Is Data Mesh Losing Its Momentum? A Shift Back to Centralised Data

Posted 2 months ago
by Tom Brammer
by Tom Brammer

Over the past few years, Data Mesh has been a hot topic in the world of data engineering, promising decentralisation, autonomy, and domain-driven ownership.

However, recent conversations with data leaders suggest a shift in the opposite direction—many teams are moving back to a centralised data model, favouring Data Lakes and Data Warehouses.

Why Are Companies Reverting to Centralised Models?

Several key challenges with Data Mesh are prompting organisations to reconsider their approach:

Scalability of Self-Service Data Products

  • Building scalable, self-service data products for the entire organisation is proving more manageable in a centralised model, where data teams can work together rather than in silos.

Developing a Semantic Layer

  • A centralised model enables a unified semantic layer, making it easier for non-technical users to access and interpret data without domain-specific barriers.

Consistency in Data Governance & Cataloguing

  • Ensuring consistent governance and metadata management across multiple decentralised teams is a challenge. A central repository allows for more structured and standardised cataloguing.

Simplifying AI & GenAI Implementation

  • With AI and GenAI becoming more prevalent, having a single data repository makes it significantly easier to implement machine learning models without fragmented data sources.

Was Data Mesh Over-Sold?

While Data Mesh introduced valuable principles around domain ownership and federated governance, some argue it was presented as a universal solution rather than a framework suited to specific organisational structures. A few key points raised by data leaders include:

  • End-to-End Use Cases Are Harder in a Decentralised Model
    • Many organisations struggle with cross-domain data use cases, which are significantly more complex in a decentralised structure.
  • Data Mesh Doesn’t Solve Integration Challenges
    • Some argue that while Data Mesh decentralises ownership, it doesn’t inherently make data integration easier—if anything, it adds more friction when attempting to unify datasets.
  • Companies Confused Architecture with Organisation
    • Data Mesh is not about decentralising data storage, but rather decentralising data ownership. Some organisations misunderstood this and ended up with highly fragmented systems.

The Future: Centralised, Decentralised, or Hybrid?

While some are returning to centralised models, others suggest a hybrid approach—leveraging Data Mesh principles for ownership and accountability while maintaining a central data repository for consistency and AI scalability.

This shift raises important questions:

Are you seeing the same trend in your organisation?

How do you structure your data teams—Data Mesh, Centralised, or a Hybrid approach?

The Product Talent Gap

Could the UK’s product job market be heading for trouble? Personally, I can count on one hand the junior roles I’ve worked on since before Covid. And with some fairly elementary digging on Linkers, the data speaks for itself: 614 Senior Product roles posted last month. Only 28 Junior Product roles. That’s a 22:1 ratio! […]

SUBMIT A VACANCY

Send us the details of your job opening and one of our consultants will be in touch to discuss suitable candidates.

UPLOAD YOUR CV

Send us your details and one of our consultants will be in touch to discuss suitable roles.