles
Master data management (MDM) is a business practice that ensures that an organization's data is accurate, consistent, and reliable. It involves creating a single master record for each person, place, or thing in a business.?
How does MDM work?
Benefits of MDM
MDM implementation styles
MDM can be implemented in 4 styles and each has its purpose. You would see the combination of styles in an enterprise depending on the domain and purpose it serves.
Registry pattern:
The registry pattern is the simplest pattern. Data gets created in the source systems whether it is ERP, CRM or an operational database. The data then flows into the MDM and on the way the data gets de-duplicated, cleansed, matched and merged. The cleaned data is not fed back to the source systems. This style has no impact on the golden source applications. Each source system remains in control of its own data and remains the golden source. Since there’s no distribution of data from the registry back to the source systems, there’s also no direct impact on data quality.
In this style the MDM system is used as an index.
Consolidation pattern:
Consolidation style is an extension of the registry style. Data is consolidated from multiple sources and called as a master data hub. The hub uses data integration patterns to continuously pull together master data from the operational systems to improve quality and manage governance. Data from this consolidated repository is available for consumption by downstream applications. This style is typically implemented for analytics, business intelligence (BI), and reporting. No effort is made to clean up or improve data in the golden source systems. Improvements made to the data are limited to the hub, so only consuming applications benefit.
Co-existence pattern:
In this pattern, the cleansed data from MDM is synchronized back to the source systems. The coexistence style is used in situations where the master data can’t be used centrally and must be distributed in multiple locations throughout the enterprise. In this style, improvements find their way back to the original systems, so updates are performed within the golden source systems. Coexistence is the most complex implementation style because complex data integration patterns must be implemented for distribution and to ensure data consistency in the golden source systems and master data store. There is a single version of truth in both MDM and source systems.
Centralized pattern:
In this style, the master data is directly managed in the MDM system. Data is authored in the MDM system and not on the source systems. This approach is sometimes called as transactional MDM, which is incorrect. The biggest concern about this style is that it requires intrusion into the golden source systems for two-way synchronization: applications need to conform to use the master data repository as their backend. This conformity pattern can be difficult to implement because application designs can’t always be changed. Off-the-shelf products, for example, usually can’t be modified. Another concern is tight coupling: if you want to change the centralized repository, for whatever reason, you’re forced to change all the other systems too.
Achieving the centralized pattern means the organization has reached a nirvana state in MDM.
These four implementation styles aren’t mutually exclusive. You can start, for example, with the registry style and move to coexistence later. You can also use different styles for different datasets. You might choose to manage a reporting structure with the consolidated style but customer data with coexistence. When implementing a style, you can also blend different product implementations. It’s important to note that the implementation style shouldn’t be driven by technology, but rather by your organizational and data needs. That said, not all MDM vendors allow you to switch between or combine the different styles easily. You have to be cautious about which vendor you choose and consider which style(s) you might want to adopt.