For years, enterprise data architecture has followed a centralized model: data flows from source systems into a central data warehouse or data lake, where a central data engineering team curates datasets for consumption by analysts and data scientists. This model worked well when data volumes were manageable and use cases were limited. It buckles under the weight of modern enterprise data realities.
What is Data Mesh?
Data mesh, coined by Zhamak Dehghani, is a sociotechnical approach to data architecture built on four principles: domain-oriented decentralized data ownership, data as a product, self-serve data infrastructure as a platform, and federated computational governance. In essence, data responsibility moves from a central team to the domain teams that best understand the data — with shared infrastructure and governance ensuring interoperability.
The Data Product Concept
The most impactful element of data mesh is the concept of treating data as a product. Rather than data being a byproduct of operational systems, domain teams publish well-defined, well-documented, versioned, and SLA-backed data products that others can discover and consume. This shift in mindset — from data as a side effect to data as a deliberate, quality-managed output — is transformational.