GS1 SSCC label on a multi-SKU mixed pallet — one error is not an incident, it is a pallet rejected at the dock.

The problem

A mixed pallet is built twice — once in the WMS and once on the floor. They are almost never the same.

A correct GS1 SSCC label on a mixed pallet is the crossing of four data points that live in four different places:

1. What the store ordered — the retailer order in the WMS, with the expected GTINs, quantities and batches.
2. What was actually picked — the real picking on the floor, where the picker swaps, changes, breaks or adjusts cases depending on what is on the rack.
3. The available SSCC range — the SSCCs the company has not consumed yet, managed in a spreadsheet or in an ERP module.
4. The retailer ASN format — EDIFACT DESADV for European grocery chains, X12 856 in the US, proprietary XML for other retailers and marketplaces, each with its own mapping.

The operator knows all this, but cannot keep the four in sync by hand. They try with a scanner and a spreadsheet, and most of the time it works. The classic system works 99% of the time. That 1% is the pallet rejected at the dock — the truck comes back, the retailer files a claim, the SSCC stays «dirty» in the national database, and the cost of the incident is far higher than the cost of the transport.

How it fits into the IRIS system

iLEAN Tracer does not replace your WMS — it seals the cracks between WMS, picking, SSCC range and retailer ASN.

The mixed pallet problem is not a lack of information: it is information living on four islands that gets reconciled too late, already at the dock, when the truck is about to leave. iLEAN Tracer acts as the filler that closes those gaps, without asking you to change your WMS, your TMS, or the flow of your platform.

Edge sees every case as it goes onto the pallet. Connect captures the retailer order and the ASN formats. Tracer assigns the SSCC against the GS1 range and builds the list. The agent reconciles and holds the closure if it does not add up. The person signs — never the other way round.

The iLEAN pieces applied to the multi-SKU mixed pallet for retail:

• Tracer (functional layer over Edge + Agents) — reserves the SSCC from the corporate GS1 range when the pallet is opened, builds the content list case by case, and at closure reconciles GTIN, batch, quantity and expiry date against the store order. Every pallet carries its number, and every number carries its evidence.
• Edge — terminal with machine vision (CNN) over the pallet build area. Reads the GTIN of every case as it goes up, validates batch and expiry when they come in the GS1-128, and triggers a light or a rejector if the content does not match what was expected. It works with no network. If the platform loses its connection to the WMS, Edge keeps reading and reconciling with its local buffer.
• Connect — captures the order of every retailer whether it arrives via modern EDI, via an email from the buyer, or via a customer web portal. And it also captures what arrives from outside (a damage alert on the rack that changes the picking, a transport message about a dock change) at second zero.
• Agent — cross-checks the real pallet content, the retailer order, the available SSCC range and the rules of each ASN format. If there is a deviation, it does not wait for the dock: it holds the pallet closure and alerts the shift supervisor through whatever channel they use. The person validates and signs; the SSCC is not printed without reconciliation.

See the full IRIS architecture →

Before and after

SSCC label rewritten at the dock vs. SSCC built on the line with iLEAN Tracer

Aspect	WMS + scanner + spreadsheet	With iLEAN Tracer + Edge + Agent
SSCC assignment	Reserved when the pallet is closed, sometimes duplicated	Reserved when the pallet is opened, against the GS1 range, no duplicates
Reconciliation with real picking	Manual rewrite when a difference is spotted	Case by case while the pallet is being built
Swap due to damage or change	The picker adjusts, the system never finds out	Edge sees the swap, the agent updates the list
ASN to the retailer	Generated from the theoretical data, frequent errors	Generated from the real data, validated against retailer rules
Rejection at the dock	Truck comes back, cost of return transport	Closure held before the truck is sealed
Batch→store traceability	Rebuilt by hand if there is an incident	Evidence pack per SSCC, automatic, with a photo of the pallet

Impact estimate

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

The block below is an estimate to be validated with the actual data of your platform. We put it forward so the committee has an order of magnitude; we refine it during the diagnostic.

• Logistics platform with 3-5 retailers, several thousand pallets/day, mostly multi-SKU mixed, heterogeneous ASN (DESADV + proprietary ASN).
• Tracer pilot over one picking area (Edge cameras over the pallet build area + scanners + closure actuator + integration with the WMS and the ASN connectors). First value expected within a few weeks.
• Expected reduction in pallets rejected at the dock because of an SSCC with wrong content: ≥ 30% over your platform baseline incident rate.
• Indicative payback between 4 and 9 months, depending on the frequency of documented rejections and the average cost of a return from the retailer dock.
• The hard lever is the compound cost of a rejected pallet: return transport, rework, retailer penalty, a «dirty» SSCC in the national system, and reputation with the customer. Every rejection avoided goes straight to margin.

And the reasonable doubt of the operations director

«What if the AI gets it wrong and lets a pallet out with the wrong list?» — hallucination is a problem of free generation, not of anchored tasks. In tasks where the AI is limited to recontextualising a data point from one system into another (reading the GTIN and comparing it against the expected order), the best models brought the error below 1.5% ^[1]. And even so, what is critical is never decided alone: iLEAN Tracer holds the pallet closure and the person signs. The three safety rings are there for exactly this.

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