Several master data management (MDM) and data governance practitioners took to the stage at the TDWI Solution Summit in Savannah, Ga. last week to share their experiences and MDM best practices.
The panel speakers -- who represented well-known companies, including financial services giant Sallie Mae, software company National Instruments and Harvard Pilgrim Health Care -- also answered audience questions about related data management topics ranging from logical data modeling to data warehousing.
A definition of master data management (MDM)
The practice of MDM combines technology, data governance and business processes in an effort to ensure the accuracy and consistency of master data across business units. Master data might include information about important business entities, such as customers, suppliers and products. Industry observers say the benefits of a well-executed MDM program include greater visibility into the overall health of the business and improved customer experiences.
Proper documentation can avert trouble
Conference speakers stressed the importance of documenting the lineage of data and being consistent about creating audit trails. It’s a point that Philip Russom, TDWI’s research director for data management, learned well “a few years ago” when he was working for a company that was investigated by the federal government. The company's lack of an audit trail eventually cost it millions of dollars.
While he didn’t give the business’ name, Russom said it was “kind of a small company” with about $12 million in annual revenue that had bought four competitors over two years.
“Based on that information, the Federal Trade Commission decided that we were a monopoly,” Russom said. “But it’s kind of hard to think of a $12 million company as a monopoly.”
During the course of the investigation, representatives from the Securities and Exchange Commission showed up to view the company reports. One of the first things the investigators wanted to know was where the data that was used to compile the reports came from. Russom’s team had no idea.
“Because we didn’t have proper metadata, we didn’t have proper master data, we didn’t have a data lineage audit trail, [it took] three years for the FTC to conclude their audit,” Russom said. “In the end, the report said there was no monopoly.”
For Russom, the lesson was clear.
“For anything that is going to go public or anything that is going to be given as a financial statement -- you better have airtight documentation and an audit trail,” Russom said, “and master data management is one of the pieces that will help you document the data.”
MDM best practices: Pay attention to data governance
Barbara Deemer, the chief data steward at Sallie Mae, stressed the importance of the data governance efforts that go into maintaining an MDM program.
In her company’s experience, data governance supports MDM by giving organizations a formal way to communicate, collaborate and establish the metrics by which progress will be measured. It also allows organizations to formalize the practice of bringing people together and prioritizing projects.
“There is no question at Sallie Mae [that] governance is a key to ensuring that MDM and data quality happens,” Deemer said. “The formalization of data governance allows you to establish the communication change that you will need and the metrics that you possibly need to get your MDM and data quality programs off the ground.”
For more on MDM best practices
Get more of Sallie Mae’s tips for launching a data governance program
Read about more MDM best practices
Learn about the link between MDM and business process management (BPM)
Making the case for data modeling
Organizations that embrace the process of data modeling can expect to be more successful in their MDM initiatives, according to the conference speakers.
One audience member asked the speakers for advice on how to handle pushback when trying to encourage her midsize organization to create logical data models for data warehousing and MDM initiatives.
John O. Biderman, an information architect with Harvard Pilgrim Health Care, said getting users
to be more helpful when creating data models is largely a matter of properly making the case and
exercising strong leadership skills.
“I just have a view that you cannot do enterprise information anything without an enterprise data model of some sort -- at least a conceptual model,” Biderman said.
When he first began working for Harvard Pilgrim Healthcare, Biderman found that the organization already had a well-thought-out enterprise logical data model and had also made great progress getting different business units to agree on the meanings of specific terms.
“It’s an incredibly valuable asset when you have the same business concept represented in many business systems,” he said. “It’s incredibly valuable in MDM and [services-oriented architecture] and I think it’s sort of a vital part of data management.”
Christine McClary, a data manager with National Instruments, added that the process of creating conceptual data models has helped her organization with its “customer insights initiative,” a data integration project aimed at giving the company a 360-degree view of its customers.
“The whole process of conceptual data modeling and discussing the vocabulary and the relationships has been very fruitful for us,” she said.