In the realm of “cart-and-horse,” “chicken-and-egg” analogies, it’s hard to argue that anything goes first or comes before data quality.
Some business decision-makers who suddenly or finally “get” the gospel of master data management (MDM) want to plunge right in with the data modeling and integration tasks to get the initiative going. But ensuring the quality of the data going into the system first is essential, as is keeping the data clean through ongoing data governance.
Philip Russom of The Data Warehousing Institute puts it very succinctly:
To reduce the risk and increase the prospects for success when implementing a master data management (MDM) project, keep this tip in mind: Implement data governance (DG) first, master data management second.
That said, it doesn’t do your enterprise any good to have isolated pots of pristine data sitting in stovepiped applications that end-users can’t easily access. Once the data have been cleansed, it’s ready to be managed in an MDM platform. Only then can data quality tools and initiatives truly benefit the organization.
Here’s a way to visualize it. On the left in the diagram below are sources of unclean data being aggregated in an MDM repository, making it easy for personnel and applications throughout the enterprise to work with data that’s incomplete, inaccurate, and inconsistent. MDM analyst Dan Power calls this “a fast, automated way to shoot yourself in the foot”:
On the other hand, it doesn’t do much good to have clean data that few users or applications have convenient access to. That’s what happens when you instill data quality tools and techniques without taking steps to aggregate and manage the data in a larger MDM initiative:
Chicken or egg, cart or horse…whichever your favorite analogy is, data quality and governance logically come before any enterprise information initiative – master data management included. A good summation of this point is a quote from David Loshin, president of Knowledge Integrity:
In order to benefit from MDM, there must be a solid foundation of data quality control, semantics for data sharing, and data governance…MDM as a tool is never going to be effective in the absence of data governance and data management. Governance policies and processes for master data can improve data utility and make sharing more effective.