EU, UK and Swiss nutrition labelling per SKU — a pallet that crosses the border with the wrong label is a rejection, not an incident.

The problem

Three regulatory frameworks for the same SKU — and the label reel does not wait.

A European food company exporting to the United Kingdom and Switzerland no longer makes one yoghurt, it makes three regulatory yoghurts:

1. EU — Regulation 1169/2011: nutrition table per 100 g, 14 allergens highlighted, batch and date in specific formats.
2. UK — after Brexit, the FSA keeps most of the EU framework but adds the voluntary traffic-light scheme and HFSS rules (its own thresholds for fat, sugar and salt) that affect how the product may be advertised.
3. CH — the Swiss FIO: nuances in declarations, possible DE/FR/IT languages, and ingredient-origin requirements that are often stricter for dairy and meat.

The quality manager knows this. The packaging operator does not — their screen only says "SKU 47213, reel A, start". If reel A carried the EU label and the sales order was for the UK, the pallet leaves, reaches Dover, the inspector opens it and it goes back. If the reel passed as Switzerland but did not carry the language of the destination canton, the pallet sits in Swiss customs paying storage. The classic system — a master spreadsheet plus a checklist plus the shift manager's eye — works 95% of the time. That 5% is the rejected pallet.

How it fits into the IRIS system

iLEAN Tracer is not a new printer — it is the filler between your ERP, your packaging machine and each country's rules in force.

Multi-country labelling is not a printing problem: it is a problem of cross-checking data that lives on separate islands. The recipe is in the ERP, the destination in the sales order, the rules in a regulatory PDF, the reel in the operator's hands. iLEAN Tracer is the piece of the IRIS category that seals that gap — without asking you to change ERP, MES, packaging machine or printer. It is the filler, not the wall.

Edge sees the reel. Connect reads recipe and destination wherever they live — ERP, sales order, the exporter's spreadsheet. The agent composes the correct label for that country. A person signs off the reel change — never the other way round.

How the work is shared between the iLEAN pieces in this case:

• Edge — a terminal with vision at the packaging machine. It reads the loaded reel (artwork, language, batch) in seconds. If the reel does not match what the agent expected for the SKU on the order, it triggers the traffic light and holds the line before the first sealed pack. It works with no network.
• Connect — captures the batch recipe from the modern ERP, from the vertical MES or from the R&D manager's spreadsheet. It also captures the destination on the sales order and the exporter's notices (an email from the UK importer with the new HFSS rule, a message from the Swiss distributor) at second zero.
• Tracer agent — cross-checks recipe + destination + country rules in force, composes the correct label and sends it to the thermal printer with the ZPL or the template it already uses. If there is a deviation (the loaded reel is for another country), it holds the change and alerts the quality manager over the channel they use. The person validates and signs off.

See the full IRIS architecture →

Before and after

Manual multi-country labelling vs. segmented labelling with iLEAN Tracer

Aspect	Manual labelling + master spreadsheet	With iLEAN Tracer
Source of the recipe	R&D manager's spreadsheet, copy on the shop floor	Batch recipe unified at second zero
Source of the destination	Sales rep's order, on paper in the office	Sales order read by Connect, cross-checked with the SKU
Per-country rules	Regulatory PDF in the quality manager's drawer	Versioned library of EU/UK/CH templates
Reel change	Operator loads the right one, shift manager signs off	Tracer composes, Edge verifies, a person signs off
Error detection	Inspector at the UK border, Swiss customs	Before the first sealed pack, on the line
Regulatory update	Manual rollout of the new PDF to the plant	New template uploaded and promoted without stopping the line

Impact estimate

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

The block below is an estimate to be validated with your plant's actual data. We set it out so the committee has an order of magnitude; we refine it during the diagnostic.

• Food plant with several packaging lines and an EU/UK/CH mix, an SKU mix with variable recipes and different reels per destination.
• Tracer pilot on one line (Edge at the packaging machine + integration with the ERP sales order + library of per-country templates). First value expected within a few weeks.
• Indicative payback between 4 and 9 months, depending on the documented frequency of pallets rejected at the border and the average cost of return/rework.
• Expected reduction in the cost of labelling non-conformities of ≥ 30% in the first year — a single pallet rejected at the UK border usually covers several months of licence.

And the regulatory manager's reasonable doubt

"What if the AI picks the wrong template and sends a UK label to the EU pallet?" — hallucination is a problem of free generation, not of anchored tasks. Composing a label from a template plus ERP data is exactly the kind of anchored task in which the best models brought error below 1.5% ^[1]. And even so, nothing critical is decided alone: Tracer holds the reel change and a person signs off. The three safety rings exist precisely for this.

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