Organizational Architecture can Make or Break your Data Governance Program

ABSTRACT: Consider key trends and challenges as you design an effective organizational architecture for data governance while generating value with pervasive analytics.

Introduction

Cloud computing, open-source development, and low-code tools have lowered the barriers to acquire data and analytics capabilities. More users now can access multi-structured data for a variety of purposes. To thrive, enterprises must capitalize and optimize the value of data assets. They need robust data governance – the establishment of a coherent structure for exercising decision-making authority and accountability over how data is managed – as a source of competitive advantage.


Data governance is the establishment of a coherent structure for exercising decision-making authority and accountability over how data is managed.


Data governance is usually run by a core data governance team working with a community of data stewards across business and technology functions. To operate well, they need a sound organizational architecture. This is the systematic arrangement of individual and group roles, responsibilities, inter-relationships, skills, goals, key measurements, and decision rights in managing data assets.


Organizational architecture is the systematic arrangement of individual and group roles, responsibilities, inter-relationships, skills, goals, key measurements, and decision rights in managing data assets.


Unfortunately, enterprises tend to focus prematurely on the other priorities – such as acquiring a data catalog, launching metadata collection, or improving data quality processes – before getting a robust organizational architecture in order. As a result, many data governance programs become sub-optimal, unsustainable, and unable to adapt to business reorganizations, leadership changes, and budget cuts.

Trends That Made Organizational Architecture More Important Than Ever

Conway’s Law reminds us that the structure of a system will mirror the structure of the organization that built it. Hence, a data governance program’s organizational architecture (“governance architecture”) is key to ensure data quality, security, and interoperability across the enterprise. The importance of governance architecture is further augmented by the following trends:

Expansion of governance stakeholders. The first trend is the expanding scope and complexity of data governance stakeholders. In its early days, data governance was primarily focused on improving data quality, documenting lineage and impact analysis, and streamlining reporting development practices. The stakeholders were concentrated in the data warehouse and business intelligence teams.

Today, the scope and demand triggers for data governance have expanded. These include data discoverability and access, secure data sharing via open data forums and marketplaces, data protection and privacy compliance, algorithmic transparency and fairness, as well as ensuring safety and trust in digital and AI-enabled products. Data governance helps prevent and solve crises such as:

  • An urgent migration to exit legacy platform that is facing end-of-life

  • Enterprise application implementation failure due to lack of master data management

  • Inability to access and leverage data assets in the data lake in a timely and secure manner

  • Failure to show up-to-date data map and audit documentation as required by regulations

  • Operational inefficiencies and dysfunctions due to pervasive data quality issues

Figure 1: CxOs stakeholders of data governance with their objectives, risks, and regulations

As data governance stakeholders have broadened to include almost all of the senior leaders and CxOs, the organizational and operating model becomes even more crucial to ensuring effective communication, coordination, and collaboration across the enterprise.

Rise of the citizen data consumer. The second trend is the significant rise in the number of “citizen” data consumers (analysts, scientists, and developers). Powered by low code tools, open-source, and easy-to-access cloud data services, citizen data consumers are empowered to build data pipelines and analytics solutions without having to learn and apply solid data modeling and engineering experiences.

Some enterprises misinterpret what it means to be agile, allowing teams to operate in silos without oversight by IT or a governance body to advocate and enforce standard good practices. The result is inconsistent policy adherence, non-interoperable architecture, and disjointed actions in managing data.

Thus, it is crucial for business and technology teams to establish a clear and practical governance framework, layered on top of an organizational architecture that is not too academic nor complicated. An effective governance architecture must empower stakeholders to collaboratively prioritize collective goals, synchronize processes, integrate tools, and rationalize regulatory requirements without bureaucracy, complex interdependencies, or contentious processes.

Challenges in Data Governance Organizational Architecture Design

Organizational architecture design is inherently complex, requiring a deep understanding of business legislature, corporate culture, and enterprise architecture. For data governance, the task is further complicated by the following considerations.

Lack of individual incentive: Data governance depends mainly on volunteer-based participants. Aside from a small core team, data governance relies on data stewards who are business or technology team members with day jobs. Most of their involvement is ad hoc and grass-roots. Smaller enterprises may not even have a dedicated data governance team or resource. There is no formal job role, career path, reward, or compensation scheme for data stewards. Often the only carrot to motivate a steward is the intrinsic satisfaction that comes from serving and collaborating across teams.

Overreliance on heroics: The most capable data stewards are at risk of burn-out. Since the most capable data stewards possess higher data literacy and analytics maturity, teammates and bosses consider them the go-to resource. They are often pulled into special projects or unplanned incidents to assist in troubleshooting complex issues. Hence, data stewards struggle to exercise their responsibilities, for which there's little career upside for excelling.

Unclear organizational hierarchy and accountability: Data governance often runs as a pseudo matrix community on top of a formal corporate organizational structure. There is no formal reporting line or chain of command for data stewards. Any notion of accountability exists as a matrix of informal dotted lines at best. There is no natural hierarchy for structuring stewardship teams. Instead, nebulous tribal communities form around data domains or enterprise projects. Hence, when reorganization or reprioritization of projects occurs, the data stewardship teams are often significantly disrupted.

Figure 2: Hierarchical vs. Matrix vs. Domain-Based Organizational Structure

Driven by urgencies of the moment: Most data governance initiatives tend to be tactical and reactive rather than strategic and proactive. Enterprises rarely recognize their need for data governance until there is a crisis. By then, their people are critically overstretched, facing an insurmountable volume of issues in an ecosystem full of undocumented band-aid solutions. Planned funding is limited, resulting in underinvestment. As a result, there’s hardly any time and budget for a thorough organizational architecture design.

Developed by IT for IT (i.e., lack of business-inclusive model): Most data governance frameworks and reference models were built by IT vendors or practitioners. Their pervading mental models and terminologies tend to be IT-centric. They are not necessarily incorrect but are difficult to comprehend, and less inclusive to, the business audience. In tech-savvy or digital product-focused organizations, people might be familiar with concepts such as data domain, owner vs. steward vs. custodian, federated vs. centralized, data consumer vs. producer, data mesh vs. fabrics, and data ops. For many parts of the business outside of IT, especially in smaller and less analytically mature enterprises, these are foreign ideas and terminologies.

Figure 3: Business-Centric Data Governance Roles Model

Conclusion

To realize value with pervasive analytics, data governance is a critical source of competitive advantage. Unfortunately, it often fails due to a lack of focus on organizational architecture. A critical analysis of the current trends (i.e., expanding stakeholders and rising citizen data consumers) and unique features of the operating environment reveal several key considerations for designing a successful governance architecture. These include a lack of incentives, reliance on heroics, unclear accountability, fire drills, and lack of business inclusivity. 

In future blogs, I will discuss ways to overcome these challenges and exploit the opportunities they present. I will present sound methods and strategies to design a robust yet agile, multifaceted yet intuitive, and broadly-scoped yet clearly-focused organizational architecture for data governance.

David Hendrawirawan

David helps clients architect data lakes, optimize BI reporting practices, automate data quality and master data management, and engineer cybersecurity, privacy, and responsible AI/ML controls. Having previously been with...

More About David Hendrawirawan