DataMesh in Practice: Lessons from Deploying Decentralised Data Ownership Across a 200-Team Organisation
Main article
Abstract
Data mesh is an organisational and architectural paradigm for data management that distributes data ownership to domain teams, treats data as a product, implements self-serve data infrastructure, and enforces federated governance. Introduced conceptually by Dehghani in 2019, it has attracted considerable practitioner attention but relatively sparse published evidence on real-world implementation outcomes. This technical communication reports on a three-year data mesh implementation at a large European e-commerce firm (approximately 15,000 employees, 200 engineering teams), detailing the architectural decisions, governance challenges, and measurable outcomes of the deployment. We document four phases of implementation, describe the specific technical choices made at each phase, and report quantitative metrics including data product discovery time, pipeline incident rates, inter-domain data quality agreement compliance, and developer satisfaction scores. The most significant finding is that the organisational transformation required for data mesh — specifically, the shift of accountability from a centralised data engineering team to domain teams — was more difficult and took longer than the technical infrastructure work. We identify three critical success factors and two anti-patterns that generalised across multiple domain teams, and we offer specific technical and governance recommendations for organisations considering similar deployments.
