Empowering data products with comprehensive governance and stewardship

Executive summary

Data products are reshaping how businesses operate, but their success depends on effective stewardship. Without it, organizations face risks such as poor data quality and compliance failures.

Key highlights:

• The opportunity: Data products, such as dashboards and predictive models, are the backbone of modern decision-making, enabling innovation and efficiency.

• The risks: Poor data governance can lead to degraded data quality, compliance risks, and inefficient use of resources. The cost of poor data quality alone is estimated at $12.9 million annually per organization (Sakpal, 2021).

• The solution: Embedding stewardship at every stage of the data product lifecycle ensures data products remain reliable, relevant, and strategically aligned.

• The benefits of effective stewardship:

  Enhanced data quality: Reliable, accurate data products aligned with organizational goals.
  
  Improved compliance: Adherence to regulatory standards and internal policies.
  
  Operational efficiency: Streamlined governance processes and reduced duplication of efforts.
  
  Increased trust: Confidence in data products that fosters broader adoption and usage.

We will examine the lifecycle of data products through the lens of stewardship, offering actionable insights to help organizations enhance their governance practices. It provides a detailed framework, explores hybrid governance models, and shares practical strategies for mitigating risks and maximizing value.

Introduction

The rise of data products

The shift to a product-centric approach to data management is transforming how businesses operate. Data products—such as dashboards, predictive models, and recommendation systems—have become essential tools for deriving insights and driving innovation. By focusing on delivering specific, actionable outcomes, these products enable organizations to better meet their strategic goals.

Architectural patterns like data mesh have emerged as powerful enablers of this transformation, decentralizing data ownership and empowering domain teams to autonomously create and manage data products. However, decentralization comes with its challenges: ensuring data consistency, maintaining quality, and aligning individual efforts with organizational objectives.

Without effective governance, data products can quickly lose relevance, introduce risks, or fail to deliver value. Issues like outdated metadata, poor documentation, and inadequate oversight can undermine even the most promising initiatives.

Data stewardship serves as the glue that holds the data product ecosystem together. It provides the governance structures, roles, and processes necessary to manage data effectively throughout its lifecycle. By embedding stewardship at every stage, organizations can ensure their data products remain accurate, reliable, and aligned with business objectives.

Data in the dark: The risks of ungoverned data products

Ungoverned data products create blind spots, erode trust, and introduce compliance and efficiency risks, leaving organizations to navigate a fragmented and unreliable data ecosystem.

• Lack of visibility: When data products lack adequate governance and stewardship, they often remain uncatalogued, undocumented, and hidden from the data catalog. This absence of visibility hinders users from discovering and comprehending the available data products, resulting in inefficiencies and the risk of duplicated efforts across teams. Without a clear understanding of what data products exist, organizations may inadvertently waste resources by creating similar products or failing to leverage existing assets effectively.

• Outdated metadata: Even if data products are initially catalogued with the required business metadata, without proper oversight and stewardship, this metadata can become incomplete, inconsistent, or outdated over time. Outdated metadata can lead to confusion, misunderstanding, and misuse of data products, eroding trust in the data ecosystem.

Pic: Impact of Poor data governance

• Lack of trust due to quality issues: While data engineering teams may run basic data observability checks to ensure only fit-for-purpose data is ingested through pipelines, the business-oriented data quality can degrade over time without proper oversight and maintenance. Degraded data quality can lead to unreliable insights and poor decision-making.

• Compliance risks: Inadequate stewardship can expose organizations to compliance risks, such as violations of regulations like HIPAA, CCPA, or internal data governance policies, standards, and guidelines. Non-compliance can result in hefty fines, legal consequences, and reputational damage.

• Duplication of efforts: Without clear ownership and visibility, multiple teams may create similar data products, wasting valuable resources and time. This duplication of efforts can lead to inconsistencies, inefficiencies, and a fragmented data landscape.

• Inefficient decision-making: Poor documentation and lack of context surrounding data products can lead to their suboptimal use in decision-making processes. When users lack understanding of a data product's purpose, limitations, or underlying assumptions, they may make decisions based on incomplete or inaccurate information.

The journey of a data product

Every data product undertakes a dynamic journey to transform raw insights into strategic assets that drive decisions and innovation.

A data product consists of the following six stages across its lifecycle:

1. Ideation and definition
   As the foundation of the data product lifecycle, this phase focuses on identifying business needs and brainstorming potential solutions. Stakeholders collaborate to clearly define the product's purpose, target users, and expected outcomes to ensure alignment with strategic objectives.
   
   

2. Design
   In this stage, the product's architecture is carefully crafted. Key activities include:
   Outlining data sources and processing requirements
   Defining the product’s metadata, schema, and access controls
   Detailing delivery mechanisms to ensure scalability and usability

   
   

   

   
   
   

3. Development and validation
   This phase covers the construction and quality assurance of the data product:
   Data engineering, analytics development, and UI/UX creation
   Conducting rigorous testing to validate accuracy, reliability, and compliance with quality standards
   
   

4. Deployment and adoption
   The product is launched for end users in this phase, with emphasis on:
   User training and accessible support resources to drive adoption
   Seamless integration into existing workflows and systems to maximize usability and impact
   
   

5. Evaluation and maintenance
   Ongoing assessment ensures the product remains effective and relevant:
   Monitoring performance, usage metrics, and business impact
   Implementing updates and optimizations based on feedback and changing requirements
   
   

6. Iteration or retirement
   Based on evaluations, the data product either:
   Iterates with substantial enhancements to align with evolving needs
   Is retired if it no longer provides value or aligns with business goals

Defining data stewardship

Behind every trusted data product lies effective stewardship—a blend of governance, accountability, and collaboration that transforms raw data into strategic value.

Data stewardship in the context of data products refers to the management and oversight of the data that powers these products, ensuring its quality, integrity, and security throughout the lifecycle of the product. It encompasses a set of practices and processes aimed at maintaining the accuracy, consistency, and accessibility of data within products such as dashboards, reports, machine learning models, and automated systems. Effective data stewardship ensures that data products remain reliable, relevant, and compliant with regulatory requirements and organizational policies, thereby enabling them to deliver maximum business value.

The hybrid stewardship model

The hybrid stewardship model is a tiered governance model between source and domain teams to address specific challenges and optimize data governance for data products. This is complemented by the oversight from the centralized data teams for effective management of data products. The approach aims to balance the need for standardization and consistency with the desire for agility and innovation. Some key considerations:

1. Shared accountability: Have cross-functional teams that include representatives from both the source and domain teams. This facilitates collaboration and joint decision-making on key data governance aspects such as data quality standards, data usage, and compliance, fostering a sense of shared responsibility.

2. Clarity on RACI: Outline the role, responsibility, and processes involved in data stewardship for the source and consumable data products, ensuring consistency and clarity.

3. Joint planning: Collaboratively develop a lifecycle for data products that outlines their creation, maintenance, and eventual retirement. This ensures that the source data is used, managed, retained, and archived in the data product lifecycle in accordance with the policy and guidelines of the enterprise.

Below is an illustration of a typical two-tiered stewardship model:

The roles and responsibilities of the source and domain data stewards are mentioned in the table below:

What are the key factors to consider while designing a tiered stewardship model for data products? 

1. Organizational data governance maturity

   Early stage: A simpler, two-tier model (source-centric and domain-centric) can provide a solid foundation for emerging data governance practices.

   Mature stage: A more complex, three-tier model (source-centric, product-centric, and domain-centric) can offer a granular approach to address intricate data landscapes and sophisticated data products.

   
   

2. Organization size and complexity

   Smaller organizations: A simpler model with one or two tiers can be efficient, with domain-centric stewards overseeing data quality and governance.

   Larger, complex organizations: A hybrid model with multiple tiers can effectively manage diverse data domains, complex data pipelines, and numerous data products. However, careful alignment between domain-centric and product-centric teams is crucial to avoid confusion and ensure consistency.

   
   

3. Data complexity and technology infrastructure

   Simple data landscapes: A two-tier model can be sufficient for governing relatively straightforward data sets and simple data pipelines.

   Complex data landscapes: A three-tier model can help manage complex data pipelines, integrations, and diverse data sources, ensuring data quality and consistency throughout the data lifecycle.

   
   

4. Resource allocation

   Implementing a multi-tier model of stewardship requires significant investment in terms of people, processes and technology.

   Personnel: Hiring and training data stewards with the necessary skills and knowledge

   Processes: Developing and refining data governance processes and procedures

   Technology: Investing in data management tools and platforms to support data governance activities

   
   

5. Organizational culture and mindset

   If the organization has a history of working in silos or driving centralized decision-making, introducing a three-tiered stewardship model may not be well received. It requires a cultural shift towards data-centric thinking and a willingness to share data and insights. Effective communication and alignment with centralized governance principles are essential.

Additional new roles for establishing accountability for data products

In addition to data stewards, organizations should consider following critical governance roles for data products:

Data product owner

This role is typically a business function responsible for the strategic direction and overall success of the data product. The data product owner defines the value proposition, sets requirements, prioritizes features, and ensures alignment of the data product with business objectives. They also manage access to the data product, ensuring that it meets the needs of users while adhering to governance policies.

Data product manager

Often originating from IT but possessing strong business acumen, the data product manager oversees the day-to-day development and maintenance of data products. This role involves close collaboration with both business and technical teams to ensure that the data product aligns with user needs and organizational goals. They are responsible for translating complex data requirements into actionable tasks, managing the product backlog, and ensuring the timely delivery of data products.

These two roles have seen a rise in demand with the growing adoption of a data product mindset within organizations.

Case Study

How a pharmaceutical company turned data chaos into clarity, leveraging stewardship to enhance trust and compliance.  

Background

A leading pharmaceutical company wanted to leverage data products to unlock valuable insights and drive business growth. However, in their initial rollout, the company overlooked critical aspects of data stewardship, leading to significant challenges. This case study explores the challenges faced, solutions implemented, and benefits realized by prioritizing data governance over a two-year partnership. 

Data governance challenges:

Some of the critical data governance issues encountered were: 

• Outdated and lack of standardized metadata: A lack of accountability for metadata maintenance resulted in inaccurate and unreliable information. 

  
  

• Inadequate data classification: Sensitive information, including Protected Health Information and Personally Identifiable Information, was not properly classified, exposing the organization to compliance risks. 

  
  

• Poorly captured business metadata: The lack of clear business context limited the adoption of data products and their value realization. 

  
  

• Inadequate access controls: Weak access controls and lack of oversight on data product usage compromised the data security and privacy of sensitive data products. 

  
  

• Slow data quality remediations: Unclear accountability and inefficient processes delayed data quality improvements in data products. 

Solution:

To address these issues, the company partnered with the Fractal data governance team to implement a comprehensive solution: 

Data governance by design: 

• Integrated data governance checkpoints into the data product delivery lifecycle, ensuring adherence to standards and best practices 

• Mandated the involvement of data steward in the product pods right at the planning stage of the data product delivery 

• Introduced standardized templates for capturing essential data governance artifacts such as data dictionaries, data lineage, and data quality metrics 

• Conducted awareness and training sessions to empower product managers, product owners and technical teams with data governance knowledge 
  
  

2. Introduction of the three-tiered stewardship model

The roles and responsibilities of different types of data stewards are mentioned below: 

This interaction model shows how different personas would interact to resolve data issues related to a data product: 

3. Training and change management: 

   Provided targeted, persona-based training programs to strengthen data governance skills and knowledge 

   Implemented change management strategies to drive the adoption of data governance processes and tools, such as the enterprise data catalog 

   Enabled community of practices for data stewards and product owners to empower them and facilitate peer learning 

Benefits

The implementation of this governance-centric approach yielded significant benefits: 

• Improved data quality: With dedicated stewards overseeing metadata accuracy and adherence to standards and established processes, the overall quality of data products improved dramatically from 55% to 85%. 

• Increased user trust: Users gained confidence in the integrity of datasets, leading to more usage of data products. With dedicated stewardship, users were able to resolve their queries more efficiently, leading to a 60% increase in the usage of the data products since their launch. 

• Enhanced compliance: By ensuring oversight on proper data classification and access controls, the organization mitigated risks related to data protection. 

• Optimized data utilization: Data discoverability and utilization improved through the cataloging of 80% of data assets with comprehensive business and technical metadata. 

• Time savings: Time to market for data products was reduced by 30% through streamlined governance processes, and cut data issue resolution time was cut by 40% due to clearly defined roles and responsibilities in the stewardship model. 

Conclusion 

Integrating robust governance practices into the lifecycle of data products not only addresses immediate challenges but also establishes a sustainable framework for the ongoing success of data product initiatives. By embracing domain stewardship, organizations can unlock the full potential of their data products and drive innovation, growth, and competitive advantage.