Traditionally business areas within an organization individually manage data essential for their operation. This data may be incorporated into specialized software applications, MS Excel or MS Access etc., e-mail filing, and hardcopy documents. These applications and data stores support the local business area decision-making and add to its knowledge. There have been problems with this approach. Data, knowledge and decisions are only captured locally within the business area and in many cases this information is not easily identifiable or available for enterprise-wide sharing. Furthermore, individuals within the business areas often keep “shadow files” of data and information. The state of accuracy, completeness, and timeliness of the data contained within these files is often questionable. Information created and managed at a local business level can be lost when a staff member leaves his or her role. This is especially significant given ongoing changes in today’s workforce.
Data must be properly managed and maintained to retain its value within the organization.
The development and execution of “single version of the truth” or master data management requires a partnership between the business areas, records management, legal, and the information technology groups of an organization. Master data management is expected to yield significant gains in staff effectiveness, efficiency, and productivity.
In 2011, Enbridge Pipelines applied the principles of master data management and trusted data digital repositories to a widely used, geographically dispersed small database (less than 10,000 records) that had noted data shortcomings such as incomplete or incorrect data, multiple shadow files, and inconsistent usage throughout the organization of the application that stewards the data.
This paper provides an overview of best practices in developing an authoritative single source of data and Enbridge experience in applying these practices to a real-world example. Challenges of the approach used by Enbridge and lessons learned will be examined and discussed.
TrustNet: Learning from Trusted Data Against (A)symmetric Label Noise
