The Data Governance Board appoints one or more Data Stewards to implement the Data Governance policies. Each master data entity must have a data steward appointed, who is a subject domain expert responsible for maintaining the entity on a continuous basis.
The data stewards typically perform the following functions:
Respond to any alerts (for example, an automated e-mail) generated by the MDM system
Take corrective action on master data entities in order to over-ride the "best guess" made by the automated processes
Monitor and recognize data quality issues with a given entity
Continually improve an entity's data quality by suggesting enhancements to the data quality process
The data steward is typically a business user who possesses enough knowledge of the source data to make the important business decisions that ultimately result in the whole business consuming a given entity in one way as opposed to another. We can go back to earlier in the chapter to see a good example of this, where we had a Customer entity, sourced from the ERP, but the 'Industry Classification' attribute did not exist in the ERP, meaning it had to be added in the master data model. It would be the responsibility of a data steward to interpret the other Customer attributes, and then pick an appropriate Industry Classification. Although picking an Industry Classification is a simple task, the effect of getting it wrong could, for example, mean that any BI reports by Industry Classification would be incorrect.
Due to the important nature of ensuring high quality master data, the data stewardship process often involves workflow. In this situation, the data steward continues to maintain the master data entity, but certain actions may be pushed to another user to approve or reject the decision taken by the data steward.
Even though data stewards are typically business users, they will often have expert knowledge of the source data, as well as how to interpret it in a business context. This is often the case in businesses or departments where the production of reports is not from a central data warehouse, but through an expert user using Excel to arrange and clean the data. Given the data stewards, high level of knowledge, they should be involved in the requirements and design of any data capture processes in an MDM program right from the start.