Informatica Enterprise Data Management

At a conference last fall, I heard Martin Brodbeck, executive director for strategic architecture at Pfizer, describe how his company, a $48-billion pharmaceutical giant, was able to employ master data management (MDM) to bring together data assets from across its global enterprise into a single, centralized data definition.

The key ingredient to Pfizer's success in this area, Brodbeck said, was not technology by itself, but enterprise governance. Pfizer's MDM effort was led by an internal business sponsor, who helped promote the concept to the rest of the global enterprise. "Master data management is much more about governance than it is about technology," he pointed out.This same sentiment was echoed a couple of months ago here at the EDM blogsite by Rick Sherman, who also pointed out that MDM wasn't solely a product or technology, but a methodology that required enterprise governance. As Rick put it: "MDM is not a product solution, but a process with the key ingredients being people and politics. If you don't have data governance and an organization ready to commit to an ongoing effort to implement and keep MDM going, then it does not matter what product you buy."

As its name suggests, master data management focuses on creating and maintaining a consistent, accurate and standardized view of reference data in business systems across the enterprise. In fact the challenges and opportunities related to MDM are similar to those of enterprise data warehouses and service oriented architecture – both can lead to enterprise transformation, but also require deeper enterprise involvement than typical IT or data management initiatives.

Unfortunately, when it comes to addressing MDM on an enterprise scale, many companies appear to be falling short. Noted business intelligence analyst Jill Dyche, for one, says she is seeing many companies are taking a spotty or piecemeal approach to MDM, in hopes of eventually bridging these islands. "It was interesting to hear stories of how companies had tried to enable pieces of MDM in bite-sized chunks, only to confront the reality that the small components of MDM don't really solve the entire problem of reconciling and integrating disparate master data on behalf of the enterprise."

However, MDM usually doesn't start out as a global initiative that immediately changes the data model for the entire enterprise in one fell swoop. Jill’s company, Baseline Consulting, in fact, has identified an evolutionary path that MDM takes, along five levels of maturity.

Level 1: List Provisioning: A company with no MDM typically has "no common agreement of data definitions across organizations," and "no common agreement of data definitions among or between different systems." All efforts to migrate data between systems are manual, and the quality of any data that is brought over is suspect. Company’s first foray into MDM may involve processes to update and change data that is moved between applications. The ability to exchange data between business units depends on the cooperation and relationships between data owners within these business units (who may deliver lists to each other via email attachments). However, "there is no systematic and rigorous way of ensuring changes to the master list," and "defining and maintaining master lists involves significant meetings and human involvement." Data issues are handled manually.

Level 2: Peer-Based Access: Centralized data definition and standardization is introduced. At this stage of MDM, "applications are aware of a central repository or list of master data, and have hardcoded logic in order to interact with that system," and "a data model exists and serves to uniquely identify each master record." However, data management is still tightly tied to individual application environment. Data access and integrity rules "must be managed by the individual application systems."

Level 3: Centralized Hub Processing: While companies at MDM Level 2 centralize data access and control across various applications, companies moving to the next level actually support a consistent view of data through a centralized environment. In this stage of evolution, "master reference data is centrally managed," and "business-oriented data rules and the associated processing have been centralized." In addition, there is a data steward or owner that manages and maintains the quality of the master data.

Level 4: Business Rule and Policy Report. Centralized rules management is established for master data. Level 3 ensures a centralized view of data with a single version of the truth, and companies moving to Level 4 ensure that the master data “reflects the company’s business rules and processes to substantiate accuracy.” At this level, companies have "implemented a process-driven data governance framework that supports centralized business rules management and distributed rules processing." In addition, the company has implemented SOA as "a means to integrate common business methods and data across applications." The MDM environment is also highly automated.

Level 5: Enterprise Data Convergence. At the highest level of MDM, all data is entirely independent of applications and systems. Master data and application data “are one and the same,” and are indistinguishable from one another. The MDM hub is "fully integrated into the application system environment, propagating data changes to all the systems that need the master data," while MDM data access is fully integrated with application processing. "Application development is able to capitalize on both business-based application services and business-based data services."