We often talk about a “golden version of the truth” when it comes to master data management’s goals. But in a recent webinar, two experts suggested a best practice is to focus instead on creating the best version of the truth.
What the heck does that mean?
It means that those following best practices will adopt both styles of MDM: a registry that retains that so-called golden record and the more flexible, virtual style that offers the best record on a more ad hoc basis.
The physical master record gives you performance and scalability, explained Aaron Zornes in “Top 10 Best Practices for Successful MDM Implementations.” The other gives you high-speed availability, explained Zornes, who is the founder and chief research officer at the MDM Institute.
It’s called “best version of the truth” to reflect the fact you are often picking the best values from individual cells across duplicate records, explained fellow presenter and Informatica representative Ravi Shankar. Shankar is responsible for product and technical marketing activities for master data management (MDM) as part of the data quality business unit at Informatica.
Shankar shared that one company he worked with took six weeks to identify its top 400 business customers, but after adopting both styles of MDM, they were able to identify those top 400 customers instantly.
Another best practice: Ensure you support reference data with your MDM initiative.
Reference data includes things like lookup tables, metrics and measurements, country codes, medical codes and other types of relatively static data. Since it is static, you may wonder why you’d want to support it with MDM, but Zornes pointed out master data often “bumps against it” when you’re sharing data during e-commerce, with trading partners, business partners or suppliers.
“Reference data is something that’s been raising it’s head the last 12-18 months as a standalone topic, and in reality, it’s actually a subset of master data management,” Zornes said. “The software vendors have realized this as well and so they’ve been busy promoting the notion reference data management and how well their product, their MDM hub in particular, supports the various types of reference data.”
In short, it lets information flow between applications, ensuring they can speak a common language, he added.
For instance, right now many health care applications still use ICD-9 codes, even though the government has switched to ICD-10 codes. You need something to help manage that “translation.”
The best practices are culled from the MDM Institute’s research and work with an advisory council composed of 150 large companies. You can hear the full hour-long webinar on Dataversity or just read the slides for a quick explanation of the 10 best practices.
After you’ve learned the best practices, you might want to take a look at what John Owens calls the “Seven Fatal Errors in MDM.”
This list focuses more on MDM implementation issues, but it’s a very unique list. For example, his first fatal error is that you’re even calling it MDM rather than master entity management.
Those types of statements can be a bit of a turn-off because you immediately feel like you’re entering a semantics disagreement — but there is a bit more to his point than mere word choice.
He contends calling it master data sends you immediately to the wrong place — your own data.
“Unless Master Entity Management is already well established in your enterprise, then your existing data on your Master Entities is probably as far from what it ought to be as it is possible to get,” he writes. “So, looking there for your Master Entities makes about as much sense as looking in a scrap yard for the components for a Formula 1 racing car.”
The man has a point; although, he does seem to be focused on entities like customers, employees, etc., and not, say, product data.
He also talks about how silo applications — from CRM to HR to Supply Chain Management — can create a silo mentality that thwarts MDM.
But for Owens, the biggest error — the original sin from which all other errors flow — is not using the logical data model.
Like I said, it’s a bit of a down-in-the-trenches discussion, but if you’re a CIO or knee-deep in MDM, it’s worth your time because he also offers a solution to each of these seven “fatal errors.”
