Prove every shipment stayed in the cold chain.

A frozen load almost never arrives obviously bad, so failures go unbilled until a customer complains about texture. The harm is recrystallization from short partial thaws — a propped door, a reefer cycling off — that a once-an-hour logger sails straight past.

Read the reefer over CAN / J1939 (setpoint, supply and return air, run-hours) and correlate it with in-load probes and door events; IoT Logic fires on rate-of-change and on door-open dwell, not a static −18 °C line, so a thaw is caught while it is still minutes old.

Produce and flowers respire — they generate their own heat and ethylene — so the reefer's return-air sensor reads in-spec while the pulp in the center of the pallet self-heats out of range. 'The reefer said it was fine' is exactly how a living load arrives over-ripe.

Instrument the load, not the unit: independent probes front / middle / rear plus a humidity sensor, with pre-cool verified before loading. IoT Logic watches the divergence between supply air and in-load temperature, so a hot core trips an alert long before the unit's own sensor notices.

Chilled fish and meat live in a band barely three degrees wide where both edges are destructive: dip below and the texture is wrecked, drift above and shelf-life burns down by the hour. With a window this tight, an undramatic slow drift does as much damage as a spike — and average-only monitoring hides it.

Use high-resolution 1-Wire probes (to ~1/16 °C) at the two diagnostic spots — by the door and at the evaporator — and alert on rate-of-change so a gentle drift is flagged early. Every reading is evaluated against the tight per-load corridor, not sampled or averaged away.

On a multi-drop route the enemy isn't a broken reefer — it's thirty open doors, a contractor fleet that rotates weekly, and drivers who switch the unit off to pocket reimbursed fuel. The chain breaks operationally, in the leg you don't own and can't rewire.

Plug-in socket terminals install in minutes and swap between contractor vans; stick-on BLE sensors need no wiring. Correlate reefer power-state with the temperature curve to expose an off-in-transit unit, add door, geofence and standstill rules, and give every alert an owner with an escalation ladder.

Download-after-trip loggers only confess once the load is already lost. Trailers sit unhooked in a yard with no tractor power, cellular dead zones blow holes in the audit trail exactly where coverage is worst, and one fleet-wide threshold on a multi-temp trailer both misses excursions and cries wolf.

A reefer J1939 gateway plus independent per-zone probes, on solar or battery so an unhooked trailer keeps reporting; store-and-forward backfills every dead-zone gap on reconnect, and each zone carries its own threshold — so the record is complete and the alerts are real.

Plenty of regulated cargo isn't refrigerated at all — controlled-room-temperature biologics, humidity-sensitive electronics, chemicals with narrow stability windows. With a single load worth six figures, the expensive part is rarely the degree or two; it's the dispute over who caused the excursion and who held the goods when it happened.

Track temperature and humidity with custody and location on one timeline, scope rules with geofences so an excursion is only allowed inside an unloading zone, fan alerts to multiple recipients, and let IoT Query assemble the chain-of-custody and MKT proof that settles the argument in minutes — fed straight to QA over the API.