Kimball University: Pick the Right Approach to MDM

It's time to migrate master data management upstream to an integration hub or, ideally, an enterprise MDM system. And if you have yet to do anything about data consistency, take these four steps toward integration and stewardship.

Four Steps Toward MDM

Regardless of how far you go towards creating an Enterprise MDM, there are several common steps you need to take:

Set up a data governance function. First, create a permanent data governance function that includes, or is managed by, the business side. This group must drive agreement on names, definitions, systems of record and business rules for every entity and attribute in question. Create an organizational process for ongoing review and enforcement to ensure correct usage and appropriate sharing of data.

Integrate existing data. You must take a first pass at integrating data from the various sources in the organization. This may be fairly simple for some attributes, like Product Category, once the business rules are defined. However, it is a massive effort if you need to match and deduplicate more complex entities such as Customer. You will never be 100 percent right. Remember that the resulting integrated data must be maintained. If this is as far as you get, consider yourself above average.

Work toward the ideal. Rewriting existing systems and reworking business processes to use the master data system rather than local data sets is a massive undertaking. The organization must be committed to correcting the problem at the source. This effort involves data governance resources, development resources, purchased software and an overall redesign of the core systems and the development process.

Enlist others. If you are getting no help from the larger organization, remember that the results of integration have value to downstream systems such as CRM and Customer Service. Try to enlist them in the fight for creating an integration hub or enterprise MDM system. All organizations should have a data stewardship function to define shared data elements and monitor their usage. You should also be practice MDM in the form of creating conformed dimensions in the data warehouse ETL process. Beyond this, the organization should be working on moving the MDM function upstream into the operational systems, initially as an integration hub, and ultimately as an enterprise master data management system.

Warren Thornthwaite is a member of the Kimball Group, teaches at Kimball University and is coauthor of The Microsoft Data Warehouse Toolkit (Wiley, 2006) and The Data Warehouse Lifecycle Toolkit (Wiley, 1998). Write to him at [email protected].