Perhaps the two most important decisions you can make concerning data stewards are how to select them, and how to manage them.
Some organizations select data stewards based on the subject area of the business data element (BDE) involved, while others assign the role based on functional area. In either case, the approach can create awkward overlaps, redundancies and gaps, resulting in a lot of needless confusion; what’s more, one can never make the boundaries distinct enough for the individual data steward to be successful. More importantly, neither approach will help the organization find the individuals who actually know the most about a given BDE, or feel the greatest sense of investment in helping to improve its quality.
Instead, we believe that the best way to assign data stewards is to start by building an inventory of data elements — a record of all the hundreds of distinct BDEs as they occur throughout your organization’s information systems. The next step is to implement a process for determining how the data is created, used, and transformed as it moves through the organization (in our experience, facilitated face-to-face meetings work best). Last but not least, you must enrich the data elements and address any anomalies. In the course of this process, the individuals who are best prepared to serve as data stewards for each data element emerge naturally.
This approach requires the organization to methodically examine its data landscape and identify the experts for each given area of data through a rigorous collaboration process — with the goal of extracting knowledge from employees’ minds that contributes to understanding the organization’s data. Typically, organizations want to take a shortcut — assigning the work described above to a group of “named” data stewards and “holding them accountable” to deliver. This approach is ineffective from the outset, as the data stewards will almost certainly not know what to do in order to succeed.
Managing for Effective Data Stewardship
The other part of the equation is how you manage your data stewards. Again, organizations vary widely in how they attempt to do this, but our experience has shown that the most typical approaches are ineffective in helping the organization achieve the desired impact on data quality, and likewise frustrating for the individuals involved.
We have found that the best way to manage data stewards is to establish detailed data quality goals for the specific BDE for which each is accountable. The most meaningful metrics to track are Validity, Conformance, Completeness, Integrity, Accuracy, Timeliness, and Availability (see definitions below).
Seven Metrics of Data Quality
Validity — adherence of data values to an existing set of approved values
Conformance — structure of data values with respect to some known standard, expected pattern or shape
Completeness — whether all expected values are present
Integrity — whether data values maintain their identity as they move through or across various tables or systems
Accuracy — whether data values are what they are supposed to be, or what is expected from the data consumer’s perspective
Timeliness — whether data values are received when they need to be by the primary consumer groups associated
Availability — the sentiment of the data community with respect to whether data is easily accessible for the specific types of consumption that they execute
Overall Data Quality Score — the overall health and quality for a given BDE, derived by the aggregation of the seven data quality indicators mentioned above, for all of the BDE’s physical versions
Ideally, an organization should set realistic goals around each of these metrics and then aggregate them for each data steward on a weekly, monthly or quarterly basis. In addition, an overall metaquality score should be calculated that measures how completely the data steward has enriched or supplied metadata to that BDE since the last evaluation. With these two scores, one can derive an overall Data Quality Score, the one metric to which the individual’s bonus should be tied.
This approach allows all parties in the arrangement to know exactly what their goals are, and to make adjustments as needed if results are not where they should be. By getting the right people in the role to begin with, and incentivizing them on clearly-described metrics that are within their control, the organization can get the data quality outcomes it seeks.
Seeing Data in All its Dimensions
The real value of investing in data quality assessment and improvement is that it gives your organization the insights and perspective needed to optimize the value it derives from its data.
It’s one thing to know where each data element resides and how it’s labeled. But it’s an entirely different level of insight to understand all its unique instances, discrepancies in field names, and nuances in how values are presented in it. That’s the level of insight that is so critical to truly improving the quality of data, and the value it delivers to your organization.