You don't implement MDM by application or even database. You start with one subject area-customer, product-Levy told me during a recent interview.
If I sell product to customers, product is a subject area. Customer is a subject area. If you're a data modeler or a database person, it's the main entity. It's really the things that we use or sell in business. It's people, places and things.
The problem is, IT doesn't think in nouns. IT is all about the verb: Defining, coding, testing, supporting. What's more, IT departments tend to view the world in terms of projects-fulfilling this feature request, upgrading to this release, migrating to this server. And that makes sense-projects can be funded. Nouns ... well, nouns aren't exactly a budget line item.
But as with so many enterprise-wide disciplines, master data management requires an approach that's independent of projects, Levy said. You adopt a standard for one subject, and then you force every application to use that standard. "Dura Lex Sed Lex (lat.'The law is harsh but it is the law')" warns a recent CIO.com article, "Five Practical Tips for a Master Data Management Solution":
MDM works best and produces the most noticeable result on the plain game field when no one is exempt from governance policies and everyone has to obey the established rules.
But before you get to enforcement and governance, you have to get to the master data, and that's harder than it sounds. First, it's not always clear what constitutes master data and what doesn't. Second, starting MDM is like trying to run before you've walked for many organizations, since few organizations have taken the first vital steps. Levy said:
Data standards is having metadata and I'm going to agree that everything has a title, I know what the title means and what the definition is. It could vary system to system. Data management says we're going to have these details. Master data management says these details will match everywhere and the value means the same thing everywhere.
Master data management doesn't just take time-it takes attention, patience, persistence and an ability to work across all projects.
"You do it one subject at a time. You don't bite off everything," Levy said.
What's funny is when data warehousing starts, they come up with this data management, but it's only inside the skins of the data warehouse. So you've got all this ETL logic, the transformation logic, it figures out how to link all this data from all these systems so it can then be formatted correctly.
The premise of master data management is, if I've done it correctly, I don't need to come up with that logic because I know how to link product ID across these different systems to pull the product information. I know how to link all the different customer IDs across all these different systems. I don't have to invent that anymore. It's done once and it's done.
Levy will be leading seminars on master data management and data integration issues at the upcoming TDWI World Conference, May 9-14 in Chicago. To read more of my interview with Levy, check out "Understanding Metadata and the Business Problems It Creates" and "The Role of Metadata in Integration and MDM."