Dairy cold chain — a break discovered at unloading is milk already lost.

The problem

The logger says everything went fine — but the rejection at the dock says otherwise.

A cold chain break in dairy transport is almost never seen in transit. The tanker's logger records, but nobody looks at it until the truck reaches its destination. And at that point, one of three things happens:

1. Rejection at the dock — the receiver downloads the logger, sees a spike outside the threshold, and rejects the truck. The transport drives back loaded, the product goes to destruction, the plant is left unsupplied.
2. Accepted with an incident — the receiver unloads but penalizes with a discount or a non-conformity. A hard cost and a damaged commercial relationship.
3. It slips through unseen — the logger records fine but the incident happened in that stretch without coverage and nobody ever knows. The product reaches the consumer with the chain broken. If it is caught in later lab analysis, there is a recall.

The carrier's dispatcher runs a mixed fleet (own tankers, subcontracted ones, different brands of refrigeration unit), each with its own telemetry portal, and none of those portals talks to the TMS. Watching trip by trip is impossible — and nobody finds out until the unloading dock.

How it fits the IRIS system

iLEAN does not replace the TMS or the logger — it seals the crack between the mixed fleet and a decision made in time.

The problem is not a lack of sensors: it is data scattered across manufacturer portals that nobody cross-references with the TMS, the GPS and the product inside. iLEAN acts as the putty that fills that gap. It does not ask you to change your fleet, your TMS or your refrigeration equipment.

Connect captures the telemetry wherever it comes from. Edge adds a sensor where there is none. The agent decides what is a break and what is noise — and alerts the dispatcher with time to act.

The iLEAN pieces applied to the cold chain in transport:

• Connect — captures the telemetry data from every brand of refrigeration unit (Thermo King, Carrier, Frigoblock, etc.), the available APIs, the incident emails from the subcontracted carrier, the TMS records and the planned stops. And it captures what arrives from outside: the driver's WhatsApp saying "the unit has dropped into demand mode" enters at second zero.
• Edge — optional, on the tanker or in the cab, when you need your own sensor (a tanker with no usable telemetry) or local processing (long stretches without coverage). It processes locally and reconciles on reconnection. What is critical cannot depend on WiFi.
• Cold chain agent — cross-references temperature, duration of the excursion, GPS position, planned vs. actual stops and the product inside. It tells a real incident (3 °C above range for 30 minutes on the road) from a false positive (1 °C at an authorized 2-minute stop). When it is real, it alerts the dispatcher and proposes actions. The dispatcher signs — the truck does not reroute itself.

See the full IRIS architecture →

Before and after

Passive logger checks vs. active tracking with iLEAN

Aspect	Classic logger (downloaded at the dock)	With iLEAN Connect + Edge + Agent
Moment the break is detected	At unloading, with the truck already at destination	In transit, with time to act
Multi-brand fleet	Each brand in its own portal, never unified	Single view, wherever the data comes from
Real break vs. false positive	A single threshold line — many false alarms	Temperature × time × GPS × product cross-referenced
Communication with dispatcher / customer	Reactive, already at the dock	Proactive, in transit, through the usual channel
Cold chain certificate per trip	A spreadsheet after the fact	Automatic signed dossier, ready for the auditor
Loss of GPS / telemetry coverage	A gap in the certificate, risk of rejection	Local buffer, reconciliation on reconnection

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 fleet. We put it forward so the committee has an order of magnitude; we refine it during the diagnostic, reviewing your historical incidents and your retailers' specs.

• Dairy cooperative or dairy processor with 15-50 tankers (a mix of own + subcontracted), 100-400 trips/week, 3-8 brands of refrigeration unit in the fleet, growing retailer spec requirements.
• Connect + cold chain agent deployment (TMS integration + APIs of the majority brands + Edge sensor on tankers with no usable telemetry). First value expected within a few weeks: a unified view of the whole fleet + the first in-transit alert avoiding a rejection at the dock.
• Indicative payback between 4 and 9 months, depending on the number of historical rejection incidents + non-conformity discounts + the cost of the product lost.
• Expected reduction in cold chain rejections ≥ 30% after the first cycle (the in-transit alert lets you adjust, redirect or renegotiate before unloading).

And the logistics director's reasonable doubt

“What if the agent hallucinates and fires a break alarm that isn't one?” — hallucination is a problem of free generation, not of anchored tasks. Cross-referencing real temperature with real GPS and regulated per-product thresholds is exactly the anchored task in which the best models brought error below 1.5% ^[1]. And even so, the agent does not act alone: it proposes to the dispatcher, who decides. The three safety rings exist precisely for this: no truck redirection goes out without the responsible person's signature.

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