The traceability marker, stitched to the ERP by API

The problem

The API exists. The integration never reached the top of the list.

Double entry between ERP/MES and line equipment is the classic silent error of a machining shop. It is not a technical problem — the API exists — it is a priority one: doing the integration properly never reached the top of the list. Every manual change is a chance for a mismatch between what the system says and what the part carries. And on a brake disc or a caliper, false traceability is not a clerical error: it is a containment you cannot bound and an audit non-conformity waiting to happen. It is usually the customer who finds it when scanning the DataMatrix, which is the worst possible place to discover it.

• The marker that engraves heat number and DataMatrix on every disc and caliper is five to fifteen years old and has a documented API. And somebody still types every part-number change into its screen, copying what the ERP already holds.
• Every manual change is a chance of a mismatch between what the system says and what the part carries. On a safety part, false traceability is a containment you cannot bound.
• It is usually the customer who finds it, scanning the DataMatrix at incoming inspection — the worst possible place to discover it, and the one that goes straight into the supplier scorecard.
• It is not a technical problem, it is a priority one: doing the integration properly never made it above the daily fires of the machining shop.

How it fits the IRIS system

Connect in integration mode — two modern systems stitched, neither replaced.

Connect in integration mode, stitching two modern systems without replacing either:

The active order publishes part number, heat number and date; the marker receives them through its API at every change; and what was engraved comes back to the central system as an explicit confirmation. If the marker does not accept the change, the plant knows before production starts, not a thousand parts later. Marking stops being an assumption and becomes evidence. In a shop that already runs slower than the molding line, the minutes recovered at every part change are minutes on the critical path of the whole plant, and the mismatch that can no longer happen is the one the customer used to find first.

• The active ERP/MES order publishes part number, heat number and date.
• The marker receives them through its API at every part change, with no manual intervention.
• The marking result comes back to the central system as an explicit confirmation, not an assumption.
• If the marker does not accept the change, the system says so before production starts, not a thousand parts later.

See the full IRIS architecture →

Before and after

The isolated marker versus the stitched marker

Aspect	Today	With iLEAN Connect
Manual entries per part change	One, on the marker screen	Zero
Mismatch between order and marking	Found by the customer	Cannot happen by design
What was actually engraved	Assumed	Confirmed and returned to the order
Marker refusing a change	Discovered a thousand parts later	Flagged before production starts
Time at every part change	Minutes on the critical path	A tap in the ERP
Either system	—	Neither is replaced

from one manual entry per part change to zero · from mismatches the customer discovers to mismatches that cannot happen by design · from marking assumed correct to marking confirmed against the order.

Impact estimate

Impact estimate — to be validated with your 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 5-10 months.
• Time recovered at every part change in the machining shop, which is the plant's slowest stage and where every minute is a part not machined.
• The item that dominates the calculation: the customer claim avoided for a lot marked with the wrong heat or part number.
• And marking that becomes evidence for the audit instead of an assumption.

estimated payback 5 to 10 months, combining the time recovered at every part change and the customer claim avoided, which is the item that dominates the calculation. *Estimate to validate*.

And the fair question from the production manager

“Isn't a classic ERP-to-machine integration cleaner?” — it is, the day the project reaches the top of the IT list, and that is precisely the problem. This stitching uses the API the marker already exposes, replaces neither system and asks for no shutdown window. Where a value has to be read from a machine response, that is an anchored task where the best models drop below 1.5% error [1], and the marking confirmation itself is a hard message from the machine, not a model's guess. If the marker answers nothing, that silence is flagged too.

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