Companies get into master data management (MDM) for some pretty vastly different reasons.
Some want to conquer the poor data quality that exists throughout the organization and recognize that only by centralizing the process will they ever get a handle on the problems. Some don’t necessarily have egregious data quality issues but are finding that because there are so many different sources for the same data, they need to centralize management of their data models and data origination activities to alleviate the disconnects.
Still others are looking for a way to collect master data for a particularly important operational system implementation and want to separate that task from the main system development work. And some want to change business processes so they can be supported within the workflow environment of an MDM system.
All of those scenarios require the integration of MDM processes and data with operational systems. So, what kind of data is usually managed in an MDM hub in support of operational systems?
Generally, it’s data that is destined for use in multiple systems – what you might call “core” data. For customer data, it’s the name, address, some demographic and psychographic details and other information that is easily correlated on a one-to-one basis to individual customers, with integration but not calculation.
That is basic MDM – mastering data for integration.
On the other hand, some organizations want to run deep business analytics calculations as part of their business intelligence program and see a rich MDM database as the vessel for collecting and consolidating the required information. Customer lifetime value, propensity to buy, likelihood to churn – such metrics can be calculated against mastered analytical data and made available for reporting, quite possibly becoming the catalyst for great returns in the way that customers are managed.
That is also basic MDM – mastering analytical data.
And these two MDM world-views need not be mutually exclusive. The full benefit of MDM comes from combining the operational and analytical MDM approaches. The resulting enterprise MDM data model is robust with both analytics and calculations, and a combination of automated processes and ones born of manual, right-time entry into an MDM workflow environment.
Involving operational systems in the analytical MDM process
An analytical MDM project can be a starting point on the way to operational MDM. But from the analytical viewpoint, it also can be only a brief time before that data is made available and put to use by operational systems themselves.
Some operational systems are going to be more ready than others to absorb analytical MDM data, and their data models and associated processes may need to be expanded. In addition, some operational systems are beating the analytical MDM process to the punch and creating disputes in enterprises as to where analytical MDM data should be originated and distributed.
Customer relationship management (CRM) systems are a case in point. Modern CRM software has a plethora of analytical values built into its data model, most of which predate the MDM era and overlap with out-of-the-box attributes of MDM models. CRM proponents advocate the origination of these analytical values in those systems and may only begrudgingly turn to an analytical MDM system for its distribution capabilities, whereas MDM proponents argue that the MDM hub should be a central origination point for all analytical master data.
Either system is going to need transactional data before it can support analytics calculations. By one definition (mine), analytics refers to a sourcing method that calculates values based on transactional data; whatever you originate by gleaning (i.e., analyzing) transactional data about a customer is analytics. If you append demographics data to a customer record based on an interaction with the customer or input from a third-party system, you can use that data in an analytical fashion -- but it is originated much differently than something you need to calculate from transactional data.
I personally find the MDM argument more compelling in most cases. Regardless, it’s important to understand that analytical data has a significant sourcing component for MDM. Some systems that subscribe to an MDM system’s mastered analytical data are ready-made for analytics and can easily absorb and use those fields, while others (and the personnel who use the systems) will have more difficulty.
So, while the analytical MDM process can be a stepping stone to operational MDM, the reverse is also true. Once an MDM system masters operational data, the information can take on analytic values. Regardless of the order, maximum value comes from fully utilizing MDM’s abilities across both operational and analytical functions.
About the author: William McKnight is president of McKnight Consulting Group, which focuses on data warehousing, master data management and business intelligence. McKnight has written hundreds of articles and white papers and has spoken at conferences worldwide; he also is the author of the book 90 Days to Success in Consulting. He can be contacted at email@example.com.