Coder without typing

The problem

The last machine before the customer is still fed by hand.

The coder is the last physical step before the customer. One mistyped digit means part of the run ships with the wrong date or batch. Manual typing remains the standard method because integrating it properly "takes time" maintenance doesn't always have.

• The coder prints expiry date and batch on every bag of pan loaves and buns, right after the slicer and the bagger. It has an API, but it is isolated from the ERP and nobody has had time to connect it.
• Every batch change is typed by hand on its keypad, with the same data already sitting in the Production Order a few meters away. The same information is entered twice, and the second time under pressure.
• On short-shelf-life bread, one wrong digit means part of the run ships with the wrong date or batch — and that ends in a withdrawal from the shelf or a recall.
• Manual typing stays the standard because a proper integration takes time that maintenance rarely has, and the coder project keeps slipping down a list topped by urgent repairs.

How it fits the IRIS system

Connect stitches ERP and coder through their APIs — neither one is replaced.

Connect stitches two modern systems together via API. When the ERP marks a Production Order as started, it automatically publishes date, batch and format to the coder — no manual typing in between.

The date printed on the bag stops depending on a keypad. When the ERP marks the Production Order as started, date, batch and format travel on their own, and what the coder confirms it printed comes back to the system. Marking becomes evidence instead of an assumption.

See the full IRIS architecture →

Before and after

The keypad coder versus the coder stitched to the ERP

Aspect	Today	With iLEAN Connect
Expiry date on the bag	Typed on the coder keypad	Published by the ERP
Batch change between runs	Re-entered by an operator	A tap when the order starts
Wrong date reaching the shelf	A periodic risk	Zero errors of human origin
Shelf-life rule per format	In the operator's head	Applied from the item master
What the coder actually printed	Assumed	Confirmed back to the system
Coder and ERP at the bakery	—	Both kept as they are

Impact estimate

Impact estimate — to be validated with your bakery's 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.

• Estimated payback 3-6 months, one of the fastest in the matrix.
• A drastic cut in recall risk from a wrong date or batch, which on short-shelf-life bread is the most expensive labeling failure: product already spread across hundreds of stores and routes.
• Typing errors of human origin in coding go from periodic to zero, and with them the night calls to pull product from the shelf.
• And maintenance does not have to wait for an integration project slot: this stitching asks for no shutdown window and is set up between production runs.

Estimated payback of 3 to 6 months, drastic reduction in recall risk from wrong date/batch. Estimate to validate.

And the fair question from the production manager

“What if the date the ERP sends is wrong?” — then it is wrong in one place and visible, instead of being retyped by hand on each machine. Building the date from the start time and the format's shelf life is an anchored task, where the best models drop below 1.5% error [1], and the coder returns what it printed so any mismatch with the order raises an alert before the cases are palletized.

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