The blog.
Briefs from the team building the event record with an attested who on every event. Each one makes one claim and carries its proof: a spec citation, validator output, or a worked example from the tested spine.
-
Seen here, then not: what read-point data cannot tell your investigators
A worked custody chain across one vendor handoff, read by read — which questions the reads answer, which they structurally cannot, and what a performer-attributed record would add to a chargeback nobody can currently win.
-
The 2D-readiness questionnaire is coming: what your retailers will ask and when
Retailer 2D-readiness requests arrive as questionnaires with dates. Because the lane's obligations flow uphill, the questions are derivable in advance — symbology, URI form, lot encoding, timeline — and a mid-market ops leader who can predict them answers from capability once instead of scrambling per retailer. The model questionnaire, worked.
-
Answering the traceability clause: how a 3PL turns an RFP liability into a win theme
A model shipper traceability clause dissected obligation by obligation — the answer that wins versus the soft answer that survives — and the custody handoff where the typical record goes silent.
-
An agent seat is a bounded mandate, not a shared login
Giving an AI agent your password creates an unaccountable ghost user; provisioning it a seat — a mandate with a ceiling, a scope, and a distinct identity on every act — is the difference between delegation and impersonation.
-
Anatomy of a 24-Hour Mock Recall That Took Four Days
Mock recalls fail at organizational boundaries — the franchise line and the co-man line — because the record that crosses them carries no performer, so every hop becomes a phone call.
-
The artwork cycle is the decision window: 2D for co-manufactured brands
For an asset-light brand the 2D decision is made in the artwork cycle — your cycle, not your co-man's and not your retailer's — and it is the rare, cheap moment when a compliance square can also become the brand's first line to the consumer of a unit. The cycle, worked, with the decision placed at its only cheap insertion point.
-
The ASN's Hierarchy Is Your Custody Tree: Pallet, Case, Each
The shipment-order-pallet-case-item hierarchy your plant already emits in every ASN is structurally the same tree as the aggregation events a traceability record needs. You author the custody tree today — and throw it away after transmission.
-
Chargeback disputes where both sides have data and neither has evidence
A model short-ship claim replayed twice — once as dueling CSV exports, once as hash-identified, independently verifiable events with an attested receiving observer. The second version ends in minutes.
-
The cheapest window in a decade to fix receiving capture
Sunrise 2027 is not aimed at a QSR — no lanes, no POS clock. But your suppliers are re-labeling cases on the Sunrise timeline anyway, which makes 2026–2027 the cheapest window in a decade to change what gets captured at receiving. A timing argument, with Chipotle's public rollout as the evidence.
-
Cold chain: where custody and condition are the product
One worked reefer-to-cooler handoff as an event sequence — what read-point-plus-temperature data answers, and the two questions it cannot: who accepted custody, and under whose warrant the observation was recorded.
-
Deputized triage — an agent working your exception queue inside granted scope
An exception queue is the ideal first agent seat — bounded queries, already-granted read scope, a human-set mandate ceiling — because the agent works the volume while every escalation lands on a human by construction, not by hope.
-
The developer-EDI field left supply chain. Your documents didn't leave with it.
The best developer-EDI platform of the last decade now sells healthcare clearinghouse transactions, and nobody left in the field connects trading documents to custody events — a dated, checkable account of where the supply-chain join now lives.
-
Direct Store Delivery Is Where Your Traceability Record Goes Dark
DSD receiving is the thinnest link in grocery traceability because the receiver is a store associate the record has no field for — the standard structurally cannot say who signed.
-
A 2D barcode full of element strings resolves to nothing: the URI-form decision
The same GTIN and lot can be encoded as GS1 element strings or as a GS1 Digital Link URI. A lane reads both; only the URI form opens in a browser. Which form your pack carries is the GTIN owner's explicit choice, with a decade of consequences — shown here as two actual payloads.
-
An event's identity is computed, never assigned: the CBV §8.9 hash, explained for executives
The standardized event hash gives every record an identity derived from its content — duplication is detectable, tampering is self-evident, and two parties can agree an event is the same event without trusting each other's databases.
-
The first-party unit-level signal: the option is the fact, the volume is the bet
A 1D UPC is a checkout token that teaches a brand nothing. A Digital Link QR creates the option of a first-party, unit-level signal class no loyalty program or panel resells — and an honest brand plan prices the option while treating scan volume as the open hypothesis. Stated concretely, then the falsification plan.
-
Five Dimensions, No Performer: The One Fact to Check Before You Buy Anything
EPCIS 2.0 defines what, when, where, why and how — and no performer is among them. The section numbers to check it yourself, in ten minutes, before any vendor conversation.
-
Franchise data rights — who owns the record when the boundary is a contract?
In a franchise system the traceability record crosses a boundary where data rights are contractual, not organizational — the workable architecture grants scoped views across the franchise line rather than pretending one party owns the whole chain's history.
-
From stripes to grids: who touched it?
The founding manifesto. Fifty years of class-grain stripes end on GS1's clock, not ours; the grid with a record behind it answers what, when, where, why, how — and leaves exactly one question standing.
-
Complying With a Mandate You Didn't Negotiate — FSMA 204 Flow-Down for Co-Manufacturers
A co-manufacturer's traceability obligation is compliance without ownership — so the winning posture is to emit a conformant record once and grant each customer a view, not to join each customer's system.
-
Which of Your Foods FSMA 204 Actually Covers — A Fresh-Department Scoping Guide
The Food Traceability List bites grocers in fresh, deli, and prepared — and scoping which SKUs are covered is a half-day exercise most chains have never formally run.
-
The KDE Gap — What FSMA 204 Asks For vs. What Your Suppliers Will Actually Send
FSMA 204 compliance is decided upstream — the delta between the KDEs the rule reaches and the data your suppliers can emit is the real project, and it is measurable today.
-
FSMA 204's Real Date — Why July 2028 Is a Floor, Not a Deadline
The compliance date rests on an appropriations rider, not a finalized rule, and FDA is soliciting flexibilities — so the rational program builds for the operational return now and treats 2028 as the de-risking floor.
-
You Own the GTIN, So You Own the 2D Decision Nobody Downstream Can Make for You
Which symbology, which URI form, and which AIs travel in the 2D barcode are decisions only the GTIN owner can make — converters and retailers execute them — and the decision window is your artwork cycle, not GS1's calendar.
-
The lane transaction is not an event record
A POS transaction log answers 'what sold.' An EPCIS 2.0 event answers 'what happened to this item.' Sunrise 2027 makes a grocer the reader of a richer barcode — and conflating the two records is the industry's quietest category error. The same scan, shown as both records, field by field.
-
Lot and expiry in the barcode are the brand's choice, not Sunrise's requirement
Sunrise 2027's baseline is GTIN extraction at POS. Lot and expiry are additional AIs a brand may choose to encode — and that choice is where item-instance data your class-grain GDSN feeds structurally cannot carry actually enters your store. One yogurt SKU, walked through all three data grains.
-
Full recall vs. lot-level withdrawal: the economics of knowing which lot went where
A worked cost model at three trace grains — SKU, lot, unit — with every assumption published so you can rerun it with your own numbers.
-
No EDI team, no problem — your commerce stack already emits the documents
A mid-market brand's existing commerce, 3PL and accounting stack already emits orders, shipment confirmations and receipts that map onto the standard transaction vocabulary EDI carries — the real object classes, mapped, with the honest gaps stated.
-
No silent grants — why agent authority must fail closed
The governing rule for agents acting on a supply-chain record — absent a pre-registered notification channel, every human-gated scope fails closed — and why a recorded refusal is worth more than an unrecorded success.
-
Nothing separate to audit: the answer you show a regulator is a view, not an export
When the executive answer is computed over the same catalog the open developer surface queries, there is no reconciliation step to defend — an auditor re-derives the answer instead of trusting the export.
-
One site, many shippers — scoped client views for the multi-tenant warehouse
A multi-client 3PL site runs one physical operation that must appear to each shipper as theirs alone — one event record with per-client scoped views is the only architecture where the same pallet move honestly serves two clients' audits without double-keeping.
-
Party Identity Without a Network: GLNs in the Documents You Already Exchange
The ship-from, ship-to, buyer and seller GLNs already riding in your trading documents are the party context a custody record needs — sourced from paperwork both sides agreed to, not from anyone's network membership.
-
Your Trace Already Has a Vocabulary for Your Paperwork: `po` and `desadv` Are Core CBV
CBV 2.0 defines po, desadv, inv and recadv as first-class business-transaction types on every EPCIS event. Joining purchase orders and ASNs to custody events is core-spec machinery, not a custom integration.
-
The POS refresh math: scheduling capital to land before December 2027
A thousands-of-lanes scanner and POS-software refresh is a multi-quarter capital program. Counted backward from December 31, 2027, six quarters remain — which means the Sunrise 2027 decision is being made this budget cycle whether or not anyone calls it that. The timeline, worked.
-
QR or GS1 DataMatrix: the one-question test your 2D vendor should pass
GS1 DataMatrix clears the lane; QR is what a shopper's phone camera reliably decodes. They are different symbols doing different jobs — and any vendor who conflates them has failed the cheapest diligence question in the whole 2D program. Here is the test, runnable in one email.
-
Reorg-Proof Records: Why the Org Chart Must Never Be Stamped on an Event
Any record that writes an organization ID onto events breaks at the next reorg, acquisition, or divestiture. A worked divestiture shows the failure — and the read-time derivation that survives it.
-
What the public record actually shows about RFID receiving ROI
The four citable artifacts on QSR RFID receiving, quoted with dates and attributions, each with its exact scope stated — and the overclaim each invites, refused in print.
-
Two vocabularies welded at the company grain — and the grain still missing
Schema.org v30.0 formally declared GS1's organization vocabulary equivalent to the web's — one equivalentClass and fifteen equivalentProperty annotations, counted from the ttl — which makes the absent worker-and-agent grain the visible open seam in the stack.
-
Shipped vs. Seen: Reconciling the ASN Against the Events
The receiving question that matters is whether what the ASN said was shipped matches what the events say was seen. One delivery reconciled line by line — two cases short, one substituted lot — and receiving becomes an exception discipline.
-
Show your customer the record without giving them your system
A SharingGrant hands a customer a scoped, verifiable view of exactly the events their mandate covers — revocable, boundary-precise, compiled to native access scopes — retiring both bad options, joining their platform or emailing them exports.
-
Read-Points in Abundance, Attribution Still Impossible: The Shrink Ceiling
Mature item-level RFID has exhausted what read-point counting can say. The marginal information left in every shrink investigation is who handled it — the one dimension the standard has no field for.
-
The Spreadsheet Trace-Back — What a Recall Drill Looks Like With No Supply-Chain Department
For a co-manufactured mid-market brand, trace-back runs on spreadsheets across a plant boundary you don't control — and the fix is an outcome you buy, not a project you staff, because the role that would staff it does not exist.
-
What Sunrise 2027 Actually Expects of a Grocery Lane — and What It Doesn't
Sunrise 2027 is GS1's program, not a law — it expects retail POS to scan and process 2D barcodes and extract the GTIN by end of December 2027, and everything beyond that sentence is someone's sales pitch.
-
Your clock is your customer's clock: Sunrise 2027 one hop removed
A co-manufacturer neither chooses the 2D barcode nor reads it at a lane — it applies what the brand specified and emits what the customer requires. Sunrise exposure arrives mandate-mediated, so the preparation that pays is capability, not compliance. The anatomy of a mandate, worked from the one on public record.
-
Supplier onboarding is an agent's job — N suppliers, N formats, one mandate
Onboarding supplier data feeds — every supplier in a different format, vintage and dialect — is exactly the high-volume, verifiable, low-authority work a deputized integration agent should do inside a bounded mandate, with humans ruling only on the mismatches.
-
The Dock Runs on Temporary Labor, and the Record Has Nowhere to Say So
Custody changes hands at docks staffed substantially by temp and agency workers — the exact place 'who handled this' matters most is the place the standard can answer it least. One cross-dock shift, reconstructed.
-
Running Two Traceability Regimes in One Plant Network
Partial FSMA 204 coverage bifurcates an enterprise brand's operations — covered and uncovered SKUs demand different records from the same lines — and the only sane architecture is one event spine with regime views, not two systems.
-
Verifiable without joining
A traceability record is worth exactly what it proves to someone who never joined your vendor's network — here is the verification a counterparty can run with no account, performed in print.
-
When the Observer Is a Robot: Custody Records for Embodied Agents
Autonomous MHE already performs custody-relevant acts, and no incumbent record has a place for a non-human observer. One pallet move, recorded three ways — only one survives an incident review.
-
`who` Is Not `capturedBy`: The Two Grains Every Custody Record Needs
The observer who performed an event and the account that warrants its capture are different parties with different failure modes. Three worked receiving events show what collapses when one field tries to do both jobs.
-
Why recording an event should cost nothing — and what you actually pay for
Capture priced per event taxes the behavior the record depends on — the durable economics put recording at zero and attach revenue to answers, which is also the only pricing a cost-to-serve buyer can defend.
-
Your next vendor evaluation will be performed by an agent. What will it find?
Procurement-class agents match on machine-checkable facts — pinned spec versions, schema provenance, verifiable event identity — so vendors whose claims exist only as positioning copy are about to become invisible to your own tooling.
No posts for that segment yet.
See your first trace.
Capture workspaces for manufacturers and brands are provisioned from the seat list — we write when your seat is ready.