The need for a Master Data Management (MDM) implementation exists for any company capturing the same data about the same entity, but storing it in different systems. I’ve seen organizations invest in MDM when they only had 2 separate applications with some overlapping master data and other larger organizations having to cope with hundreds of different systems, all recording their customers, and products, and suppliers, etc. in many different ways. Often, when multiple applications are performing the same or similar business functions with the same data, organizations tend to see this as a technology problem. The reality though is that at the core is a business problem because these systems might perform the same business functions, but most definitely they do not all follow the same business rules. Why? Well, the answer might provide enough content for an article on its own, but to cut it short for the purpose of the topic at hand, it is mainly due to a lack of data governance. MDM is definitely a great driver for starting a data governance program or beefing up an existing one, and it cannot be successfully done without data governance. Let’s see what are those 11 data governance deliverables needed for a successful MDM implementation.
Roles and responsibility
The data governance team needs to name those groups, committees, units, or individuals that will be accountable for the data in the “golden record”. The individuals could very well be assigned data stewards that will maintain the mastered data and validate, and even merge and purge records when the situation requires a manual process. Data governance will also clarify who is responsible for enforcing the business rules and policies governing the “golden record” (more on that below).
If not already in place, a business glossary is one of the most important deliverables in order to get the MDM underway. There will be a lot of business terminology to be encountered from the start of the MDM implementation and one cannot manage something they cannot understand. The organization needs to reach consensus on the terminology of the data being mastered, including definitions for customer, product, vendor, and so on. These might seem like straight forward, but their complexity can be directly proportional to the complexity of the organization and its diversity of services. The more services the organization offers, the more industries it is part of, the more departments it has, the more it would take to understand all the business requirements around their “customer” and what that business term actually means at the organization level and all the way to department level.
Checkout these 6 main benefits of investing in a business glossary
Business rules and policies
In addition to the business terminologies mentioned above, data governance needs to work with the business to document all business rules and policies required to manage the “golden record”. These are rules and policies that govern the creation, acquisition, maintenance, dissemination, and archival of master data and all data feeding into it. Similarly to the challenges faced by creating a common verbiage throughout the organization, putting these together is not easy. The business rules and policies need to address external requirements from regulatory bodies while meeting the needs of different business units.
As part of any data integration project, and yes the MDM is much more than that, the technical team needs to refer to a data dictionary to achieve better results. The data dictionaries need to capture technical metadata of the data that will be mastered (data type, size, format, etc.). I’m mentioning data dictionaries and not data dictionary, because there should be a dictionary for each system/ database that feeds data into the “golden record”. The data governance team needs to work with IT to also have a data dictionary documenting the technical metadata for the “golden record” itself.
Data integration rules
Since MDM is helping consolidating the same data stored in multiple locations, often in many different ways, there needs to be rules on how to do this consolidation and integration of data. For example, the data from one system may take precedence over the same data from another system. The rules on how this priority is determined are put together by the data governance team. Same with deciding what data to merge or purge or transform and what those data integration and consolidation rules are. On the same line, the data governance team needs to put together and disseminate to all data consumers and creators the documentation of the data lineage.
Data quality rules and metrics
The definitions for data quality dimensions is again a deliverable of data governance, but also different rules for:
- determining the level of quality of the golden record data
- performing data profiling to determine the data quality level
- implementing and running data quality audits and preventive measures
Metrics and KPIs
Adding to the above, data governance should also output and maintain a set of KPIs and data quality metrics, data issues metrics, and data governance metrics as they relate to the MDM implementation. Overall metrics on the MDM implementation are recommended in order to keep track of the progress. These could include:
- Number of integrated systems
- Number of master data records
- Percentage of records processed
- Number of systems consuming master data or feeding into
- Number of data domains covered
- Time needed for a new “golden record” setup
Regardless on what metrics and KPIs you are tracking, the data governance team should make them available and report on their progression on a regular basis.
Note: If you are interested in a lot more examples for these metrics, but from the point of view of data quality or data governance, let me know and I’ll be happy to create some content on it.
Master Data Management (MDM) and Data Governance are great partners. Even though the MDM can be a great driver for data governance (and there are a few others – read/watch more here), a successful MDM implementation cannot be achieved without data governance. If you don’t have a data governance program, what are you waiting for?