TDWI | Training & Research | Business Intelligence, Analytics, Big Data, Data Warehousing

MDM Planning

Master Data Management is Inevitable... So Get Ready

What is master data management? How should I approach it? Is it something I truly have to do?

• By Philip Russom, Ph.D.
• August 10, 2005

For many people in IT, this decade is all about integration. It's about integrating customer data, integrating application silos, integrating data for BI purposes, integrating with partners, integrating with governments, and integrating through new interfaces like Web services. As if the list weren't long enough, the practice known as master data management (MDM) has become a candidate for even more integration work.

Hence, a lot of data and integration professionals are asking: "What is master data management? How should I approach it? Is it something I truly have to do?" This article seeks to answer these questions.

Master Data Management is a Data Integration Practice

Master data consists of facts that describe a business entity. This is particularly useful when multiple IT systems across a company identify the entity differently. Master data creates a system of record for that entity, so users have a single view that is more complete, accurate, and standardized than that of individual systems. The single view enables entity analysis, data synchronization, data quality, transactional integrity, and many other uses.

Today, the business entity is usually a customer or product. The single view can be created in a physical, redundant database, typically some form of ODS or other data hub. But the trend in master data management is to use federated or virtual styles of data integration.

Master data management is the practice of acquiring and integrating master data. MDM practices can be enabled by various integration and database practices, technologies, and products. These include ETL, EII, EAI, databases, homegrown operational data stores, hand coding, modules of packaged applications, vendor products, consulting solutions, metadata management, entity modeling, and so on. Regardless of how you combine these in your MDM recipe, data integration in some form is always a prominent ingredient. Even if you buy an MDM product (from SAP, i2, Kalido, etc.), you still have to customize it to your database schema and data management practices so that it feels more like a practice than a product.

Multiple Trends Point to MDM, Making It Inevitable

Since the practice of master data management is enabled by a potentially long list of technologies, it is influenced by all these and their trends. MDM's intersection with other practices and technologies gives users common ground for understanding and practicing MDM. But these numerous intersections also make MDM inevitable for some users and the businesses they support.

• Data integration is booming this decade. On the one hand, this is partly driven by the "do more with less" mantra of the current economy, because data integration helps you get more out of preexisting applications and databases. On the other hand, many corporations are still correcting their shopping sprees of the late 1990s, where they deployed lots of new applications, and so must now integrate these silos. In fact, many corporations have spent more this decade on integrating old systems than deploying new ones. Master data management, as an integration practice, is relatively late to the integration boom, but is certainly a part of it.
• Integration is more likely than migration and consolidation. IT personnel would like to reduce the number of redundant applications and databases for better data visibility and less administrative cost. But the question is: Do you migrate, consolidate, or integrate? In these cases, integration wins, because it's cheap, fast, and non-intrusive, compared to migration and consolidation, which are time-consuming, expensive, and disruptive. As a viable integration alternative for these cases, master data management has already proven itself in the forms of customer data integration and product information management.
• Virtualization is on the rise, and data virtualization is a subset of this trend. Federated databases and virtual EII have gained ground this decade, and data grids and data services are just emerging. Data virtualization also affects master data management, which can be federated or virtual (built on EII or other metadata views). Even so, MDM can also be physical (where data is copied among persistent data stores) or a hybrid of physical and virtual.
• Many companies are still on a quest for a single view of customers and products. Master data management is one way to achieve a single view, which explains why MDM is sometimes a component of customer data integration or product information management. MDM will most likely broaden, as companies pursue a single view of an employee, supplier, distributor, partner, patient, claim, policy, and so on. Obviously, data warehousing intersects with MDM, because both seek to provide "a single version of the truth." In some corporations and consulting firms, it's the data warehousing team that designs and implements MDM.
• Vendors now offer master data management solutions of various types. These come from platform vendors (IBM, HP), application vendors (PeopleSoft, SAP), and integration vendors (Ascential [an IBM company], Informatica, Kalido), plus system integrators (BearingPoint, Cap Gemini, Wipro). Master data management is also a critical component in vendor products for customer data integration (from DWL, Siperian, Oracle) and product information management (FullTilt, i2, IBM, SAP). All these products are today inherently tied to either customer or product data. The future success of MDM—whether vendor product or independent practice—will hinge on crossing from product and customer data to data about other business entities.

Customers and Products Today; Much More Tomorrow

With so many trends intersecting with master data management, it has a high probability of adoption. In fact, this data integration practice is inevitable in many user organizations. For example, companies pursuing a single customer view or product-oriented supply chain optimization absolutely have to have some form of master data management.

Today, master data management is applied to the master data of a specific business entity, usually customer or product data. Eventually master data management will extend beyond customer and product data into broad enterprise use. As this trend wends its way, the challenge for data professionals is to develop a strategy that addresses both isolated pockets of MDM today (for customer and product data) and broader enterprise use tomorrow (as MDM is applied to employee, patient, claim, supplier, and other business entities).