Master Data Management (MDM) is often defined as "management of master data (customer, product, supplier, etc.) that is shared across disparate IT systems and groups." However, this simplistic description doesn't do justice to the complexity of the MDM's task and problem area. Master Data Management encompasses areas such as Customer Data Integration (CDI), Product Information Management (PIM), and Global Data Synchronization Network (GDSN); and partially overlaps the areas of Identity Management System (IdM), Business Intelligence systems, data quality, and data integration. This broad area of potential application causes multiple perspectives, diversity of stakeholders, and a fair amount of confusion across clients investigating an MDM solution.

The business need for MDM is made manifest both implicitly and explicitly. Its utility tends to be obvious in efforts around conformance and auditing, accurate reporting efforts, and a single view of the customer initiatives. However, MDM is often also a hidden requirement for successful consolidation projects after mergers and acquisitions. Its value in terms of return on investment, cost savings (reduced storage, reduced analysis, development, and maintenance, etc.), increased revenue (consistent view master data, reduced time to resolution, and effective decision making), and competitive advantage (operational efficiency, improved visibility to company performance, etc.) has been well documented by multiple reputable groups and authors (AMR Research, Forrester Research, Gartner, and the Yankee Group) so we won't explore the existing benefits that the reader can easily reference. We will however discuss the benefits of MDM as they relate to SOA enablement.

MDM systems can be "federated," "integrated," or "hybrid" reflecting a combination of the first two fundamental architectures. These three types of system characteristics are as:

• Federated MDM - cross references key identifying information from participating systems to implement a registry-style solution. The main benefit of a federated solution is non-intrusiveness on participating systems that maintain their original context.
• Integrated MDM - stores all master data information from all participating systems in a centralized MDM repository. This centralized repository houses the "gold copy" of all master data information. The main benefit of the integrated approach is that it provides the most complete, accurate, and consistent single view of master data.
• Hybrid MDM - stores common data elements from participating systems creating a "light gold copy" of the master data, while disparate elements are referenced from their original system of record. The benefit and drawback of the hybrid solution is the partial combination of the federated and integrated benefits.

Service Oriented Architecture (SOA)
From a systems design perspective, SOA is an architectural approach based on distributed computing principles. SOA has numerous other aspects in topics as diverse as business process design and IT governance. However, these aspects go beyond our scope here.

As an architectural paradigm, the participating components of a SOA system include: service providers, service consumers, intermediary services, and registries. A service provider publishes a service in the registry to be consumed by a service consumer who can identify the interface, purpose, and location of the service from the registry. Intermediary services intercept and handle operations that are common across services and can be leveraged instead of recreated every time. Typical intermediary services include: authentication, auditing, logging, monitoring, and message routing. All communications are done through commonly agreed on standards (UDDI, SOAP, WSDL, XML, HTTP/SSL). The design principles governing SOA are primarily object-oriented paradigms extended to address the service-oriented requirements. These service design principles include: loose coupling, service contract, abstraction, composability, autonomy, reusability, statelessness, and discoverability.

Services access information from a data services layer. A data services layer provides an abstraction layer between producers and consumers of data. The data services layer presents consumers with a virtual aggregated view of data from multiple data sources in a consistent and centralized fashion. The layer's interface supports all consumers (human, application, external parties, or business services) while providing agility to data source providers.

A data service layer offers many benefits. Consumers are insulated from complexity, location, and changes in source data systems through abstraction. Providers have the flexibility to change underlying data schemas without impacting consumers through abstraction. Companies can centrally manage, monitor, measure, and report on the enterprise view of the data and metadata.

The three main categorizations of services in the data services layer are: Enterprise Data Services, Enterprise Metadata Services, and Enterprise Data Platform Services.

• The Enterprise Data Services area encompasses all the services around the data. For example, a request to be addressed by this area would be: Retrieve "gold copy" of "customer A" record.
• The Enterprise Metadata Services area includes all the services around the metadata. This area would address items such as: Retrieve master data schema of "customer A" record.
• The Enterprise Data Platform Services area supports all the services around the platform including management, monitoring, and reporting. An example of a request here would be: Retrieve MDM system, quality of service targets.

Services are defined in each area based on function (examples are shown in Figure 1). In each service and across all three areas, methods for search, access, create, update, delete, manage, monitor, and reporting functionality should be evaluated for applicability and realization.

MDM Meets SOA
MDM and SOA evolved separately but share many design principles.

• "Contract first" applies to the interfaces in MDM and the service definition in SOA
• "Reusability" applies to data through conformance in MDM and services through SOA principles in SOA
• "Discoverability" applies to data through the master data repository in MDM and services through registry in SOA
• "Abstraction" applies to source system complexity and MDM and underlying service complexity under SOA.

MDM, however, typically doesn't embrace SOA's "loose coupling" principle. Extending MDM with loose coupling allows support for SOA's semantic conformance needs.

As MDM practitioners contemplating supporting today's SOA systems, we need to become familiar with SOA standards and strive for loose coupling with external systems. Eliminating point-to-point interfaces and replacing them with service-enabled integration minimizes the impact of changes from integration partners and consumers. Loose coupling should be applied internally as well to create an agile MDM system. An agile service-oriented MDM system provides its data quality, conformance, and other MDM functionality as business or data "services" available for net-enabled consumption by external parties. Finally, MDM systems should be able to handle the extensible data types (XML, HTML, PDF, and e-mail) common to net-centric application and be able to expose the master data model as part of the enterprise canonical data model (CDM) for service consumption.