Business Intelligence Network - The Vision for BI and Beyond

The Myth of the Purebred Data Steward

Published: February 22, 2007

As organizations formalize the role of data steward, expectations vary—as do perceptions of added value. In this article, Jill Dyche discusses why the jack-of-all-trades approach to data stewardship may result in serving too many masters.

The dog in the accompanying photograph isn’t just any dog. It’s Lulu, and she’s a great dog which, if you’ve ever rescued a mutt, you know is a big relief. Lu does fun tricks, stays close on hikes, and brings the ball back when you throw it. Veterinary consensus is that she is half border collie and half German shorthaired pointer. This makes perfect sense. Lu the pointer is resolute and focused and, when she picks up a scent, will raise her right paw, straighten her tail, and jut out her snout just so. Lu the border collie can read your mind. She is in motion even before she’s been given full instructions. If she had opposing thumbs, we’d be in big trouble. 

Lulu

Lu is not as intense as a purebred pointer, nor does she have the herding instinct of a purebred border collie. Instead, she’s a rare combination, focused and smart and up to any task. Lucky for us, she represents the best of both breeds. 

Business Data Stewards
The role of data steward has been the subject of analysis and speculation from within both IT and lines of business, not to mention HR departments. Many companies have made data steward a formal job function. Nevertheless, debates rage on about whether the role is better positioned within the business or as an IT resource. 

Each assumption has its merits – and its flaws. It’s natural to assume that a good data steward will have solid business chops. Understanding real-life requirements and usage needs is fundamental to rigorous data models, accurate metadata, certified master data, and useful reports. Data stewards who can speak the language of business understand the company’s strategy and the role data plays in fulfilling it. They can articulate the specific use cases and business circumstances that require detailed and integrated data. “We have multiple relationships with our physicians,” a pharmaceutical company data steward explained recently. “They can be MDs, patient activists, clinical trial researchers, speakers, or department chairs. Or they can play all those roles. We need the data to tell us who’s who. That way, we can communicate with customers relevantly, and gain their loyalty. And that’s when our revenues increase.” 

Because of their contextual understanding of how data is used, business-side data stewards are well-equipped to participate in the often politically fraught activities of data governance. They can weigh in on ownership, access, privacy, and regulatory policy making, armed with the real-world knowledge of what’s important to the company. They can also develop practical measurements for monitoring the usability and value of key data elements. Moreover, because they come from the business side, they can usually explain the “whys” – why a data element must be delivered in a certain format, or why credit score is just as important as overall lifetime value on a customer’s profile. Business data stewards are the linchpin for representing the strategic and operational value of information. They have an eye on business benefits, can foster the goal of re-use, and articulate the strategic significance of information to executives.  

But this lofty and hard-to-find combination of skills was nevertheless insufficient for what a manufacturing client really needed. 

Source Data Stewards: The Missing Link? 

We recently worked with a large Midwestern manufacturer who had a business data steward with the previously described skills. She was bright, well-liked by managers both within and outside her line of business (Finance), and had a clear vision of the priorities for data delivery. She made things happen. 

The problem was she oversimplified data requirements. She couldn’t tell you where the data came from. She didn’t understand its lineage, and why that even mattered. She was resolute that the company’s billing system was the system of record for all data, never mind what Marketing, Sales, Operations, or R&D had to say. And, despite having painstakingly drafted detailed use cases for critical information, she couldn’t defend her data against the onslaught of questions from other business units, each of which – no surprise here – had its own credible, business-focused data steward. 

Enter the source data steward, an IT resource who understands the various operational systems of record, lineage, and formatting for heterogeneous data. Source data stewards can deconstruct data requirements to determine the optimal source or sources for data. A competent source data steward understands the operational systems, how the data is stored, its lineage across the data demand chain, and how to extract it from often proprietary systems. An effective source data steward understands the technical and operational aspects of systems. He knows the data “as it lays,” inside and out. 

So we recommended that our client define two discrete data steward roles: business data steward and source data steward. Our sponsor, the V.P. of Applications, initially balked. Why, he wondered, was the source data steward a full-time job? The reason was because the manufacturer had previously tried coaxing operational systems owners to be more data-centric, but got little traction. Similarly, they had played with the concept of data “owners” or “custodians” at the operational level, but simply adding titles and tasks to the hefty responsibilities of already-harried systems staff got little support from IT management. They needed a discrete job responsibility with its own description, tasks, and measures.  

The business data steward couldn’t grasp different interpretations of a data element. Conversely, because of his exposure to different systems across the firm, and his relationships with other system owners, the source data steward could. The business data steward never really understood the operational complexities of synchronizing financial asset data with other data in the company. The source data steward did. The business data steward wasn’t conversant about different business rules for the same data across the company’s many applications, never mind the metadata or the transformation rules. The source data steward filled these voids.  

The source data steward tracked the changes and maintenance activities that impacted data in the operational systems. He was heavily involved in data profiling and correction. Most of the company’s systems contained thousands of data elements – and the source data steward would support the many development teams that relied on access to that data. Once the scope of the source data steward was explained and the benefits clarified, our sponsor set aside headcount for three.

Don't Count on the Hybrid
We got lucky with Lu. She’s got the best of both border collie and pointer. It’s a great mix. But some dog owners aren’t so fortunate. I once met a wolf hybrid whose owner lamented the dog’s psychic conflicts, its instincts as a wild animal and those of a domesticated pet converging into canine schizophrenia.

So it is with the data steward. Our client expected its data steward to be a jack-of-all-trades, both business-savvy and technically sharp, with one foot in systems and the other in strategy. It’s rare to find a person who can explain how a new data element will enrich corporate KPIs who’s also capable of annotating proprietary coding standards. Such a combination is not only rare, it’s unlikely that you’ll be able to find enough of these “hybrids” to adequately fill the needed stewardship functions.

For the manufacturer, the result has been nothing short of tighter business and IT alignment. Business users now understand role and ownership boundaries, task assignments and rules of engagement are more straightforward, data quality is improving, and data is delivered to developers and users more quickly than ever. In effect, the entire data life cycle is represented across both business and IT functions. And that’s a very good mix indeed.

If you found this article helpful and would like to receive this special edition newsletter as well as the latest insights each week from Bill Inmon and other experts featured on the Business Intelligence Network, please subscribe to the Bill Inmon Newsletter.