Master Data: Think Globally, Start Locally

It's a fast-growing market, but it's also one that is rapidly evolving, gaining sophistication even as practitioners eschew big-bang projects. The technology in question is master data management (MDM), and it's growing at 30 percent to 40 percent per year. Fueling that growth, vendors are quickly expanding technical capabilities, but the toughest challenges in MDM remain breaking down political barriers, aggreeing on definitions and handling all the management tasks that follow in the wake of software implementations.

MDM encompasses technologies and processes that ensure that information about customers, products, suppliers, employees and other assets is current, consistent, accurate and agreed upon across the enterprise. Customer data management and product information management are the most practiced forms of MDM today, but the technology is getting both broader and more sophisticated, as evident in a number of recent product upgrades:

- Purisma Data Hub 3.5, released today, has been upgraded to automatically create and maintain corporate hierarchies that offer multiple views of customers. An organization that has a diverse company like GE as a customer, for example, can give sales a view by territory, accounts receivable a view by legal entity and marketing a views by SIC code.

- TIBCO Collaborative Information Manager 7.0, released September 24, adds "cross-catalog relationship capabilities" that let you correlate master data across multiple domains to discover, for example, which customers are being served by which locations or which vendor is supplying which products.

- Similarly, Initiate Master Data Service 8.0, released last month, offers the ability to identify relationships "within and between data domains," such as company subsidiaries to branch locations, consumers to households, distributors to customers, and customers to contracts, products and pricing.

MDM initiatives are typically rooted in one of three ambitions: 1. Gaining a cross-enterprise perspective for better business intelligence or customer intelligence; 2. Improving transaction management so things like purchase orders are consistent with invoices, easing cross-selling and upselling; and 3. Building a master data foundation for service-oriented architectures, ensuring consistent referential data for building composite applications and information services.

Early approaches to MDM envisioned a "big bang" approach of creating a universal, enterprisewide definition of, say, customers, with a single repository intended to feed all operational systems. That was possible from a technical perspective, but many projects hit the brick wall of cultural and political conflict, with disagreements on data definitions, attributes and ownership.

"What we're seeing smart enterprises doing is starting out incrementally," says Bob Hagenau, founder and vice president of products at Purisma. "They might start just with identifying customers consistently across all systems. There's not a lot of push-back on that, but you can start to do some of the business intelligence roll-ups because you can combine the sales coming out of, say, JD Edwards and SAP because you have a common customer identifier."

Even if the journey starts with a single domain, TIBCO and Initiate are among the growing ranks of vendors looking beyond the customer- and product-oriented projects that are popular today. "Our position is that you should have a common platform for managing all types of data," says Neeraj Gokhale, general manager of information management at TIBCO. "Not only can and should you store multiple domains on the same platform, there is intelligence to be gained in understanding the intersections between those domains."

Would-be MDM practitioners should keep in mind that only 10 percent of the challenge is in implementing the software, according to an April AMR Research report entitled "MDM Strategies for Enterprise Appications." The bulk of the challenge lies in "establishing governance and documenting the MDM architecture," and "data remediation" such as cleaning and conforming data to new MDM rules, resolving discrepancies, and filling in missing information.