Individual traceability for almadraba bluefin tuna — one lost tag is one fish out of the market.

The problem

The tag, the GBYP and the eBCD travel on three different media — and are reunited by hand.

Correct ICCAT traceability on almadraba bluefin tuna is the cross-reference of three data points that almost never reach the plant together:

1. The physical tag on the fish — a numbered tag that arrives with the piece from the auction market.
2. The GBYP number assigned during biological sampling — it comes on a separate sheet of paper or in a photo on the skipper's phone.
3. The eBCD (electronic Bluefin Catch Document) — the document that travels to Brussels and to the final buyer, signed by the authority and by the operator.

In practice, the plant runs on a provisional paper sheet until the good eBCD arrives, and then somebody retypes the references. The changeover from one fish to the next on the cutting line — the moment when loins, ventrescas and descargados are separated — is exactly when it is easiest for a cut to end up linked to the wrong tag. The classic system (operator + paper sheet + a trained eye) works 99% of the time — and that 1% is what reaches the authority as an untraced piece. Every incident is a kilo of the most expensive premium product in the sea taken off the market.

How it fits the IRIS system

iLEAN Tracer does not add a fourth medium — it seals the cracks between the three that already exist.

The problem with ICCAT traceability is not a lack of information: it is information that lives in islands (physical tag, GBYP on paper, eBCD in a PDF) that does not reach the fish as one unit at the moment of cutting. iLEAN acts as the putty that fills those gaps, without asking you to change your flow with the authority or your cutting line.

Edge reads the tag before the first cut. Connect captures the eBCD whether it comes by email, as a PDF or as a photo of the paper. The agent joins the three and, if something does not add up, holds the fish. The person signs — never the other way round.

The three iLEAN pieces applied to piece-by-piece bluefin tuna traceability:

• Edge — a terminal with machine vision (CNN) over the infeed belt of the cutting line. It reads the ICCAT tag and the GBYP mark, validates the format and triggers a traffic light or ejector if it does not match the eBCD. It works with no network. If the plant loses WiFi, Edge keeps reading and holding fish whose tag has not been validated.
• Connect — captures the eBCD through whatever channel it comes on (an email from the authority, an attached PDF, a photo of the signed paper), the GBYP from the biological sampling, and the buyer's changes on formats (sushi cubes for the US, whole loins for a Japanese wholesaler) at second zero, without anyone forwarding anything.
• Agent — cross-references tag, GBYP, eBCD, the photo of the parent fish and every derived cut. It generates the file per fish that the auditor asks for and that today is rebuilt weeks later. If there is a deviation, it holds the piece and alerts the person in charge through whatever channel they use. That person signs; the line does not restart on its own.

See the full IRIS architecture →

Before and after

Paper-sheet traceability vs. cross-referenced Tracer with iLEAN

Aspect	Provisional paper sheet + retyping	With iLEAN Tracer (Edge + Connect + Agent)
Reading the ICCAT tag	The operator reads and types it in — human error	Edge reads it with vision and validates the format
eBCD entry	Paper/PDF that lags behind — provisional sheet	Connect captures the document at second zero
Parent fish ↔ cuts link	Manual chain at every fish changeover	Inherited automatically, validated by photo
Detecting a misread tag	In a later audit or in a claim	Before the first cut, inline
Operation with no network	n/a	Edge keeps running on the light of its own panel
File for the ICCAT auditor	Rebuilt by hand, weeks of work	Automatic per-fish dossier, with photo

Impact estimate

Impact estimate for your plant — to be validated with your numbers.

The block below is an estimate to be validated with the specific data of your plant. We put it forward so the committee has an order of magnitude; we refine it during the diagnostic.

• Cutting plant with one infeed line and separation into loin / ventresca / descargado / akami / sushi cubes.
• Tracer pilot at the cutting infeed (camera over the tag + integration with the eBCD via Connect + file agent). First value expected within a few weeks.
• Expected reduction in traceability incidents ≥ 30% against the plant's documented pattern — a conservative floor.
• Indicative payback between 4 and 9 months, depending on season volume and average selling price. A single fish pulled for incomplete traceability pays for a good part of the pilot.

And the quality manager's reasonable doubt

“What if the AI misreads a tag and links a cut to the wrong fish?” — hallucination is a problem of free generation, not of anchored tasks. In tasks where the AI merely recontextualizes a piece of data from one medium to another (reading a tag, validating the format and cross-referencing it with a PDF), the best models brought error below 1.5% ^[1]. And even so, what is critical is never decided alone: iLEAN holds the fish and the person signs. The three safety rings exist precisely for this.

^[1] OpenAI paper “Why Language Models Hallucinate”, 2025 — on the reliability of AI in anchored tasks.