Zero grams lost: the resin balance, live, from silo to pallet

The problem

A slow hemorrhage, not a one-off incident.

Four pains that are really one. A resin inventory that is hard to govern, whose availability and consistent quality are a condition of continuous production. A volatile raw material cost that hits profitability directly. Stock levels trapped between the shortage that stops the line and the excess that ties up cash. And multi-line scheduling that breaks every time a customer changes their mind. Everything lost along the way — preform overweight, changeover purge, start-up queue, scrap from defects, held batches — appears aggregated in a month-end settlement, with no root cause and nobody to ask. By then there is nothing left to correct. Overweight deserves a separate mention because it is the quietest of all: one extra gram per bottle, at the scale of a converting plant, is tonnes of resin a year that nobody books as a loss because the product ships conforming.

• Resin is between sixty and seventy percent of the product cost, and today nobody can close live the balance between the kilos entering the silo and the sellable containers leaving the dock.
• Four pains that are really one: resin inventory that is hard to govern, volatile raw material cost, stock caught between the shortage that stops the line and the excess that ties up cash, and multi-line planning that breaks every time a customer changes their mind.
• Everything lost along the way — preform overweight, changeover purging, start-up tail, scrap — is settled at month end, aggregated and with no cause.
• What is settled in aggregate cannot be attacked: there is nobody to attribute it to.

How it fits the IRIS system

Four rings — no new pieces, it is all of the above orchestrated.

Connect + Edge + Agents + JIDOKA AI + SMED AI, coordinated into four rings. There are no new pieces: it is everything above, orchestrated against the capital pain. Ring 1 — Inlet. Every resin batch enters traced by a photo of the delivery note and the certificate of analysis, with its intrinsic viscosity, its moisture and its recycled content, linked to a specific silo. Drying conditions are read off the panel without touching the equipment. You know what material you have and where it is. Ring 2 — Transformation. The real start-up recipe is recorded; preform and bottle weight is validated inline; the Edge camera attributes every defect to its mold and cavity; the mold change does not start without evidence of a complete purge and weight in window. JIDOKA AI stops the line on a sustained weight deviation. Ring 3 — Outlet. The pallet label prints from the ERP with reference, batch, mold, count and declared recycled content. The kilo that came in through the silo and the container leaving on the pallet are linked end to end. Ring 4 — Decision. Agents cross-references the live balance against the production schedule and the order changes arriving by email or WhatsApp, and proposes which line, which mold, which sequence and which silo coverage. SMED AI groups changeovers by mold family so the resequencing does not eat in changeovers what it gains in service. The audit evidence pack comes out as a by-product. Humans in command. The four rings propose and evidence. Batch release and the replanning decision stay human.

One percentage point of improvement in the ratio of kilos of resin per thousand sellable containers is the largest economic lever a converting plant has. Well above any headcount saving, and that is why it is the capital pain the complete system aims at.

See the full IRIS architecture →

Before and after

The four rings, and what each one closes

Ring	Which cases it draws on	What it closes
1 · Input	Resin receiving (7) + dryer (2)	Kilos and quality entering, tied to silo and lot
2 · Process	Start-up (1) + laboratory (6) + validation (5)	Real recipe, target weight and signed deviation
3 · Line	Vision (9) + mold changeover (10)	Overweight, scrap and purging, stopped where they happen
4 · Output	Pallet label (8) + evidence pack (11)	Sellable containers counted and traced per pallet

Impact estimate

Estimated impact — to validate 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.

• In a converting plant, one percentage point of improvement in kilos of resin per thousand sellable containers is the largest economic lever that exists, well above any headcount saving.
• From an aggregated monthly settlement to a live figure with root cause by mold, cavity and resin lot.
• Preform overweight: from an invisible loss to a deviation stopped on the line.
• Replanning after a customer change against the real state of silo and line, not against yesterday's plan.

In a PET converting plant, one percentage point of improvement in the ratio of resin kilos per thousand sellable containers is the single largest economic lever there is, far above any headcount saving. Estimated payback 6-12 months. Estimated scrap and purge reduction of at least 30% in the first cycles, plus the OEE points recovered on mold changes and the working capital freed by silo coverage matched to the real schedule. *Estimate to be validated* against the real kilos-per-thousand ratio over the last twelve months.

And the fair question from the production manager

«Do we have to deploy all twelve cases to get the balance?» — no, and framing it that way would be the way never to start. The rings are built in layers and each case pays on its own from the first month. With resin receiving and line start-up you already have all of ring 1 and half of ring 2, which is where half the leak sits, without committing to a long project.

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