You own a Sunrise 2027 program. It has a named owner, a budget line, and a date you did not set. The symbology decision is drawable. The data-pool feeds are live. What the plan still does not name is the event layer — where the scans your new 2D mark makes possible get recorded, and who says they happened.

That is the line item a GS1-literate skeptic on your steering committee will reach for first. This page is the read that survives that question: what Sunrise 2027 requires of a GTIN owner versus the converter and the retailer, where the FSMA 204 date really stands, and how a chain of custody holds across co-manufacturers you do not own.

We conform to GS1's standards. We do not stand above them. Everything below cites a public artifact you can open yourself, or it is cut.

What does GS1 Sunrise 2027 actually require of a brand owner?

Sunrise 2027 is a GS1 program: by the end of December 2027, retail point-of-sale is expected to scan and process 2D barcodes at the lane [A5]. That is the whole of the dated expectation, and it is aimed at the retailer's scanner — not at your pack.

Read precisely, three things follow for a GTIN owner, and each one is a decision that is yours, not the program's:

  • Sunrise puts nothing on your pack. You do — if you choose to. The baseline the program is built around is GTIN extraction at point of sale. The retailer's lane must be able to pull a GTIN out of a 2D symbol. Whether your artwork ever carries a 2D symbol, and what travels inside it, is a brand-owner decision.
  • **Lot and expiry are optional Application Identifiers you elect to encode.** A GS1 Digital Link URI can carry batch/lot and expiration as additional AIs — but that is your call as the party that owns the artwork and the GTIN, weighed against what it unlocks downstream. It is not a Sunrise requirement.
  • A QR that a shopper's phone resolves is a different object than a symbol that only clears a lane. A QR code carrying a GS1 Digital Link URI is web-resolvable — a phone camera opens it. A GS1 DataMatrix clears the retail lane and is the pharma/regulated-goods symbology; it is not the consumer's symbology. And a GS1 element string (the FNC1-delimited form) is not a web-resolvable URI at all — only the Digital Link URI form is. Which of these you print is a decision the program hands to you.

The hard part of Sunrise 2027 was never the symbol. It is owning an obligation across plants you do not own.

The GS1-literate correction, stated plainly so no one on your committee has to make it for you: Sunrise 2027 is a lane-capability program with a POS-side expectation. It does not mandate a mark on your product, it does not require lot/expiry, and it does not decide your symbology. Those are brand-owner decisions — which is exactly why the program plan has to name who makes them.

Is the FSMA 204 compliance deadline still January 2026 — or did it move?

It moved. Lead with the operational reason to do this work, and treat the regulation as the floor it actually is.

The public FSMA 204 record: the compliance date in the proposed rule of 2025-08-07 moved from January 2026 to July 20, 2028 [A7]. It was never finalized as a rule. The extension rests on the Continuing Appropriations Act of 2026, and FDA is soliciting further flexibilities (Federal Register, 2026-05-28). Anyone who walks a steering committee a "gun-to-the-head January 2026" framing is one search away from being corrected in the room — and the plan built on that framing loses its funding with it.

So do not build the plan on the date. Build it on the operating cost the date merely de-risks:

  • The reconciliation meeting exists only because no record carries who observed what. Every audit that turns into archaeology, every mock recall that takes days across the co-man boundary, every spreadsheet reconciled from three systems — that is the standing operational tax. FSMA 204 does not create it; it makes ignoring it expensive on a schedule.
  • FSMA 204 is the de-risking floor, not the urgency. The Key Data Elements and Critical Tracking Events a Food Traceability List product must carry are a floor your event layer should clear as a byproduct of being built correctly — not the reason to build it.
  • Timing comes from Sunrise, cost comes from FSMA, and you never reverse them. Sunrise 2027 gives you a date the retailer set. FSMA 204 gives you a floor and a return. The urgency is the operating tax you are already paying.

The plan that survives November is the one whose first slide is a cost you can already measure, and whose regulatory slide is a floor you clear on the way past.

Who decides the 2D symbology and the Digital Link URI — us, our co-packer, or the retailer?

You do. This is the decision-boundary that the program plan has to draw explicitly, because in practice three parties touch the mark and only one of them owns it.

  • You (the brand owner) own the GTIN and the artwork. You decide the symbology, the GS1 Digital Link URI, and which AIs travel in the string. This decision is made once, here, at the GTIN-owner seat.
  • Your converter / co-packer applies what you decided. They put ink on substrate. They do not choose the mark and they do not read it at a lane. If your decision is ambiguous, their default becomes your policy — which is why the decision has to be drawn before the artwork cycle, not during it.
  • The retailer reads it at the lane, per their Sunrise 2027 timeline, and requests "2D-readiness" of you — but the retailer does not decide what your symbol carries. Their request lands on your desk as a cost line; your job is to convert it into a decision you own.

The one decision none of these three parties has assigned an owner is the event layer: when the mark gets scanned — at your plant, at the co-man, at receiving, at the lane — where does that event get recorded, and whose account says it is true? That is question five, and it is the one your plan is missing.

How do we prove chain of custody across co-manufacturers we don't own?

This is the structural problem, and it has a structural answer that you can verify in the specification yourself before you take our word for anything.

The executive gift — the gap in the standard, checkable in ten minutes. EPCIS 2.0 §7.2.2 defines an event with five dimensions — what, when, where, why, how. Read the list. There is no performer among them. The party fields the standard does provide are organisation-grain: EPCIS §7.3.6.4 and CBV §7.4.3 carry parties as source/destination organizations, and §8.7.1 identifies a party by PGLN — a company's number.

So the standard answers a Company did a Process. It structurally cannot answer a Worker (or an agent) did a Task at the moment of the scan. There is no "Who" field to leave blank — there is no field. Across a co-man boundary, that is precisely the question an auditor asks and the reconciliation meeting exists to reconstruct: not which company, but who, on whose authority, observed this.

§7.2.2 · §8.7.1 — ten minutes, the specs are public, don't take our word for it.

How a record holds across a boundary you don't own. The record carries a two-grain envelope: a who — an attested observer, whether a person, an agent, or an embodied agent — and a capturedBy — the warrantor account that stands behind the observation. Company grain and party grain are derived at read time from grant chains, never stamped into the event. The consequence is the one that matters across a co-man network: the record survives a reorganization and a grant revocation, because who-observed-what is not frozen into the row — it is resolved from the authority that held at capture time. You can answer for your own product without asking the network's permission, and the answer still holds after the org chart changes.

The honest gap, stated rather than papered over. There is no citable public artifact for an enterprise retail CPG brand's own 2D/EPCIS program. The closest public evidence — a Fortune-tier brand funding a manufacturer→restaurant RFID pilot with a named supplier [A1], and the same program seen from the plant that serves it [A4] — is QSR-supply shaped. It shows the shape of an enterprise supplier pilot; it is not a retail-CPG deployment, and we will not dress it up as one. That is a gap in our evidence, not in the market.

Where do the events behind our new 2D mark get recorded, and by whom?

Here is what the platform is, stated plainly, with the built/unbuilt line drawn where it actually falls.

The engineering ledger — for the platform team you will forward this to. Our conformance claim is a test run, not a brochure:

  • The projection law: project(event) validates against the official GS1 EPCIS 2.0 JSON schema. Validators are precompiled over pinned official GS1 schemas with sha256 provenance in PINS.json, at EPCIS 2.0.1.
  • EPCIS 1.1 / 1.2 / 2.0 XML translated to EPCIS 2.0 JSON-LD with a per-job round-trip fidelity report — a decade of your existing event exports, read without loss.
  • The CBV 2.0 §8.9 standardized event hash, with GS1 Digital Link normalisation, gated against OpenEPCIS reference vectors — so your hash matches your partner's hash.
  • Errors as RFC 7807 application/problem+json, carrying the standard's own exception types.
  • Gateway-stamped record time, capturing party, and attestation grade; caller-supplied values in those fields are stripped. Storage is append-only — no service identity holds an UPDATE or DELETE grant.
  • An MCP door so a procurement-class agent uses the same interface and the same key as a person.

schema.org v30.0 (2026-03-19) welded the physical and digital vocabularies at company grain — 1 owl:equivalentClass + 15 owl:equivalentProperty GS1 annotations [A6]. (The hasGS1DigitalLink property is pending, not core, and has shipped since v27.0 — it is not new in v30.0. We state the version history because a GS1-literate reader will check it.)

The proof line, in the open, not the footnote. The spine's stateless doors are live at epcis.dev — POST /translate, /validate and /hash — and 683/683 spine tests pass. No conformance attestation has ever been issued, and none is claimed. The durable write path (hosted capture at api.epcis.dev) is a named entry in the open P0 ledger; the dated summary is our ledger of what ships today.

On cost, as policy: recording an event never costs money and is never shown as a meter — stated ahead of published terms. Revenue attaches to answers — traces, custody evidence, partner grants — never to capture.

Is the 2D mark a consumer signal for an enterprise brand too?

Partly — and honestly, less than it is for a smaller brand, so we will not oversell it to you.

A 1D UPC is a checkout token: it talks to someone else's POS and tells your brand nothing. A QR carrying a GS1 Digital Link URI makes the same square resolve in a shopper's phone — a first-party signal class you did not have. For a mid-market brand whose only customer view is syndicated panel data, that is revelatory (that is the mid-market consumer-signal story, and if your delta is customer-blindness, start there).

For an enterprise brand that already runs loyalty and retail-media, the unlock is incremental, not revelatory: the delta is unit-level granularity — this lot, this expiry, this specific item — that no loyalty program carries. Whether consumers scan at volume is a bet, not a fact; there is no public evidence we can point you to, and we host no consumer-scan analytics today. We are telling you this plainly because a skeptic on your committee will, and the plan should say it first.

The plan, in three steps you can run before any ask

  1. Take the read. The segment-specific read of what Sunrise 2027 and FSMA 204 actually require of an enterprise CPG brand owner — including the FSMA facts most vendors will not tell you [A7]. Free, no gate. The read leaves with you whether or not you ever come back.
  2. Run one of your own files through it. translate a real EPCIS XML export and validate it against the pinned schemas — one curl each: POST to https://epcis.dev/translate and /validate, round-trip fidelity on your own data in the response. Hand it to your platform engineer via epcis.dev, the same spine from the developer side. Free, no gate.
  3. See what the record looks like with a performer in it. A sample projection carrying the two-grain envelope — who and capturedBy, party and org grain derived at read time from grant chains, never stamped — so the record survives a reorg and a revocation. Free, no gate.

Then, and only then, the ask.

What happens if the plan is built on the deadline instead

You present a 2027 plan built on a vendor's deadline framing. Someone on the steering committee searches FSMA 204 in the room and finds the 30-month slip and the un-finalized rule. The plan loses its credibility and its funding in the same meeting — and the real work restarts in 2027 with eleven months left and a co-man network that still cannot answer who observed this.

The alternative is the plan that goes from "we can reconstruct what probably happened, given a week" to "every event answers who observed it, we can prove it against a pinned schema, and the proof holds after the org chart changes."

Get the read — and get on the seat list

The interview that produces your program-office read (the enterprise-cpg branch) is live: get started — your address first under the one-message promise, then questions that branch on your answers, ending in the written read for your program. It runs on this origin, needs no script, and locks when it ends. The whole instrument is machine-readable at /flow.json.

The lighter door is the one-question list: Put me on the seat list → name your side of the handoff: Manufacturer or brand

You pack it out and you own the GTIN. That is the manufacturer branch — the one your program office sits on.

The waitlist promise, exactly: Your capture workspace is provisioned from this list, in order. One email when your seat is ready — one, not a drip campaign. Nothing else, ever — no newsletter, no sequence. If we stop working on this, you get one message saying so and your address is deleted. Either way it is one message. The named human sender and the verified sending address are a launch gate, recorded as P0-V4 in the open P0 ledger; the message is not sent until both exist.

And on the conversation you may be tempted to ask for: We answer in writing. We take at most five conversations a month, only when you ask for one, and only after you already have the written read.

Conforms to EPCIS 2.0 and CBV 2.0. GS1 is the standards body; this platform is not affiliated with, endorsed by, or certified by GS1. The developer lens on this same spine is epcis.dev.