
For two decades, "data governance" has implicitly meant centralized governance: one council, one catalog, and one set of policies enforced from the top down. That model worked when data lived in a warehouse and changed slowly. It breaks down when an enterprise runs hundreds of domain platforms, dozens of SaaS systems, and an AI program that touches them all.
The answer is not to abandon governance. It is to redesign it for federation.
What "Trust by Design" means
Trust by Design is an approach to governance in which accountability is distributed to the domains that produce the data, while a lean central function provides the standards, tooling, and arbitration that keep the federation coherent. Trust is not inspected centrally; it is engineered locally and verified centrally.
Three principles anchor the model:
- Domains own their data products. Each domain publishes data as a product governed by a contract, with declared owners, SLAs, definitions, and quality guarantees.
- The central function defines the contract, not the content. It owns the data product specification, platform standards, and cross-domain policies. It does not own the data itself.
- Trust is observable. Every product exposes machine-readable signals of freshness, completeness, lineage, and policy compliance, so consumers can decide what to rely on without asking permission.
Why federation outperforms centralization at scale
- Latency: Domain teams make changes in days, not quarters, because they do not have to wait for a central council.
- Fidelity: The people closest to the source own the definition. Less meaning is lost in translation.
- Resilience: One domain's failure does not bring down the entire governance program.
- AI readiness: Models can consume data products governed by contracts with confidence, rather than negotiating trust case by case.
The hard part
Federation works only if the contract is enforced. Without it, "federated governance" becomes "no governance" with extra steps. The central function must have the authority to reject a non-compliant data product. The platform must make compliance the path of least resistance, and leadership must be willing to let local domains say no to bad requests.
Trust by Design is not a softer version of governance. It is a more disciplined one, and the only version that scales with the data estate you actually have.
If a migration or master data effort has stalled, the Migration Rescue Sprint is designed to get it moving again.
For the basics in plain language, see the Simple Data series.
