The Doppler Quarterly Winter 2016 - Page 27

uniformity in implementation across systems that may contain duplicate sets of the same data. original data should be weighed in the event future workflows require the original form of the data. Set a lifecycle and stick to it Model Management Setting a lifecycle for data that determines the point in which data is retired and no longer needed ensures stale data is not floating around incurring costs, as well as driving decisions. Predictive models drive many organizations. These models are used to define many things from recom- mendations to risk profiling. These models are just as critical as the data feeding them, if not more so. These models should be considered in a data governance strategy to account for who can approve new model deployment, how they are tested and what documen- tation is required for all models produced. Track metadata across the organization Metadata has become more critical in recent years with the increase in unstructured data being stored and analyzed. The metadata about creation, owners, and topics is key to understanding and increasing the value of a data set. Having an organization-wide pol- icy and single instance for tracking all metadata will enable anyone in the organization to quickly locate information that is relevant to their work. Track copies/instances of the same data set with locations and times of creation As information systems increase in complexity, it is more and more common that a dataset will be copied multiple times within an organization. These replica copies are k