TDWI | Training & Research | Business Intelligence, Analytics, Big Data, Data Warehousing

Five Steps to Balancing Centralized and Decentralized Data Governance

Organizations must realign responsibilities and decision rights between the central data governance authority and the decentralized business functions. Here’s how.

• By Diby Malakar
• June 12, 2023

Data governance is critical in today’s data-driven world. It defines how data should be gathered and used within an organization.

For Further Reading: Three Ingredients of Innovative Data Governance Executive Q&A: Data Governance and Compliance Setting Up for Success: Governing Self-Service BI

Growing data sprawl causes users to act on whatever data is available: decisions are made in isolation and data is improperly used. Additionally, the rise of the data mesh forces a fundamental shift in how companies manage their data governance processes, which means it’s no longer the responsibility of a core data and analytics function to create data products that are then delegated to business functions. Organizations must realign responsibilities and decision rights between the central data governance authority and the decentralized business functions.

To strike this balance, centralized data governance teams and representatives from business groups must take five key steps.

Step 1: Revamp data governance policies

Although most centralized data governance teams have long-standing, well-understood policies, they can seem arcane to a newcomer. Organizations' challenge is people coming in and out of roles that intersect with governance. Once some of the day-to-day responsibility of data governance is decentralized, data and metadata governance policies should be carefully examined by both data leaders and users -- because they’re the closest to the data -- and reworked with the broad community in mind. This means that policies must be more prescriptive, sufficiently clear, and specific if they are to be used properly.

Organizations also need to create an ongoing orientation or training plan to keep the organization up to speed. You will have higher quality, more accurate data, and better visibility into what data to use and how to use it, resulting in better decision-making.

Step 2: Decentralize decision rights and responsibilities

A centralized data governance team must switch from implementation responsibility (centralization) to facilitation and oversight (decentralization). The result: a maturity curve in which data experts must redefine their roles. The data governance team needs to work through this challenge to define decisions, rights, and levels of autonomy to understand the split between centralized and decentralized data.

There is a delicate balance between centralized and decentralized governance, and the maturity curve of the customer plays a critical role in how implementation pans out. Rights and responsibilities need to be clearly defined. To help business functions understand this shift in responsibilities, a decision rights and responsibilities matrix shows who is responsible and accountable for data governance. Thereafter, the reconstituted central team can concentrate more on helping business functions recruit, train, and support decentralized governance personnel.

To illustrate this, take the example of the U.S. Department of Agriculture (USDA). The USDA sets the policies and standards for food production, labeling, handling, expiration, and disposal. The USDA defines governance policies and focuses on communication, education, and enforcement. The farmers, ranchers, and grocers -- executors of the guidelines -- help with the execution. They work well together to maintain quality and consistency instead of working in silos, which allows for faster production and higher-quality products than in the past.

Step 3: Gain transparency and clarity through service-level agreements

When continuously creating data products, an enterprise must provide transparency, reliability, and accountability that consumers can trust -- such as a service-level agreement (SLA) associated with a data product. An SLA should be required throughout the governance process – which includes a detailed workflow step (for a business term) to govern change, ensure the data catalog is reviewed and trustworthy, and have rich context (such as producer, frequency of data updates, data quality commitments, and critical use policies and restrictions).

For example, stewards are responsible for maintaining a data catalog that contains high-quality, governed content. To do so, they must monitor and measure statuses against policies, standards, and defined targets. Business stewards guide community curation participation; technical and compliance stewards largely maintain the attributes they are responsible for and respond to community expectations and requests.

Step 4: Implement and maintain a governed catalog as a centralized system of reference

A significant risk of decentralizing data governance responsibilities is losing enterprise transparency and reducing the visibility required to satisfy regulatory authorities. Organizations can implement a centralized data catalog used by all decentralized business functions to address this challenge. This does not limit the business functions’ autonomy or responsibility.

A governed catalog is a centralized repository of data assets -- such as databases, tables, files, and data pipelines -- that have been identified, classified, and managed by data stewards according to a set of policies, standards, and best practices. Essentially, it’s governed data that is stored in a catalog and is a business enabler and increases productivity by providing a path to trusted data and better relationships.

However, consistent maintenance of the governed catalog is crucial. Inaction poses a risk. This could happen in any part of the organization that deals with or produces data. When data leaders fail to ensure that data is governed effectively, data ends up in siloes instead of the catalog, hindering its ability to be used effectively. Transparency around how data is governed is critical. More specifically, following the rule of thumb that activity is decentralized, but knowledge needs to be centralized.

Step 5: Perform governance audits

Previously, the data governance team handled all responsibilities. The shift toward decentralization holds all teams accountable for correctly applying governance. The challenge is to teach those teams the responsibility for overseeing the data. This is the company's next maturity level and will mitigate risk.

The shift to decentralized responsibility also means that central governance teams move into the role of “auditor,” monitoring the work of other teams and holding other business operators accountable. With an audit process, alignment with policies, SLAs, and the use of the catalog as described above is of utmost importance. If new governance regulations are enacted, the data governance team will oversee and enact changes to the policy to ensure compliance and train others within the organization on new processes.

If your company struggles to manage unwieldy data, now is the time to audit structuring data governance policies. This change is significant, and assessing readiness and maturity is imperative before taking on a decentralized approach.