We’ve added another new session to our upcoming WWDVC conference! We want to welcome STEPHANE VIVIEN to our conference. Here’s a sneak peek at what they are bringing:
How to manage BK changes?
The DV bible (Building a scalable Datawarehouse with DV2.0) states that a BK, by definition, should have a very low propensity to change.
But there are cases when a BK is changing along the process; for example the contact acquisition from a marketing/CRM perspective.
Let’s imagine our client is a famous Business School that has to follow the changes (salary, work country…) before/after the student have followed an MBA.
To make it simple, let’s say that the Contact object will exists in only 3 different source systems: marketing, CRM and a campus solution.
We have defined that the BK is composite with fullname, DOB, email address.
Not all the 3 of them will be fed at first client sight (in the MKT system), probably only email address and maybe fullname with luck.
At the second sight, we could have the DOB information but…
… in the DV system, we will have 1st row with the first BK/HK (without DOB) and then a second row with the DOB
=> how do you manage to consolidate these 2 keys on a unique one?
Using the source system ID’s in addition…
We create additional helper tables (one per source system) which will contain the BK, the HK and the contact source system ID
In our case, the HLP table will also contain two rows (still because 2 BK/HK) but related to the same contact ID
So, using a Same-As Link we can determine that the first BK represents the same business object occurrence than the second one; we always refer to the oldest BK (the first one that existed considering all the source systems) using hub dates and satellites business dates.
So, at a business vault layer, we can build a “gold” contact hub with deduplicates keys including indirect relationships (BK1 => BK2 => BK3; BK1 => BK3) in case we have more than 2 rows for the same BK.
Then, defining a gold contact satellite, we can move on to a Data Governance level as the Contact object can be the feeder of a first Contact implementation in an MDM system for example and it allows to have a full consolidated view of the operational data propagation.
You WON’T Want to miss this years’ conference. If you haven’t registered yet, please do so today – just click the link below.
See you at #WWDVC – the best and ONLY authoritative #datavault conference in the world.