The five essential elements of MDM
At the core is the data hub, which is built in one of three styles: a persistent hub, which stores all of the critical data from each source system in one place; a registry-style hub, which stores only the directory information and the foreign keys; and the hybrid, a mixture of both.
In addition, every MDM initiative relies on some form of middleware to synchronize the data going in and out of the hub.
 | ||||
| ||||
| ||||
That, of course, demands a data quality initiative, which could mean using a homegrown tool or buying packaged software.
MDM also depends on external data sources.
"You probably have D&B or something like it in your data warehouse," Power said. "But if you're building a central repository, don't just leave it there."
Finally, and perhaps most important, an MDM initiative demands a data governance program. The business side of the organization is not going to accept an MDM initiative unless they're driving the project, and the way they drive the project is through data governance, Power said.
"Bringing together all these different technology elements and designs is hard enough without a data governance organization," he said. "However you arrange this, think about it from the beginning. I would say don't go back to your business and launch an MDM initiative; launch a data governance initiative."
Ten MDM best practices
- Secure active involvement and executive sponsorship.
"This can only be driven from the top down," Power said. "Bottom-up projects tend to run out of steam and don't work. You may not need to have [executive sponsors] at every meeting, but you do need them in your corner when the going gets tough."
- Business has to own the data, data governance and MDM. It does not work when IT is pushing it.
If the business side doesn't want to own it, have them own the data governance side and IT own the technology component. One technique Power advised is tying compensation or bonuses to the project.
- Be prepared for the skeptics.
"There are anti-champions," Power warned. "These opponents kick up a lot of angst. If you're prepared, then when those champions of the status quo start to throw stones, they won't be able to do any harm."
- Take a holistic approach. What matters is the end result, Power said. Tie financial and time investments to the end result.
"Harley Davidson tied $70 million in increased revenue and decreased cost to MDM," he said. "I don't think [that project] has any problem getting funding. It's intangible and fuzzy enough that if you don't publish it, people don't believe it."
- Be ongoing and repeatable.
"A lot of people still treat this as a project that goes live and is done," Power said. "It's not like that. You're going to be creating and updating data as long as you're in business. From the beginning, plan for a way of life and not a project."
- Dedicate a team of data stewards.
At Tektronix, the data stewards have been involved from the beginning and are closely monitored. Data management is part of their annual review process.
"We have data managers and they own and are responsible for it," said Tektronix's Hughes. "Plus they have me, Mother Oracle. They hear from me sometimes nicely and sometimes not so nicely."
- Create a data model.
- Resist the urge to customize.
"It's getting easier because these hubs are getting more mature," Power said. "If you do [customize], do it carefully. Make sure they're documented and easily upgradeable. Sometimes it's easier to push the vendor to improve further releases."
- Plan for at least one upgrade during the implementation.
"There's nothing worse than contacting support and hearing this problem is fixed in the next release but you can't apply it for months or years," Power said.
- Test.
"[In] real estate, it's all about location -- in software it's about testing," Power said. "Vendors are getting better at testing and QA, but the burden is on you and your testing team."
No comments:
Post a Comment