We want to enable a culture of innovation in a data-driven company by making data valuable and useable for Data & AI use cases.
CTO, Client
Background: From Central Data Provisioning to Domain-Driven Ownership
The client is a large enterprise operating across multiple business domains, each with distinct data needs and analytical ambitions. Leadership had set a clear direction: move beyond centralised data provisioning toward decentralised ownership in line with Data Mesh principles according to Zhamak Dehghani.
For years, data had been managed through central teams. While this provided control, it created growing dependencies – every new use case required coordination across multiple teams, platforms, and approval chains. Without end-to-end ownership, accountability for data quality, freshness, and structure was diluted across teams that often had limited direct relationship with the data they were handling. As demand for data-driven insights accelerated, the central model could no longer keep pace.
The organisation recognised that domain teams, closest to the data, were best positioned to own and deliver it. Shifting to distributed ownership required more than a mandate – it required architecture, governance, enablement, organizational change and trust.
Pain Points: Central Dependencies and Governance Gaps
The most visible symptom was speed – or rather, the lack of it. Central analytics teams had become delivery bottlenecks. Every request, regardless of size, funnelled through the same overloaded pipelines and the same handful of specialists. New use cases required coordination across multiple dependencies and manual handoffs before any value could be delivered.
The lack of end-to-end ownership compounded the problem. With multiple teams involved in each data journey – most with limited context about its origin or intended use – accountability gaps meant issues were discovered late, often only when a downstream use case failed.
Beneath the surface, the challenges ran deeper. Technologically heterogeneous platforms operated with limited interoperability, making lineage difficult to trace. Governance was fragmented: GDPR, the EU Data Act, and the AI Act all demanded compliance, but enforcement relied on manual effort and institutional knowledge. The existing platform landscape had grown organically and was not yet fully aligned with modern cloud-native capabilities. A scalable operating model for delivering data products that are easy to understand and consume had yet to be established.
Goals & Strategic Guidelines: A Scalable Foundation for Data-Driven Excellence
The transformation was guided by a set of principles balancing autonomy with consistency. Domain teams would own end-to-end delivery and quality of their data products. Unified patterns for ingestion, transformation, and provisioning had to work across multiple clouds and platforms. Design patterns to document business semantics had to be in place to enable data owners to describe their data efficiently. Governance needed to be federated by design – automated policies, fine-grained permissions, and transparent lineage built in from the start, not added afterwards.
Data products had to be discoverable, well-documented, access-controlled, and fit for purpose. The organisation aimed for a single logical data environment that minimised replication while ensuring quality-assured consumption. Clear roles, enablement programmes, and well-defined processes would give domain teams the confidence to deliver independently.
Impact: Autonomous Teams, Governed Data, Scalable Delivery
The modular Data Mesh approach significantly accelerated the delivery of analytical insights across the organisation enabling wide spread use of available data. Today, over 300 data products serve more than 100 use cases. More than 30 domain teams deliver independently, with transparent ownership from creation through consumption. Governance is unified and automated across the primary data platforms, ensuring compliance without slowing delivery. Every reused data product saves the company significant cost, as existing data pipelines can be leveraged without duplication.
The operable semantic layer positions the organisation for AI-powered analytics and agentic interactions with enterprise data. Time-to-value has improved significantly – standardised data discovery, streamlined approvals, and centralised discoverability mean teams spend less time searching and more time delivering insight.
