The coder and labeler integrated with iLEAN — the operator stops typing the SKU change.

The problem

The modern coder is already on the line — but disconnected from the ERP.

It happens in a great many breadstick and crispbread factories: the investment in the Markem-Imaje, Domino or Videojet coder is already made. The machine has an Ethernet port, speaks OPC UA, exposes a REST API. And still, at every SKU change, the operator climbs onto the line, opens the coder's panel and types by hand the new batch, the new date, the new expiry, the new SKU's allergens.

• One wrong digit in the batch — the operator types 251207 when it was 251208, and 1,200 bags print with a batch that does not exist in the ERP. Traceability broken.
• An incorrect expiry — the previous SKU's shelf life (90 days) is applied to the new SKU (60 days, because it carries seeds and oil). Product on the shelf with an impossible date.
• The previous SKU's allergens — the line switches from classic to sesame breadsticks and the label still declares "sesame-free". A recall.
• The previous batch's label reused — the new SKU's first box goes out with the previous one's label because nobody validated the first print.

The production manager knows it, and keeps an SKU-change checklist pinned to the wall. The checklist works 99% of the time. That 1% is what reaches RASFF — and one wrong digit in the date is a recall, not an incident.

How it fits the IRIS system

iLEAN Connect integrated mode — the coder and labeler become just another ERP peripheral.

The modern coder does not need an operator to type into it: it needs a layer that speaks the ERP's language to it. iLEAN Connect in integrated mode is that layer. It is the putty covering the dead zone between the ERP's production order and the job the coder prints — without asking you to change the coder, the labeler or the ERP.

The order changes in the ERP. Connect translates. The coder prints the correct SKU from the first bag. The operator signs — they do not type.

How the iLEAN pieces are laid out in a breadstick and crispbread factory with a modern coder/labeler:

• Connect (integrated mode) — talks to the coder over its native interface: OPC UA with Markem-Imaje series 9, Ethernet/IP with Domino Ax-Series, REST/MQTT with Videojet 1860, IO-Link for peripheral actuators. On the other end it reads the ERP's production order (SAP B1, Sage, A3, Navision, Odoo, Datisa) and the SKU's template. It translates and loads the job in milliseconds. The coder stops having a panel to touch.
• Agent — governs the SKU change as a process, not a keyboard. When the order changes, the agent: validates that the format is approved for that SKU, recalculates the expiry per the defined shelf life, verifies that the recipe's allergens match those declared on the label template, holds the first print for visual validation and, if everything adds up, releases the line. If something does not, it holds and alerts the quality manager through whichever channel they use. The person signs — the line does not start on its own.
• Edge (optional, recommended) — a terminal with machine vision over the coder's exit that reads the first printed bag and compares what printed against what was expected. If the modern coder already carries its own vision camera, Connect connects to it; if not, an Edge is mounted. It closes the cycle: the guarantee that what leaves the head matches the order.

See the full IRIS architecture →

Before and after

A hand-typed SKU change vs. an SKU change triggered by the order

Aspect	A coder typed by hand	With iLEAN Connect integrated mode
The SKU change's trigger	The operator reads the order and types at the panel	The ERP's order triggers the job automatically
SKU change time	5-12 minutes per change (climb, type, verify, climb down)	Seconds — the line stops only for the physical format change
Batch transcription error	Inherent to manual typing	Removed — the data travels hands-free
The expiry	The operator calculates or types the date	Calculated by the SKU's template per shelf life
Allergens on the label	A generic template updated "when there is time"	Synchronized with the active SKU's recipe in the ERP
First label validation	The operator's glance, no record	Vision on the first bag + the manager's signature
The file for the IFS/BRC auditor	Rebuilt by hand, weeks	A per-batch dossier, automatic, with the first bag's photo
A reel label format change	The designer edits and resends to the line lead	A per-SKU template, triggered by the order

Impact estimate

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

The block below is an estimate to be validated against your plant's actual data. We put it forward so the committee has an order of magnitude; we refine it during the assessment.

• Breadstick and crispbread factory with 2-4 packing lines, a modern thermal coder (Markem-Imaje series 9 / Domino Ax / Videojet 1860) and/or a reel labeler, 15-30 active rotating SKUs.
• Connect integrated mode deployment on one pilot line. First value expected within a few weeks: the SKU change goes from typing to a job triggered by the order.
• A reduction in SKU change time in the order of ≥ 50% (the stretch eliminated by climbing to the panel and typing).
• Coding errors (batch, date, allergen) trending to zero on the transcription side — the residual remains only in errors in the order itself, caught at the first-label validation.
• Indicative payback between 4 and 9 months, depending on monthly SKU change frequency, the average cost of a stopped line minute and the historical exposure to labeling recalls.
• The hard lever, as in the other food landings, is a single recall avoided: product recovered from the shelf, reverse transport, destruction, brand damage. One recall pays for the pilot.

And the fair question from the production lead

"What if Connect errs translating the order into a coder job, and the line prints wrong at production speed?" — a fair question. Two complementary answers. First, technical: Connect in integrated mode does not "interpret" the order, it recontextualizes it from one format to another. Hallucination is a problem of free generation, not of anchored tasks; in tasks where the AI limits itself to moving a data point from one system to another (the ERP's order into the coder's "batch" field), the best models brought the error below 1.5% ^[1]. Second, architectural: the first bag of every SKU is validated with vision and the manager's signature before releasing the line. The three safety rings are there precisely for this — the person signs the critical.

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