Datavault - hard rules (rawvault) vs soft rules (businessvault)

Question

I have a question on hard rules (rawvault) and soft rules (businessrules).

The example I have is a source system has a denormalized table called Pets where Pets contain Cats, Dogs, and Birds where they are distinguished by a Type Code (1 – cat, 2 – dog, 3 – birds).

My question is regarding hard rules vs soft rules when loading the data into the Rawvault vs business vault. When loading the Pets table, can you create a h_cat, h_dog, and h_bird hub in the rawvault and filter the source table pets base on the Type Code of 1 into h_cat, type code 2 into h_dog and type code 3 into h_bird? Is this a hard rule or a soft rule?

Or

Should we be creating a h_pet hub in the rawvault keeping the data as close to the source as possible creating the h_cat, h_dog, and h_bird in the businessvault when filter the data based on type code because this would be classified as a soft rule?

Answer 1

I think I'd go with the second approach to stay as close as possible to the source data structure.

Then split the tables in the business vault if needed. This also doesn't need to comply to the Data Vault model, I suppose. In the end the main question is how you want to use the data. If for reporting then I suppose a virtual cube would work as well.

Answer 2

In your case you would have a single hub for all three animal types. Any logic that changes the data is added after the raw vault.

Here's a good distinction between soft rules and hard rules:

Hard Rules

These should be applied before data is stored in the DataVault. Any rules applied here do not alter the contents or the granularity of the data, and maintains auditability.

• Data typing
• Normalization / Denormalization
• Adding system fields (tags)
• De-duplication
• Splitting by record structure
• Trimming spaces from character strings

Soft Rules

Rules that change, or interpret the data, for example adds business logic. This changes the granularity of the data.

• Concatenating name fields
• Standardizing addresses
• Computing monthly sales
• Coalescing
• Consolidation