What must exist before decentralising data ownership makes things better rather than worse?

What is being tested

Whether you know that the platform is a precondition rather than a later phase.

What must exist

1. A self-serve data platform. Domains cannot each build ingestion, storage, cataloguing, access control and quality tooling. Without it, decentralisation means every domain solving the same problems badly, and the estate is worse than the centralised version it replaced.

This is the precondition that is most often deferred and most often fatal.

2. Capability in the domains. A team with no data engineering skill will not produce a good data product. Either the skill exists, or it is hired, or an enabling team transfers it — and that is a resourcing decision that must be made explicitly.

3. Incentives. If nobody is rewarded for data quality, data remains exhaust. The team producing it must be accountable for its published form — schema, freshness, quality, documentation — and that accountability must be real.

4. Federated governance with actual standards. Global agreement where interoperability requires it: identifiers, security classification, interchange formats. Without it, decentralisation produces datasets that cannot be joined.

5. Sufficient scale to have the problem. In a small organisation a central data team is more efficient. The mesh addresses a coordination problem that only exists above a certain size, and adopting it below that adds overhead for no benefit.

The principle worth adopting regardless of scale

Data as a product. A published dataset has consumers, a documented contract, quality guarantees, a freshness SLA and a support path.

That converts data from a by-product into something with an owner and a standard, and it is valuable at almost any size. It can be adopted without any of the organisational restructuring.

The failure modes

Adopted as a technology. There is no data mesh product; buying a catalogue and declaring a mesh changes nothing.

Decentralisation without the platform, which is the common failure.

Domains without capability or incentive.

Applied at the wrong scale, adding overhead to a problem that does not exist yet.