Traditional centralized data warehouses often struggle to keep up with rapid growth. Monzo, the digital bank, faced this challenge as its team expanded. In response, they implemented a 'meshy' governed data mesh that not only cut costs by 40% but also accelerated data delivery by 25%. Read on to learn how they achieved this and what it means for modern data architecture.

What drove Monzo to redesign their data warehouse?

Monzo experienced explosive growth in both customer base and internal teams. Their original centralized data warehouse became a bottleneck: over 100 teams were competing for access, leading to slow query performance, long wait times for new data models, and rising costs. Engineers spent excessive effort on coordination and governance rather than delivering insights. The existing system simply couldn't scale efficiently, prompting Monzo to explore a more distributed approach that would empower teams while maintaining control over data quality and compliance.

What is the 'meshy' governed data mesh that Monzo implemented?

The 'meshy' approach is a variant of the data mesh architecture, where data ownership is decentralized across domain-specific teams. Unlike a pure data mesh, Monzo added a layer of governance to ensure consistency, security, and reusability. Each of the 100+ teams owns and manages its own data products using dbt models, but shared infrastructure and standards are enforced centrally. This balances autonomy with control, allowing teams to move fast while maintaining trust in data. The result is a flexible system where data is treated as a product, with clear ownership and accountability.

How did Monzo manage to involve over 100 teams and 12,000 dbt models?

Monzo achieved this scale by adopting dbt as the standard transformation tool across the organization. They built a modular framework where each team could create and maintain their own dbt models independently, but within shared naming conventions and quality gates. Centralized metadata and discovery tools helped teams find and reuse existing models, reducing duplication. Regular training and documentation ensured that even non-experts could contribute. The governed data mesh provided the guardrails, while the teams had the freedom to innovate, leading to the rapid growth from a few dozen models to over 12,000.

What specific results did Monzo achieve with this data mesh?

The transformation yielded impressive quantitative gains: warehouse costs dropped by approximately 40% because each team's data products were more efficiently scoped and queried, reducing idle compute. Data delivery speed improved by 25%, meaning insights reached decision-makers faster. Qualitatively, teams reported higher satisfaction and reduced time spent on data pipeline maintenance. The governed mesh also improved data quality and compliance, as teams were responsible for their own data products but adhered to shared standards. Overall, Monzo turned a scaling challenge into a competitive advantage.

How does the governed data mesh differ from traditional data warehouse architectures?

In a traditional centralized warehouse, a single team manages all data pipelines, leading to bottlenecks and a lack of domain expertise. The governed data mesh flips this model: each domain team owns its data—from ingestion to modeling to serving. This distributes the workload and aligns data ownership with business knowledge. Governance is applied through shared policies, tooling, and a central catalog rather than a central team controlling everything. The result is greater agility, scalability, and data quality, as demonstrated by Monzo's ability to support 100+ teams and 12,000 models without sacrificing control.

What lessons can other organizations learn from Monzo's data mesh journey?

Monzo's experience offers several key takeaways: First, governance doesn't have to slow things down—when designed as shared standards rather than rigid controls, it actually enables scale. Second, invest in tools like dbt that empower domain teams while maintaining consistency. Third, prioritize a strong metadata and discovery layer so teams can find and reuse existing data products. Fourth, be prepared for cultural change; moving from a centralized to a mesh model requires trust and training. Finally, start small: Monzo iterated on their approach. Other organizations can adapt these principles to their own context, aiming for a balance between autonomy and governance.