2

So I'm trying to implement a data vault. I have some difficulties modeling the hub. I have 2 systems that don't talk to each other. I want to know how to represent the same item from different sources in the hub.

In our ordering system, it generates a number to distinguish customers. So lets say we have a customer named "Bob" and he needs our service in 5 locations. So we treat those 5 locations as separate customers. So we get 5 different customer IDs(10004-10008). Easy peasy so far in the Data Vault.

Now the tricky part is in our operations system. We have a compound business key. So we get a unique customer name "bob' and a sequence number (1-5) for each of his locations. So (bob,1) is unique but we can also have a (joe,1).

In our operations system, there's no way to put in the ordering system's ID. In our ordering system, the (bob,1) information isn't filled in until well after (bob,1) was already created.

So the timelines can be like this Bob comes in and says I need 5 locations. Ordering system gives him IDs (10004-10008). A week later he gets created in our operations system (bob,1), (bob,2)...(bob,5) Then maybe 2 weeks later we'll see (10004,bob,1), (10005,bob,2) in the ordering system.

How would one deal with this scenario?

Thanks

BarneyL
  • 1,332
  • 8
  • 15
Restie
  • 51
  • 4

2 Answers2

3

Data Vault literature says you store each key in the hub and use a same-as link to do arbitrage. Perhaps you can partially solve this in your ETL by filling the link automatically.

Is there no way to split the business key from his locations? It would seem that you're concatenating different core business concepts into the same key which you'd probably would like to split into several hubs in your DV. (customer, customer location)

Onno
  • 295
  • 3
  • 21
0

I would possibly create two hubs in the raw data vault, aligned with the source systems, and then create a single hub in the business vault where you apply the necessary business rules to build a single hub view of the customer. That way you preserve source-system semantics completely, but still add value to the business by trying to merge the data sets.

As mentioned above, some same-as links could also be used.

Chris KL
  • 4,882
  • 3
  • 27
  • 35
  • Don't create separate hub to represent the same entity (entity, not occurrence). DV is designed to be system independent. Implementing the SAL is a way better solution – MLeblanc Aug 11 '20 at 17:10