Cut-flower reefer: the auction price is decided in the cracks of the cold chain.

The problem

The flower arrives beautiful and still loses the price.

A reefer of rose, lisianthus or gypsophila travelling from an African or South American farm to a European auction passes, between the farm cooler and the buyer's trolley, through at least four custodies: farm → origin airport handler → carrier (sea or air) → destination airport handler → auction platform. The cold chain does not break in one of them — it erodes at every hop:

1. Farm: precooling takes as long as it takes — it depends on how loaded the cooler was and on the ambient temperature that day. The operator writes it down on a sheet.
2. Airport handler: ULD or pallet waiting for the plane or the ship. A six-hour wait outside the 0.5–4°C range eats vase life.
3. Reefer datalogger: it emits the curve at the end, as a PDF or a Cold Chain EDI — it arrives when the batch is already at the auction.
4. Auction: the buyer projects the vase life they will offer the florist or the retailer. They penalise with price, not with a claim.

The exporter knows this, but cannot be in four places at once. When the buyer penalises, no email arrives — the auction clock price arrives, and there is no renegotiation. The classic system works 95% of the time. That 5% is what erodes the margin batch after batch — and an eroded margin in cut flowers is the difference between exporting next year or not.

How it fits into the IRIS system

iLEAN Tracer does not add one more system — it seals the cracks between the four custodies already involved.

The vase-life problem is not a lack of information: it is information living on four islands which, at the critical moment (the batch entering the auction trolley), does not reach the person who sets the price in time. iLEAN Tracer acts as the filler that closes those gaps, without asking the carrier, the handler or the auction to change their software.

Edge reads the MPS label, the reefer seal and the batch QR. Connect captures the datalogger PDF, the handler's message and the airline EDI. The agent projects vase life before the clock. The person validates.

The iLEAN pieces applied to a cut-flower reefer heading to auction:

• Edge — a terminal with computer vision (CNN) at the farm cooler and at auction reception. It reads the CC label, the MPS-Florimark code, the batch QR, the reefer seal and the photo of the datalogger display. It works without a network: at an airport handler with no stable WiFi, the read-and-sign cycle carries on.
• Connect — captures everything arriving through loose channels: the reefer datalogger PDF, the carrier's Cold Chain EDI, the airline AWB, the handler's message at the airport, the customs agent's email. Every message enters the batch custody chain at second zero, with nobody forwarding anything.
• Agent — crosses farm precooling, the reefer curve, route events and the species/variety of the batch. It projects the expected vase life in the vase before the batch enters the auction. If the projection falls below the threshold agreed with the buyer, it holds the batch for human review and alerts the quality manager. The person validates and signs; the batch does not go to the clock on its own.

See the full IRIS architecture →

Before and after

A blind cold chain vs. a cold chain signed with iLEAN Tracer

Aspect	Farm sheet + datalogger PDF	With iLEAN Tracer
Farm precooling	In/out times on a sheet	Real curve linked to the batch and the MPS
Reefer curve in transit	PDF after the fact, arrives late	Captured by Connect, crossed with route events
Wait at the airport handler	Visible only if the handler warns you	Detected and signed as an event
MPS-Florimark label	Validated by eye at reception	Edge reads it and crosses it with the farm origin
Projected vase life	Estimated by the buyer at the clock	Projected before the auction, defensible
Post-auction claim	Rebuilt by hand, weeks	Dossier per batch, automatic, with photo and human signature

Impact estimate

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

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

• An exporter with 3-6 reefers a week (rose, gypsophila, lisianthus), two farms, one or two transit hubs and a European auction as the main market.
• A Tracer pilot on one farm and one reception at the auction platform (Edge for MPS + seal + datalogger on display, Connect for datalogger PDFs + AWB + the handler's messages, an agent projecting vase life). First value expected within a few weeks: the first batches already arrive with a defensible projected vase life.
• Indicative payback between 4 and 9 months, depending on the share of batches auctioned today below the expected price because of undocumented cold chain incidents.
• Reduction of vase-life penalties at auction of ≥30%, with a defensible floor well below the cost of every post-auction claim lost for want of evidence.
• The hard lever is the price defended at auction: two or three cents more per stem, multiplied by the weekly boxes, pay for the pilot in a few months.

And the quality manager's reasonable doubt

«What if the AI projects vase life badly and we leave a good batch on the floor?» — hallucination is a problem of free generation, not of anchored tasks. In tasks where the AI merely recontextualises a piece of data from one system into another (reading the datalogger curve and crossing it with the farm precooling), the best models brought the error below 1.5% ^[1]. And even so, what is critical is not decided alone: iLEAN holds the batch and the person validates. The three safety rings are there precisely for this.

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