Transform Your BI Team with Data Mesh

Introduction: The Enterprise BI Challenge

Modern enterprises struggle with siloed data, lengthy backlogs, and business teams frustrated by slow, centralized BI processes. The journey to a flexible, scalable, and collaborative data platform is often blocked by technical debt, organizational complexity, and conflicting priorities. Data mesh offers a transformative approach, decentralizing data ownership and enabling domain teams to drive innovation—yet the path is rarely straightforward. This in-depth guide draws on the real multi-year transformation of a 30-entity holding company, distilling key stages, lessons, and frameworks for your own data mesh journey.

Stage 1: Preparation—Laying the Foundation

• Assessing the Landscape: The initial BI team was small, with 5-8 people supporting 20-30 companies, each running disparate tools such as Keboola, SAP BW, Hadoop, and Tableau. The backlog spanned years, making it impossible to meet business demands through a centralized model.
• Vision Setting: Leadership recognized that data was critical for transforming business operations—starting from basic reporting up to advanced use cases like dynamic pricing and personalization. However, the central team simply could not deliver at scale.
• Scalable Architecture: A key insight was the need for a flexible, scalable platform. Acquisitions, divestments, and shifting priorities demanded an architecture that could adapt. The team chose Keboola for its modularity and ability to support both central and domain-specific data solutions.
• Non-Technical Stakeholder Engagement: Instead of starting with technology, the BI lead engaged business stakeholders in over 80 workshops, gathering pain points and requirements. This ensured the new data mesh vision was grounded in real business needs, not just IT aspirations.

Stage 2: Revolving—From Central BI to Empowered Domains

Reimagining BI Roles: Rather than scaling the BI department to meet ever-growing demands, the team shifted focus. The central BI team became stewards of platform governance, certified data sets, and quality, while empowering domain teams (marketing, sales, logistics, etc.) to build their own analytics and data products.

• Domain Definition: Domains were organized not only by business functions (e.g., marketing, sales, IT) but also by solution areas (e.g., pricing, call centers). Ownership and responsibility were clearly assigned for each domain, making it easier to manage complexity and enable autonomy.
• Flexible Operations: Different domains had distinct levels of data maturity. Marketing, for instance, quickly built a team of data scientists and analysts to extrapolate costs and optimize spend, while other domains like sales remained more reliant on central BI support and passive data consumption.
• Data Mesh in Action: The core BI team provided certified, well-documented data marts (e.g., sales, marketing, stock turnover), which domain teams could enrich and extend. For example, logistics developed a unit economics dataset, allocating true logistics costs per order item—a task requiring deep domain knowledge and close collaboration between BI and business experts.

Stage 3: Expansion—Scaling Data Ownership and Innovation

After 18-36 months, the company reached a new stage: domain teams were building and maintaining their own solutions, with over 60-80 active Keboola users (up from 5-8 BI engineers), and 600 Tableau users consuming insights.

• Autonomy and Collaboration: Domain teams had the data access and tools needed to innovate independently. The central BI team focused on major projects and platform stewardship, drastically reducing backlog and friction.
• Community Building: A 'data champions' program was launched, with each manager nominating progressive team members to participate in ongoing meetups, training, and best practice sharing. This community approach reinforced knowledge transfer, accountability, and culture change.
• Documentation as a Product: Recognizing that technical documentation quickly becomes obsolete, the emphasis shifted to business-oriented, outcome-focused documentation. This made data products discoverable and understandable for new users and teams.

Key Success Factors and Frameworks

1. Scalable, Modular Architecture: Choose platforms that let you easily add or reconfigure domains without major rework. Keboola's modular approach enabled fast onboarding and adaptation as the business changed.
2. Clear Domain Ownership: Define responsibilities and assign owners for each data domain or product. This decentralizes accountability, speeds decision-making, and reduces bottlenecks.
3. Community & Training: Invest in building a community of data champions. Provide regular training, peer support, and forums to share successes and address challenges.
4. Iterative Mindset: Treat the transformation as a journey, not a one-time project. Start by enabling a few domains, learn from their experience, and expand iteratively. Expect and adapt to setbacks.
5. Business-Driven Design: Start with business pain points and objectives. Involve stakeholders early and often to ensure your data mesh strategy aligns with real-world priorities.
6. Governance Without Bureaucracy: Use lightweight governance to ensure data quality, consistency, and discoverability—without creating bottlenecks. Data catalogs and product documentation are essential, but so is a culture of collaboration.
7. Change Management: Encourage dialogue between teams when business rules or data products change. Use versioned APIs or duplicate datasets when needed, allowing smooth transitions and minimizing disruption.

Practical Examples and Outcomes

• Logistics Innovation: The logistics domain developed advanced unit economics models, transforming how costs were allocated and driving better decision-making across the business.
• Marketing Analytics: The marketing domain centralized data from Google Analytics and ad platforms, then extended it with domain-specific models for forecasting and optimization.
• Business-Led Data Product Development: Domains not only consumed reports but engineered data products reused across the organization, feeding data back into operational systems and driving business value beyond dashboards.
• Reduced Backlog and Friction: By distributing data engineering across domains, the company virtually eliminated its multi-year backlog and transformed BI from a bottleneck to a strategic partner.

Challenges and Lessons Learned

• Dependency Management: Domains often developed overlapping or similar data products. The solution was to allow parallel innovation, then consolidate proven approaches into core datasets over time.
• Cultural Change: Transformation required ongoing relationship-building, workshops, and executive support. Change did not happen overnight; it took years of deliberate effort.
• Balancing Autonomy and Consistency: Too much autonomy risks fragmentation; too much central control stifles innovation. The right balance is achieved through clear ownership, community, and pragmatic governance.
• Continuous Improvement: Business needs change, and so do data products. Open communication, versioning, and a product management mindset are essential to avoid disruption while evolving the data landscape.

Why Keboola Is Critical for Data Mesh Success

Keboola’s platform was an enabler for each stage of the transformation:

• Fast Onboarding: Teams could quickly spin up new domains, connect data sources, and build data products without deep technical expertise.
• Collaboration: Shared data catalogs, reusable components, and integrated documentation supported both central and domain teams.
• Governance: Keboola provided the tools for lightweight governance, certification of data marts, and auditability without unnecessary bureaucracy.
• Scale: As the organization grew, Keboola scaled to support hundreds of users, dozens of domains, and diverse data products.

Conclusion: Key Takeaways for Your Data Mesh Journey

• Start with business needs, not just technology.
• Empower domain teams while providing central stewardship and governance.
• Invest in community, training, and documentation as first-class citizens.
• Be patient and iterative—transformation takes time and ongoing effort.
• Choose platforms like Keboola that support flexibility, collaboration, and scale.

By following these principles and learning from real-world examples, your organization can move from BI bottlenecks to a collaborative, data-driven enterprise where innovation flourishes and business value is unlocked at scale.