MDM and Data Vault EDWH: master data, business keys, and a complementary approach

There are similarities between the approach to implementing an MDM system and the approach to implementing an EDWH with Data Vault, especially speaking about the business key concept.

This for a very simple reason: these 2 subjects are complementary!

Why MDM and Data Vault EDWH naturally complement each other

Indeed, a Data Vault designed EDWH will provide centralized and common definitions (here again our concept of business key 😉) for all business objects of the company, including those that are not intended to be managed in an MDM solution (the “non master” data – we will come back to this a little later)

In addition, a Data Vault designed EDWH will provide all the required data history to set up an MDM system – or at least to set up a new “master” data in the MDM system – through its Business Vault layer in which the business object will have been qualified, cleansed, deduplicated… in one word, identified as the unique reference of the business occurrence – OK there is more than one word 😉

Master data vs non-master data: what’s the difference?

But this leads us to ask ourselves: what is a “master” data and what is a “non-master” data?

What is master data?

A “master” data will be qualified as such if it intervenes in most of the company’s business processes, if it is identified as critical within the company.

Examples: a customer, a product (or service) for sale, a supplier, an intermediary, the organizational structure of the company…

What is non-master data?

At the opposite, a “non-master” data will be a business object that is not critical in the business process, or specific to a particular process (not shared), but which nevertheless exists and therefore needs to be designed in the EDWH.

Feeding a Data Vault EDWH once MDM is implemented

To feed Data Vault designed EDWH once an MDM system is implemented (I’m only talking about the “consolidated”, “co-exists”, “centralized” approaches), the business objects (HUBs in Data Vault language) related to the “master” data can only be supplied by the MDM system when the “non master” data must be supplied by the other operational source systems.

Which should come first: MDM or EDWH?

In summary, as much as it is possible to set up its MDM system before setting up its EDWH, as much as it is entirely conceivable to start an MDM initiative with the implementation of an EDWH (thus allowing a global retreat on the “master” data) before entering the MDM project as such but having capitalized on data quality and merging issues.