Four coordinated rings against scrap and the customer line stoppage
The capital pain of a plant that paints and assembles for an OEM customer is double and linked: scrap eats the margin, and when FTQ drops the sequence tightens and the ghost of the customer line stoppage appears. The iLEAN flagship coordinates the 11 previous cases into 4 verification rings that cross-check each other — and if they do not add up, the plant knows before the rack goes on the truck.
FTQ and delivery performance define the customer relationship — and are managed with systems that do not talk.
The two metrics an OEM customer really watches are FTQ and adherence to the delivery plan. Everything else is internal. The problem is that both are managed today with loose pieces — a panel, a spreadsheet, a sheet of paper, an EDI — that never cross-reference each other:
- The reaction margin is hours, not minutes — when FTQ degrades, the figure arrives once the shift has already produced. You correct the next shift, not the rack currently going through the oven.
- Each system holds a share of the truth — the oven knows its curve, the lab knows its film thicknesses, inspection knows its rejects and the EDI knows the sequence. None of them knows whether the whole thing adds up, because nobody cross-references them in real time.
- The risk concentrates at the end — the moment an error stops being cheap is when the rack goes on the truck. Everything not caught before that gets caught at the customer, with the cost multiplied.
The customer relationship ends up being managed on effort and craft, with the permanent feeling that the system is not helping.
Four rings that cross-check — and a plant that finds out before the customer does.
The flagship is not one more case: it is the coordination of the previous eleven into four verification rings that check each other. Each ring contributes a share of the truth, and the value appears where they cross: a discrepancy between two rings is an alert before it turns into scrap or an incident.
Ring 1, real batch conditions. Ring 2, color changeover with evidence. Ring 3, 100% Edge inspection with JIDOKA AI. Ring 4, sequence stitched to the EDI with an evidence pack behind every rack. If they do not add up against each other, the plant knows before loading the truck.
The four rings, and what each one contributes:
- Ring 1 — Real batch conditions: the order startup sheet, the oven curve and the lab measurements define the conditions under which each batch was actually painted. Without this ring, the rest of the data has nothing to be explained against.
- Ring 2 — Color changeover with evidence: visual validation of purge, booth, bells and first part guarantees the batch starts clean. It is the ring that stops a whole batch being born condemned.
- Ring 3 — 100% Edge inspection: every part inspected at the oven exit, with the defect classified by type and position. If the pattern spikes, JIDOKA AI alerts and blocks before producing scrap in bulk, instead of counting it afterwards.
- Ring 4 — Sequence: labeling stitched to the EDI and customer alerts read within seconds ensure the rack going on the truck carries the right reference, color and position — with the evidence pack behind it, ready for any complaint.
- SMED AI and people in command: color changeovers are accelerated so that verifying does not cost capacity, and the earpiece and the tablet keep operators and supervisors deciding with the information in front of them. The system proposes and holds; people sign.
FTQ managed after the fact vs. four rings cross-checking in real time
| Aspect | Loose systems and paperwork | With the iLEAN flagship (4 rings) |
|---|---|---|
| Reaction margin | Hours: the next shift | Minutes: the rack going through now |
| Cross-check of conditions, defect and lot | Manual, if done at all | Automatic and continuous |
| Detection of the sequence incident | On the customer line | In the plant, before the truck |
| Paint scrap | An accepted structural cost | Reduction in the order of 30% or more |
| FTQ trend | Measured, not explained | Measured and explained by cause |
| Evidence for audit and complaints | Reconstructed over days | Generated at source, downloadable |
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.
- Tier 1 plant painting and assembling components with sequenced delivery to an OEM customer, where FTQ and plan adherence define the commercial relationship.
- Phased deployment: start with the shortest-payback Connect cases, raise rings 1 and 2, add Edge inspection and close with the sequence ring and the evidence pack.
- Sustained FTQ improvement — the order of improvement this sub-sector has shown to be reachable through process stabilization is counted in tens of points — and scrap reduction in the order of 30% or more. Estimate to be validated against your starting ratios.
- Indicative payback for the whole of 6 to 12 months, with the Connect cases delivering return from the first months while the Edge rings mature.
- And the asymmetric protection that does not fit on a spreadsheet: a single sequence incident avoided — penalty, urgent freight and commercial damage — comfortably exceeds the cost of the project.
And the fair question from plant management
"Is this not too much of a project to start with?" — you do not start here. The flagship is the destination, not the first step: you get there in phases, starting with the shortest-payback Connect cases, which finance the next ones. And on the reliability of the whole: everything the AI does across these four rings is an anchored task — reading a known figure, classifying against a trained pattern, cross-referencing two records — where the best models brought the error below 1.5% [1], with a person signing every critical decision.
[1] OpenAI paper "Why Language Models Hallucinate", 2025 — on the reliability of AI in anchored tasks.
What people ask about the flagship paint and assembly deployment
Do all twelve cases have to be deployed at once?
No, and it would not be advisable. The flagship describes the destination, not the starting point. The usual sequence begins with the Connect cases — startup sheet, oven panel, lab, receiving, labeling — because they have the shortest payback and require no investment in vision hardware: between three and eight months each. That return finances the next phase, which is the Edge rings for inspection and color changeover. The evidence pack closes at the end, once there are recorded events to order. Each phase delivers value on its own, even if the plant chose to stop there.
What does it actually mean that the rings "cross-check"?
That the information from each ring serves to verify the others. A concrete example: if ring 3 detects a spike in craters on a rack, ring 1 can say which basecoat lot and which oven curve painted it, and ring 2 whether the previous color changeover came out clean. None of those three figures explains anything on its own; cross-referenced, they point at the cause. And the other way around: if ring 4 detects that a rack's sequence does not match the release, the alert fires before loading the truck, which is the only moment when correcting is still cheap.
Can FTQ really be raised by tens of points?
That is the order of magnitude this sub-sector has shown to be reachable when the process is stabilized, and it is worth understanding where it comes from: not from inspecting more, but from explaining the defects. Today FTQ is measured but not explained, because the axes — lot, color, position, oven curve, lab conditions — that would turn a list of rejects into a map of causes are missing. When those axes exist and cross-reference each other, patterns appear that no sampling inspection could show. Even so, it remains an estimate to be validated: the real distance depends on each plant's starting ratio.
What about the risk of a customer line stoppage?
It is the risk that justifies the project on its own, and it is attacked from two sides. On one, ring 4 removes sequence errors of human origin by stitching the label to the active release and checking before printing. On the other, customer alerts are read within seconds instead of being discovered in an email hours later, which turns a crisis into an adjustment. The economic logic is asymmetric: this is not a recurring saving but the fact that a single incident avoided — penalty, premium freight and wear on a relationship that took years — comfortably exceeds the cost of the deployment.
What role is left for people in a plant like this?
Deciding — which is the role they often cannot exercise well today for lack of information. The principle is humans in command: the system proposes, holds and alerts, but does not execute what is critical. Quality signs off a color start; a person decides the fate of a pulled part; a manager validates the ERP transfer. What changes is not who is in charge, but what they are in charge with: the earpiece brings the alert to the operator whose hands are full, and the tablet puts the right instruction in front of the supervisor. The AI removes the mechanical work and gives judgment back to those who have it.
Book the masterclass: we show you the four rings running on your own flow.
We work on your plant's real data, not ours. We walk your paint and assembly flow and put numbers on every ring. Assessment with no commitment.
See how we apply it in your plant — masterclass on your own flow ‹ See all 12 paint & assembly cases See automotive