MS BI Conference 3: MDM Chalk Talk
The chalk talks are held in the main exhibition hall, "rooms" with about 50 chairs in them, fabric drapes enclosing them, with a small sound system and a large monitor (small widescreen TV size) on a stand at the front.
They are supposed to be more interactive and more demo and whiteboard based than the main breakout sessions are, and that's promising.
Unfortunately, there's usually not enough seats, you can't hear the speaker for the noise from the chalk talks happening on the other side of the fabric, and you can't really see what is being demoed on the screen.
The first chalk talk I went to today was about Master Data Management. I hadn't heard of this term before, so I thot I'd better find out what it is. It turns out it is the same problem we've been dealing with for a long time now, having the same entities in different repositories and trying to have a consistent view of that entity. It's just that now there's a trendy label for it, MDM.
Roger Wolter introduced himself and said that the session would be a few slides and then mostly demos.
Master data management is about having a single view of the "master data" you have in your enterprise, things like Customers, Products, and so on. It usually also encompasses the hierarchy management of that too - all the different ways that the org slices/dices/rolls up/groups the master items.
Master data, if it wasn't already obvious to you, usually is the basis for your dimensions in your OLAP cubes, but maintaining it usually means an ETL process to build the master data in the first place (and dedupe and version and consolidate and make consistent) and then a process to continue to maintain a single view of the entities.
We did this with a customer when I first started with IMaginet, we made a process to keep their Active Directory consistent with their HR database, their accounting database, and some other sources of "people". I can see it being an issue at the furniture manufacturer that is our customer now too - maintaining their product list across two different systems with two different representations of it.
Master data must be surrounded by stewardship and governance. Stewardship means that the data has owners that are responsible for its quality. Governance means that there are rules and processes and quality and service expectiations to ensure the master data stays consistent. There is usually workflow processes around MDM, where the stewards of the data participate in maintaining the quality, particularly in the conflict resolution process.
MDM usually has four P's:
Politics
Policies
Processes
Products
What was most unfortunate about this session is that he ran out of time and he was still on the slides. Almost no demo at all. We spent all the time discussing the issues and what I was hoping to learn was some of the solutions to these things.