In short: UK frozen and chilled warehouses are legally obliged to record air temperature, and most of them do it with wireless sensors hung in an environment engineered to defeat wireless. Cold storage temperature monitoring fails at the network long before it fails at the sensor.
Key Takeaways
- The law asks for records, not thermometers — Quick-frozen food must be held at -18°C or colder with air temperature monitored and recorded, so a gap in the data is a compliance gap even when the product never warmed up.
- Cold stores are built the way radio hates — Dense steel racking, ice on every surface, insulated panel walls and moving high-bay cranes turn a warehouse into a set of metal corridors where 2.4 GHz Wi-Fi rarely reaches the far aisle.
- One excursion can cost more than the network — A rejected 26-pallet load of frozen product, plus the investigation and the retailer's view of you afterwards, buys a private 5G deployment several times over.
In a nutshell
What is cold storage temperature monitoring, and what does the law require?
Cold storage temperature monitoring is the continuous measurement and recording of air temperature inside a chilled or frozen warehouse, in the racking, at the dock, in the blast freezer and in the vehicles that load against it. What the operator is really producing is a defensible record: proof that a specific pallet sat within its specification for every hour it was in their custody.
The UK requirement sits in two places. Quick-frozen foodstuffs must be held at -18°C or colder, with brief upward tolerances for practical handling, and the operator must monitor and record the air temperature in storage and transport at frequent intervals; that regime came into UK law through the Quick-Frozen Foodstuffs Regulations and the retained EU rules on monitoring, and the records must be kept and made available to the enforcement officer. Everything else runs through hazard analysis under the food hygiene regulations, where temperature is almost always a critical control point, which means you have to prove that you monitor it, that you set limits, and that you did something when a limit was breached.
Then there is the commercial layer, which is usually stricter than the statutory one. A BRCGS audit will ask for the records. A retailer's technical specification will ask for the records with the delivery. And any dispute about a rejected load, which is where the money is, gets settled by whoever can produce a complete temperature trace with a credible chain of custody. Whoever has a gap in the data pays.
Why do wireless sensors drop out in a frozen warehouse?
Because a cold store is a near-perfect radio obstacle course, and it gets worse as it fills.
Start with the fabric. Insulated composite panel walls with steel skins on both faces, a steel roof, a floor slab with underfloor heating pipes, and no windows. There is no leakage in or out, which is helpful for security and useless for coverage. Inside, drive-in or mobile racking packs steel uprights at close spacing across the full width of the building, then loads them with pallets of dense frozen product, which is mostly water and absorbs 2.4 GHz enthusiastically. An access point that surveys beautifully in an empty building can lose the far aisle the week the racking fills.
Then add ice. Frost forms on every cold surface, and it forms on sensor housings, antennas and access point radomes as readily as on the racking. Kit that cycles between a -25°C chamber and a +5°C dock, which is what happens to anything mounted on a mobile racking end or a forklift, gathers condensation on the warm leg and freezes it on the cold one. Battery chemistry does not enjoy this either: a coin cell rated for years at room temperature can lose most of its useful life at -25°C, so sensor batteries fail at exactly the moment nobody wants to send an engineer up a mast lift in a freezer.
Then add movement. In an automated high-bay store, stacker cranes travel the aisles up to around 40 metres tall at speed, and each crane is a large steel object moving through the propagation path several times a minute. Every aisle is effectively a waveguide with a moving reflector in it. Wi-Fi roaming across that environment is where handheld terminals lose sessions and where warehouse management scans get re-keyed by hand at the dock.
The practical result across a lot of UK sites is a monitoring system that is 95 per cent reliable, which sounds fine and is not. Five per cent of a 15-minute logging interval across 300 sensors is a lot of blank cells in the record that an auditor will ask about.
What does a temperature excursion actually cost?
Name the numbers and the business case usually settles itself.
A standard frozen load is 26 pallets. Frozen ready meals, ice cream or premium protein at wholesale prices puts that load somewhere between £20,000 and £60,000 depending on the product, and a rejected load is rarely reworkable: once the record shows an excursion, the retailer's technical team is not interested in whether the product is actually fine. The disposal costs money too.
Behind the load sits the investigation. A confirmed excursion triggers a non-conformance, a root cause exercise, a review of every other pallet stored in the same location over the same period, and, if the site is third-party logistics, a conversation with a customer who now wants to know how much else of their stock was affected. Sites that cannot answer that question precisely end up quarantining far more than they needed to, because the only safe answer without good data is all of it.
Behind that sits the relationship. UK cold storage is a competitive third-party market: NewCold at Wakefield, Magnavale, Partner Logistics, Reed Boardall at Boroughbridge and a long tail of regional operators are all selling the same square metres of frozen space at broadly similar rates. What differentiates them to a national retailer is service reliability and the quality of the evidence they can produce. Losing a customer over a data gap is a much bigger number than the pallets.
Set that against the cost of fixing the network. An Ofcom Shared Access licence in the 3.8 to 4.2 GHz band starts at £80 a year for a low-power licence. The radios, core and installation for a single large shed are a capital project measured in tens of thousands, not hundreds. One avoided rejection and one retained contract is the whole argument.
What about the automated high-bay store?
The newest UK cold stores are the hardest case and the clearest one. An automated high-bay freezer runs at -25°C with almost no people in it, stacker cranes serving racking 35 to 40 metres tall, pallet conveyors through an airlock, and a warehouse control system that has to know exactly where every pallet is because nobody can walk in and look.
Everything in that building is a data flow: crane telemetry, position encoders, drive temperatures, evaporator fan and defrost cycles, door interlocks, ammonia detection in the plant room, and the temperature sensor grid itself. Ammonia is the refrigerant of choice at this scale for good thermodynamic reasons, and it is a named substance under COMAH, so the larger plants sit inside a formal safety regime where detector coverage and alarm handling are inspected rather than assumed.
Cabling all of that is possible and is what the integrator will quote. The catch shows up in year three, when the operator wants to add pallet-level sensors, or track the reach trucks in the pick face, or put cameras on the inbound dock, or retrofit condition monitoring to the compressors, and every addition is another cable run through insulated panel and cold aisle at cold-store labour rates. A private 5G layer changes the arithmetic: coverage is planned once for the building, and each subsequent device is a SIM rather than a civils job. Licensed spectrum also means the sensor grid is not competing with the site Wi-Fi, the guest network and whatever the neighbouring unit installed last month.
We should be honest about the limits. Private 5G does not solve a badly sited sensor, and it does not put a radio through 30 metres of solid frozen palletised product any more than Wi-Fi does. High-bay aisles need radios planned per aisle or a leaky feeder run, and anyone who tells you a single macro cell will cover an automated freezer has not stood in one.
Can the same network pay for itself on energy?
Cold storage is one of the most energy-intensive uses of a warehouse in the country, and refrigeration load is unusually flexible: a well-insulated frozen chamber is a thermal battery. Operators have used that for years to buy power more cleverly, pre-cooling on cheap overnight units and easing back at peak, and a growing number sell that flexibility into balancing and demand-side response markets through aggregators such as Flexitricity and Axle Energy.
The requirement for doing that safely is exactly the requirement for compliance: fine-grained, trustworthy, real-time temperature data across the whole chamber, not one probe near the door. If you are going to let an aggregator shed your compressors for 20 minutes, you want to see the product temperature response in the racking as it happens, with alarms that work. Sites that install monitoring for the auditor and discover they can also trade flexibility with it tend to find the payback arrives from the energy side first.
Where should a cold store start?
With a survey done under load, not on the day the racking went in. Walk the building full, at temperature, with the cranes or the trucks working, and record where the existing network actually reaches. Most operators are surprised by the map.
Then instrument the exceptions rather than the whole estate at once. The dock and the airlock, where product changes state and most excursions begin. The blast freezer, where the process itself is the risk. The top of the high bay, where warm air collects and where nobody has a ladder. Put a private 5G radio plan behind those points, prove the data is continuous for a quarter, and use that record in the next BRCGS audit.
If the monitoring system currently produces a spreadsheet with gaps in it that someone explains to the auditor by hand, that is the thing to fix first, and the network is the reason the gaps are there.