Over the last few years, data governance has become a key watch word for any organization using data to drive its business. Not only have public and regulatory pressures put the onus on corporations to ensure data quality and privacy, but businesses themselves have realized the internal efficiency benefits of carefully organized and controlled data assets.

Governance from above. Traditional models of data governance operate from the top down. Chief Data Officers (CDOs) and governance councils work with data owners to create policies and procedures, then data stewards implement those policies at the unit level. (See figure 1.)

Figure 1. Top-Down Data Governance

Within this model, data stewards act as guardians. They work to enforce data definitions set by those at the top and attempt to control their data users’ access to the data. This model aims to reduce risk, but in the process, a disconnect forms between the analysts and data scientists who work with the data and the bodies which govern the data. Misalignments emerge, and the definitions created in abstract at the top don’t correspond to all the use cases identified by practitioners on the ground.

Governance from below. In the last few years, a new philosophy for data governance has gained steam in response to these shortcomings. Often referred to as agile data governance, this new methodology takes a bottom-up approach. Rather than starting with company-wide standards imposed from above, in the new paradigm, knowledge about data held by hands-on data users percolates up. (See figure 2.)

Figure 2. Bottom-Up (Agile) Data Governance

As shown above, data users pass their knowledge up the food chain. Governance bodies integrate these insights and adapt them into broader policies that they pass back down in a dynamic exchange. Because the governance process involves the data users themselves, the end result better reflects their needs and the reality of the data. 

As data volumes continue to grow, this model has proven far more scalable. A core data governance council and team of part-time data stewards could never sort through and define all of the data assets generated by a modern, data-driven enterprise. But a bottom-up approach embeds the task of data governance into the regular workflow of analysts and data scientists. This devolving of responsibilities also makes data governance more agile. As business needs change and demands on the data evolve, definitions and standards can shift accordingly.

Tools like data catalogs and data hubs facilitate this new model of data governance. Catalogs not only automatically gather meta-data on data assets, they also provide features like tagging, rating, and commenting that allow data users to provide insights directly. The social functions of some catalogs facilitate the easy exchange and codification of tribal knowledge across departments. This helps users from different lines of business learn from one another directly. Data hubs similarly democratize governance by lowering the bar for business managers to master data and merge records.

The new data steward. Within both models, the data steward serves as a vital link in the governance chain, but as the orientation moves from top-down to bottom-up, the stance of the steward must also shift. Rather than policing the data user, stewards must focus on empowering them. Instead of trying to enforce abstract standards, the new goal of the steward is to harmonize user-generated definitions and communicate the best practices of their team to the larger organization.

Bottom-up governance initiatives work best when implemented on a specific project or within a single team and then slowly expanded across the organization. The steward, as a businessperson with deep insights at the unit level, sits in the perfect position to spearhead this approach. Stewards also help ensure the agility of the system functions. Without the right guard rails from a data steward, empowered data users could easily create redundancies and silos, and a well-intended governance initiative could descend into chaos. By continuing to set access permissions, stewards ensure users have the freedom to innovate without accidently seeing confidential information. Ultimately, stewards must strike the appropriate balance between a governance free-for-all and the grander aims of the organization as a whole.

Conclusion. Data governance continues to become more important as companies rely on greater volumes of data to run their businesses. Good data governance not only mitigates security concerns, it also reduces the time needed to build analytics solutions by making it easier for users to find the right data. 

The role of the data steward is critical in making the switch from top-down governance to an agile, bottom-up approach. As the intermediaries between departmental data users and corporate governance councils, data stewards can ensure the appropriate balance between shared standards and agility. These stewards, however, must reshape their outlook to achieve success with bottom-up governance, focusing more on empowering than policing.