Connect meaning, not just transport
The data fabric provides a shared operating record across authorized surfaces. Its job is not moving bytes — transport is the easy part. Its job is making a fact from one system mean the same thing everywhere else it appears, inside the brand boundary it belongs to.
That is why an integration begins with authority and meaning before any connection is opened. A connection that moves records without establishing who owns them, which brand they belong to, and which system is currently authoritative has not integrated anything — it has created a second copy that will drift. The companion essay build the shared record before the dashboard makes the same argument from the operator's side.
Source orientation
Before a source is connected, record in writing:
- Owner and brand context. Who is accountable for this source, and which brand's boundary its records fall inside? Brand context is attached at ingestion and validated on every read — the same database-level isolation the rest of the platform runs on.
- Identifiers. Which field uniquely names a record, and how does it map to the shared record's identity? A source that cannot name its records cannot be reconciled.
- Update pattern. How often does the truth change, how quickly must the shared record reflect it, and what does "current enough to act on" mean for this workflow?
- Authority during migration. Which system remains the source of record while the fabric is proven — and for which fields? Authority moves field by field, on evidence, never in a single switch-flip.
- The decision each field supports. If no decision consumes a field, its freshness and accuracy will never be reviewed. Fields without decisions are how fabrics rot.
Safety questions
These are the questions an operator should be able to answer about any live connection — and the ones our own onboarding review asks:
- How is brand context attached and validated on every read and write?
- Which identity and role may read or write each record?
- How is lineage preserved from source to shared record — can you trace a value back to the system and moment it came from?
- What counts as stale, partial, duplicated, or contradictory state for this source?
- Does uncertainty block action, degrade safely, or require operator review — and who is that operator?
- Where can an operator inspect connection health and exceptions without filing a ticket?
Warning
A connection is not complete when data moves. It is complete when operators can verify its authority, freshness, failure behavior, and evidence — the same standard integration boundaries states for builders.
What the fabric makes possible
Once the shared record is trustworthy, the compound surfaces read from it: dashboards that need no reconciliation meeting, agents that act on verified state inside governed boundaries, and ledger surfaces like billing & treasury splits that post against the same facts operations already uses.
What the fabric is not
The fabric is not a data lake — a place copies accumulate and hope compounds. Every record in it has an owner, a brand context, and a freshness expectation; anything that cannot name all three is not fabric content. It is also not a replacement for source systems of record during migration: authority moves field by field, on evidence, and the fabric is explicit about which system wins for every fact until the move completes. And it is not an analytics side channel — derived and indexed data inherits the same tenant boundary as the rows it derives from, covered at database isolation mechanics.
Boundary
This page does not publish private endpoints or authentication material. Production connection details and live health belong in authenticated developer and operator surfaces.