Barrel and derivative traceability with AI — from incoming crude to the customer's diesel, with no gaps in the genealogy.

The problem

The genealogy exists — but it lives spread across four systems and two people's heads.

Any scheduling, quality or product manager in a refinery knows that traceability exists. The trouble is that it is not in one place:

1. The crude discharge sits in SAP IS-Oil (bill of lading, certificate of origin, lab discharge assay).
2. The pass through the units (atmospheric, vacuum, hydrotreating, reforming) sits in the DCS historian — temperatures, flow rates, conversion, yields.
3. Tank farm movements sit in the tank-farm SCADA and in the dock operator's shift reports, with blending criteria that every plant resolves its own way.
4. Lab test results sit in the LIMS, with a natural lag between sampling and result that the operator at the line never sees in real time.
5. The customer certificate is assembled by hand when the truck or the vessel is loaded, cross-referencing four sources and the shift lead's memory.

The system works — until a customer complains, an auditor asks for a specific period, or an end buyer demands to know which crude their diesel came from. Then the reconstruction begins: one week, two people, a spreadsheet with the batch genealogy hanging off it. Data existing is not the same as data being joined up.

How it fits the IRIS system

iLEAN does not replace SAP IS-Oil, the DCS or the LIMS — it seals the cracks between them.

Every system in the refinery does its own job well. The classic problem is not that one more system is missing, but that between SAP IS-Oil, the DCS, the tank-farm SCADA and the LIMS there are dead zones that the operator fills in with a notebook and the memory of the veteran on the dock. iLEAN is the putty that fills those cracks — it captures real tank measurements, operator shift reports, LIMS test results and DCS movements, cross-references them with SAP IS-Oil, and keeps the genealogy alive at every movement.

Tracer maintains the weighted genealogy movement by movement. Edge captures reality on the dock and in the tank farm. Brain cross-references SAP IS-Oil, the DCS and the LIMS. The person signs the delivery to the customer — always.

The three iLEAN pieces applied to barrel and derivative traceability:

• iLEAN Tracer — the genealogy engine: it maintains the real blend by tank and by movement, with the weighted origin of every component, without the operator having to keep count by hand. When a backward query comes in ("this batch — which crude is in it?"), it answers in seconds instead of in a week.
• iLEAN Edge — terminals on the dock, in the tank farm and at the units that read what the machines never push to the DCS: manual tank gauging, the reading from the old probe, the image of the tanker truck seal, the OCR of the dock operator's paperwork. It works with no network; on the dock that matters.
• iLEAN Brain — cross-references SAP IS-Oil, DCS, LIMS, tank farm data and crude diet to flag discrepancies in the moment (tank measurement and mass balance do not agree, the lab result is outside the expected specification), and prepares the batch data sheet for the product manager, who signs.

See the full IRIS architecture →

Before and after

Classic derivative traceability vs. continuous genealogy with iLEAN

Aspect	Classic traceability (SAP IS-Oil + DCS + notebook)	With iLEAN Tracer + Edge + Brain
Batch genealogy	Rebuilt by hand whenever somebody asks	Maintained at every movement, queryable in seconds
Blend in the tank	"Majority component" or the operator's estimate	Real weighted blend, traced movement by movement
Cross-reference with LIMS	Asynchronous, depends on the product manager's eye	Automatic, alerts if a result does not match expectations
Customer certificate	Template filled in at loading, several manual cross-checks	Data sheet prepared by iLEAN Writer, signed by a human
Auditor / end buyer query	A week of reconstruction by two people	Dossier with the full genealogy in a couple of clicks
Customer complaint	Retrospective investigation, blurry focus	Genealogy and test results available, immediate focus

Impact estimate

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

The block below is an estimate to be validated with the data of your crude mix, your tank farm and your real query rate. We put it forward so the committee has an order of magnitude; we refine it during the diagnostic.

• Mid-sized refinery or petrochemical complex running a multi-crude diet, a multi-product tank farm, exports to several markets and regulated certificates (EN 590 sulfur, FAME, a growing carbon footprint requirement).
• Pilot on one product family (e.g. automotive gasoils): integration with SAP IS-Oil, DCS and LIMS; Tracer maintains the genealogy; Writer prepares the data sheet. First value expected within a few weeks.
• Expected reduction in response time to a customer/auditor query of ≥ 30% and, on large loads, far more (from a week to a few hours). To be validated with your real rate.
• Indicative payback between 4 and 9 months, driven by two levers: avoiding specification penalties and reducing complaints by attaching the data sheet on the day of delivery.

And the product manager's reasonable doubt

"What if the AI gets the blend weighting wrong and I give the customer an origin that is not the right one?" — hallucination is a problem of free generation, not of anchored tasks. In tasks where the AI merely recontextualizes a piece of data from one system into another (reading a movement from the tank-farm SCADA, cross-referencing it with the DCS mass balance, calculating the weighted blend in the tank), the best models brought error below 1.5% ^[1]. And even so, what is critical is never closed out alone: iLEAN proposes the data sheet, the person signs the customer certificate. The three safety rings exist precisely for this.

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