Running Two Traceability Regimes in One Plant Network

There is a version of FSMA 204 that would have been easier for an enterprise CPG brand than the one that exists: full coverage. If every SKU were covered, you would run one record discipline everywhere and be done. What the Food Traceability List actually does to a large portfolio is worse — it covers part of it. Some SKUs are in scope, most are not, the covered ones run on the same lines and through the same co-manufacturers as the uncovered ones, and the Director of Food Safety inherits the operational cost of the split: two traceability regimes in one plant network.

Why partial coverage costs more than full coverage

Bifurcation is expensive in ways a compliance budget never itemizes. Line operators must know which regime the SKU in front of them belongs to. Quality documents fork. Supplier onboarding forks. The mock-recall playbook forks — and the exercise that crosses the fork is the one that runs long. Every place the two regimes touch, someone maintains a mapping between them, and every mapping is a place the record can silently disagree with itself.

The four-quadrant matrix makes the cost concrete. Plot your portfolio on two axes — FSMA-covered vs. not, owned plant vs. co-manufactured — and mark where the record actually lives today:

Four quadrants, four record postures, one brand accountable for all of them. The top-right cell is the one that fails audits — a mandate you must answer for, executed in plants you do not control (the co-manufacturer's side of that same boundary is its own read: FSMA 204 flow-down for co-manufacturers) — but the deeper problem is that the four postures exist at all.

One spine, two projections: the record does not fork, the view does

The architecture that survives is the one where the record is uniform and the regime is a view. Every lot event — production, shipping, receiving, transformation — is captured once, at one grain, into one catalog, regardless of which quadrant the SKU sits in. Regulatory scope is then a projection over that catalog, not a property of the capture.

Here is the same illustrative production event rendered both ways. First, as the FSMA KDE row a records request would ask for:

And as the very same event in its conformant EPCIS 2.0 projection — an ObjectEvent with bizStep: commissioning, the lot-level EPC class, the quantity element, the read point, and the event time. One observation, two renderings. The KDE row is not a second record that must be reconciled to the first; it is a computation over the first. visibility.cloud's every projection validates against the pinned official GS1 EPCIS 2.0 schema — superset where our record carries more, conformant always; no conformance attestation has ever been issued, and none is claimed.

The payoff shows up on audit day: when the executive answer is a view computed over the same catalog the open developer surface queries, there is no reconciliation step to defend. An auditor can re-derive the KDE table from the events rather than trust an export — recomputability is structural, which is the platform's standing argument in the answers.

What uniform capture buys each quadrant

Covered, owned: the KDE view falls out of the record your quality team already wanted for its own reasons — lot genealogy with a who at every step, the attested observer distinct from capturedBy, the warrantor account.

Covered, co-man: the mandate flowing down to your co-manufacturer becomes emit the same events you already emit for quality, grant us the view — a far smaller ask than join our compliance system, and the ask the co-man's own economics prefer, as the flow-down read argues.

Uncovered, owned and co-man: capture costs nothing extra because the discipline is uniform — and here is the option value: if FDA revises the list, scope expansion is a configuration change. The uncovered half of your portfolio is one edit away from covered, on the day the record already exists. Under a bifurcated architecture that same revision is a new project with a new deadline.

Standardize the event, not the regime

The program office's move this cycle, concretely: standardize the event vocabulary across all four quadrants — what gets captured at production, shipping, and receiving, at lot grain, with a performer — and let regime views multiply on top as regulators, customers, and your own quality organization demand them. The alternative, standardizing per-regime, guarantees you rebuild the record layer every time a rule, a customer mandate, or the list itself changes.

This is one decision inside the larger one the enterprise programme office already owns — the mark, the data pool, and the event layer, on the Sunrise clock — and that full read is the pillar: Sunrise 2027 for brand owners. Who makes each of those decisions, and why the event layer is the one still open, is worked through in you own the GTIN, so you own the mark.

Where this goes next

visibility.cloud provisions capture workspaces from the seat list, in order. The way in is the interview: email first, under a one-message promise, then a short branching sequence about your network — owned plants, co-man share, covered portfolio — ending in a written read for your situation. The final step locks.

→ Start the interview — it is questions, not a demo.