I recently attended Teradata Partners conference, which was a lot of fun, one of the things they discussed was governance, data stewardship, data ownership - and of course: Claudia Imhoff in her masterful presentation of MDM talked diligently about SoR, SoE, and a few other acronyms. The gist of the statements (across the board) was that System Of Record lines are blurring. Shifting Sands I might say...
In 2006, I blogged on my version of the SoR, and how I believe there are at least three different definitions for it. You can find the entry here. I recently received a good comment about SoE, and how these things need to be separated. The comment discussed the notion of incorporation of MDM.
I'd like to keep this blog entry unusually short (for me) - because I believe a summary is in order.
My current thoughts are shifting along with the sands of definition land... but here's my two cents on it:
1) We have systems of entry (SoE) as they are calling it
2) We have systems of Record (SoR) - but how this is defined is very shady
3) We have single versions of integrated FACTs (everyone knows how I feel about TRUTH)
4) We have historical views of integrated FACTs
5) We have non-auditable, completely altered, data in "Single versions of the Truth" that are delivered to the business.
I've assigned SoR to an integrated EDW space, single version of integrated facts, because it's the only place that 3 & 4 exist over a period of time, I've assigned SoR to an SoE - why? because frequently the operational systems do both, and are responsible for both, and once the data is fed from #5 BACK to source systems as "clean data" - that shifts it's definition to become an SoR as well.
Now we have MDM - which really, only the Master Data itself can be considered an SoR for the company, but what does that mean?
So I'll leave you with this tonight, these are questions that over the next couple months I'll blog on in depth...
1) What _exactly_ is an SoR?
2) Who is the "expert" that can accurately define a global metadata definition for the term SoR?
3) What differentiates an SoE from an SoR and why would they be separated?
4) Why ISN'T a raw-detailed warehouse (single version of integrated facts) considered an SoR? When might it be?
5) Can a Federated Star Schema, cleansed data set/altered data set be called an SoR when it feeds data back to the operational systems? Is it acting as Master Data?
6) WHEN does a Master Data "set" of information become qualified to be an SoR?
7) Where's the governance, data stewardship, data management in all of this?
Love to hear your thoughts....
Posted October 11, 2007 7:44 PM
Permalink | 1 Comment |