The ISCC mass balance is not reconciled at month end: it is reconciled tanker by tanker.

The problem

The balance lives in a spreadsheet only one person understands.

In most biodiesel and blending plants, the real circuit is this: the tanker comes in, the weighbridge operator files the delivery note in a folder, someone writes the volume on a sheet, and at month end the sustainability manager sits down to reconcile sustainable inflows and outflows tank by tank. They rebuild densities, hunt for missing sustainability declarations, call the supplier about a loading-point detail and adjust until the balance closes.

1. The close: two or three days a month of an expensive person's time reconstructing movements that already happened, with the information scattered between weighbridge, ERP, supplier email and a folder of scans. When a data point is missing, the truck left three weeks ago and there is no one left to ask.
2. The accounting risk: if an outbound movement was issued as sustainable against a balance that was not actually there, it is not discovered in the moment — it is discovered at the close, or worse, in the audit. Redoing declarations already delivered to a customer is a commercial problem, not just an administrative one.
3. The ISCC audit: the auditor does not audit the tank, they audit the documentary traceability. They pick an outbound movement at random and want to see the inbound movements that support it, with their sustainability declaration, their feedstock and their origin. If that has to be reconstructed, it gets reconstructed over the weeks before they arrive.

The spreadsheet works most of the time. The problem is that it only works backwards — when you can no longer correct anything, only justify it.

How it fits the IRIS system

iLEAN Tracer is the filler between the weighbridge, the tank farm and the ISCC dossier.

The plant already has almost all the data — the problem is that each data point lives on an island: the weighbridge with the weight, the delivery note on paper inside the truck cab, the sustainability declaration in the purchasing inbox, the tank level in the SCADA and the balance in a spreadsheet. iLEAN Tracer does not ask you to change anything. It sits on top and joins what already exists around each tanker movement.

Connect captures the delivery note with the truck still in the yard. Edge cross-references weighbridge and tank. The agent keeps the balance live. The person validates the doubtful movements — never the other way round.

The iLEAN pieces applied to the ISCC mass balance per tanker:

• Connect at weighbridge and yard — the operator photographs the delivery note and the sustainability declaration from the terminal or a phone. The system extracts license plate, supplier, document number, feedstock, volume, density, temperature and loading point, and requests on the spot whatever is missing. A data point requested with the truck in front of you costs thirty seconds; the same data point three weeks later costs a call to the supplier and a week of waiting.
• Edge at the transfer point — cross-references the weighbridge ticket, the flow-meter reading and the destination tank against the open order, and closes the movement with its evidence. If the connection to the ERP drops, the Edge keeps recording locally and syncs when the network comes back: a tank farm cannot depend on coverage to release a truck.
• Mass-balance agent — maintains, for each tank and each period, sustainable inflows, issued outflows, available balance and volume-to-mass conversions using the declared density. It flags the movement that would push the balance negative before the outbound document is issued, proposes the allocation when the tank combines feedstocks with different profiles, and leaves the ISCC dossier compiled with the evidence attached to each entry. Anything doubtful is raised for a person to validate and sign.

See the full IRIS architecture →

Before and after

Mass balance in a spreadsheet vs. a live balance with iLEAN Tracer

Aspect	ISCC mass balance in a spreadsheet	With iLEAN Tracer
Moment of capture	After the fact, at the monthly close	At the movement, with the truck in the yard
Evidence per tanker	Delivery note in a folder, declaration in email	Photo, extracted data and signature on the entry
State of the sustainable balance	Known at month end	Live per tank and period, in the moment
Outbound that would go negative	Caught while reconciling, already issued	Flagged before the document is issued
Volume-to-mass conversion	Density estimated or redone by hand	Density declared per load, documented
ISCC audit preparation	Weeks of documentary reconstruction	Dossier compiled, reverse traceability ready
Operation without network in the yard	n/a	Edge records locally and syncs later

Impact estimate

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

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

• Plant with several ISCC-certified tanks and dozens of tanker movements per week between loading and unloading. Documentary capture at weighbridge and yard + a live balance per tank as a pilot.
• First value expected within a few weeks: the first period close done on movements already captured, with no reconstruction, with the evidence attached to each entry.
• Indicative payback between 4 and 9 months, depending on the number of movements, the hours the monthly close consumes today and the effort spent preparing the ISCC audit.
• Hard levers: monthly close days eliminated by not reconciling the balance in a spreadsheet; weeks of audit preparation replaced by an already-compiled dossier; incomplete sustainability declarations caught in the yard, not at the close.

And the sustainability manager's reasonable doubt

“What if the AI misreads a delivery note and throws my balance off?” — hallucination is a problem of free generation, not of anchored tasks. In tasks where the AI merely extracts a data point that is written on a document and cross-references it with the order and the weighbridge ticket, the best models brought error below 1.5% ^[1]. And even so, what is critical is never decided alone: when the extracted data does not match the order, the weighing or the tank balance, the movement is raised as doubtful and a person validates it before it enters the balance. The agent proposes, the person signs.

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