When Master Data Management (MDM) first appeared there was great hope that it could fulfill on a promise to create the “Golden Record” for organizations. A golden record, defined by most, as the single best view of the entity that it represents. In practical
terms this meant that if you had 5 different records that represented John Smith (patient or contact, etc…) that the MDM could create a single best view out of those 5 records.
In theory this sounds perfect and has been accepted by vendor and customer alike for a couple of decades. But MDM has fallen short of it’s expectations and that has led to less acceptance than what was originally expected.
So the question is why has MDM fallen short of industry expectations?
Much of the reason is based on the fallacy that there is a single golden view that can be achieved. While a single golden view of an entity for an organization sounds like the right answer, in practice this rarely meets expectations and it’s only during implementation that this is recognized.
Let’s break down what it means to generate a golden view also known as single source of truth. For the example with John Smith and the 5 records that comprise the John Smith Entity, there are likely 5 names, 5 phone numbers, 5 addresses, 5 emails. To construct the single source of truth some logic is needed to pick the single best name, phone, address and email. This means that you are asserting that you know the best name, phone number, address, and email to use for everyone else in the organization. One common approach is to select the most recently created or updated value with the belief that the last value to be entered has the highest chances of being the most correct. But this doesn’t take into account the actual consumers of the data which are the departments throughout the organization. Each department can and most often will have different requirements of the data. A sales organization doesn’t necessarily want another departments phone number, for example accounts payable, because they use and rely on their own phone number. The accounting department is likely to have the phone number of customers accounts payable departments which are different than the sales teams CRM phone numbers for their customers.
The other assumption that is often wrong is how one defines the match logic that links record into an entity. There’s usually a general assumption that there’s one way to link records but this is often not the case. An example is the difference between what an operational team might need from an entity versus a business analytical team. A sales team might want a company to roll up into an entity that is defined by a specific address because they execute their sales at each location. But a BI team might want a single view of the company or they might want a regional location. For example, sales might have the same company at 4 different locations within the same city but the BI team might want all 4 to be rolled up into a single entity for their purposes. Both of these are match issues and require different match algorithms to address the different views.
The industry is slowly moving away from Golden View and adopting a multi-view perspective which will help advance the use of MDM. A golden view should be thought of at the department level and multi-view should be the new way to think of MDM on an enterprise level.