In short: Virtual fencing collars such as Nofence enforce the boundary on the collar itself using GPS, so cattle, sheep and goats stay contained with no mobile signal at all. What stops working is everything the farmer sees and controls from the app (moving the fence, escape alerts, battery and position reports), because those travel over the public 4G network, and on upland, wooded and coastal grazing that network is often too thin to rely on.
Key Takeaways
- The fence works offline, the farmer does not — GPS keeps animals inside the last boundary the collar received, but without a mobile link a new paddock, a strip-graze move or an escape alert never arrives.
- Collars use LTE-M on public 4G masts — LTE-M reaches further than a phone call, yet a steep valley, a wood edge or a coastal headland can still leave collars silent for hours or days.
- A farm-owned cell closes the gap — A single multi-operator small cell or a private 4G/5G network on a building or mast can put the hill back in contact, and the same network then carries trough sensors, gate alarms and cameras.
In a nutshell
What is virtual fencing, and how does it work on UK farms?
Virtual fencing replaces wire and posts with a GPS boundary drawn on a phone and sent to a collar on each animal. When an animal walks towards the line, the collar plays a rising audio tone; if it keeps going, the collar gives a mild electric pulse, and the animal learns within a few days to turn at the sound. Nofence, the Norwegian company whose solar-powered collars are the most common system on UK farms, sells collars for cattle, sheep and goats, and the approach has spread fastest on ground where conventional fencing is expensive or unwelcome: commons, moorland, conservation grazing on nature reserves, rotational grazing on dairy and beef units, and steep hill land where a contractor might quote several pounds a metre for new fencing.
The economics are straightforward. A grazier moving a mob across a 40-hectare conservation site, or strip-grazing a herd onto fresh grass every day, can redraw the paddock in the app in a couple of minutes rather than spending a morning with reels and stakes. Regulation has also left room for it. England's ban on electronic training collars for cats and dogs, in force since February 2024, does not extend to livestock, and virtual fencing systems are sold for cattle, sheep and goats across England, Wales and Scotland.
Does virtual fencing work without mobile signal?
Yes, for containment; no, for management. Nofence's own guidance is clear that the collars do not need a cellular connection to enforce boundaries, because the collar compares its GPS position against the fence map it already holds and decides on the spot whether to sound the tone. An animal on a hill with no coverage stays inside the last fence it was given.
The mobile link carries everything else. When the farmer draws a new paddock, the change sits on the server until each collar next checks in. When an animal crosses the line and the collar gives up after repeated warnings, the escape alert waits in the collar until it can reach a mast. Position reports, battery levels and the activity data used to spot a lame ewe or a cow bulling all travel the same way. Nofence notes that the collars connect over LTE-M (Cat-M), a low-power, low-bandwidth variant of 4G designed for sensors, and that collars can often hold a connection where a phone shows no bars.
In practice, therefore, a collar with no signal behaves like a well-trained animal carrying a stale map. That is fine for a week on the same paddock. It fails the day the farmer needs to move stock off a flooding riverside field, pull sheep off a site before a public event, or find the heifer that walked through the boundary at 3am.
Why do virtual fencing collars lose signal on UK farms?
The ground that suits virtual fencing best is the ground the mobile networks reach worst. Operators built their 4G grids around roads, towns and villages, and the Shared Rural Network programme (a £1 billion deal between government and the four mobile operators, aimed at 95% of the UK's land area having 4G from at least one operator) has filled many gaps, but "at least one operator" is the important phrase. A collar on one network and a mast on another does not help, and the remaining 5% is disproportionately upland, forest and coast.
Four features of grazing land cause most of the dropouts we see on rural sites:
- Terrain shadow. A collar at neck height in the bottom of a Welsh or Cumbrian valley can be hidden from a mast two kilometres away by the valley side, even where the ridge above has good coverage.
- Woodland and hedgerows. Wet foliage absorbs signal; silvopasture, wood pasture and grazing under parkland trees all cut the margin.
- Animal bodies. The collar antenna sits against the neck, and a herd lying down close together adds loss that a phone held at chest height never experiences.
- Coastal and estuary grazing. Salt marsh and headland sites often sit at the edge of a cell's planned coverage, pointed inland at the nearest village.
LTE-M helps. It was designed with coverage-enhancement modes that repeat transmissions so a weak device can still be heard, which is why collars often connect where a handset does not. However, repetition costs battery and airtime, and in deep shadow even LTE-M runs out of margin. Signal can also change with the season; a field that worked in March can go quiet in August when the canopy is full.
What can a farmer do about virtual fencing in a no-signal field?
There are four practical options, and the right one depends on how much of the farm is affected and what else needs connecting.
(i) Plan around the gaps. The cheapest answer is to check coverage before buying, using the vendor's coverage map and a test collar, and keep collared stock on fields that check in reliably. This works when the dead zone is a single corner. It fails when the dead zone is the hill grazing that made virtual fencing attractive in the first place.
(ii) Accept delayed management. Some graziers on remote conservation sites run collars knowing that fence changes will land when animals wander onto higher ground. That is workable for slow rotational grazing with long stays on each block, and unsuitable for strip grazing or anywhere a fast escape alert matters, such as land next to a road or railway.
(iii) A signal booster. Ofcom allows only certain licence-exempt mobile repeaters, and they are designed to bring signal into a building or vehicle rather than across open fields. A booster in the farmhouse does nothing for a collar on the far side of the hill. We cover the limits in more detail in our piece on mobile signal boosters for farms.
(iv) Put a cell on the farm. A small radio cell mounted on a barn, a silo or a short mast on high ground can cover several square kilometres of open grazing at the low frequencies LTE-M uses. There are two ways to do it. A neutral host small cell broadcasts the public operators' networks, so collars, phones and any other SIM-based device on the farm attach to it exactly as they would to a roadside mast. A private 4G or 5G network belongs to the farm and carries the farm's own devices, with the farm deciding where it reaches and who uses it.
Neutral host or private network: which suits virtual fencing?
For the collars themselves, the deciding factor is the SIM. Virtual fencing collars ship with the vendor's own SIM, contracted to public networks, and the farmer cannot swap it for a private network SIM. A cell that carries the public operators' LTE-M, which is what a neutral host small cell does, is therefore the route that brings the collars back into contact with no changes to the collar or the app.
A private network earns its place for everything around the collars. Water trough level sensors, gate and building alarms, CCTV over the handling pens, a weather station, an RTK base for the arable side of the business and lone-worker devices for staff checking stock alone can all run on a farm-owned network, with known coverage and no per-device airtime charge. Many of these share the same pole: one mast on the highest building with a neutral host cell for public SIMs and a private cell for the farm's own sensors.
The trade-offs are cost and paperwork. A small cell needs a backhaul connection (fibre where it exists, otherwise fixed wireless or satellite), power, and for the neutral host route the agreement of the operators whose networks it carries. A private network needs spectrum, which in the UK usually means an Ofcom shared access licence in band n77 (3.8 to 4.2 GHz) or, for wider rural reach, the 1800 MHz shared band. Neither is a weekend job. Both are cheaper than a new mast from an operator that has no commercial case to build one in an empty valley.
How do you check coverage before buying virtual fencing collars?
Survey the network from the collar's position. A handset at chest height on the farmyard tells you very little about a collar at knee or neck height in a valley bottom.
We suggest three steps:
- Walk the grazing with a logging device on a cheap LTE-M modem or a test collar, at animal height, and record signal strength (RSRP) and quality (RSRQ) every few metres. Cover the fields at their worst: bottoms of slopes, under trees, along the shoreline.
- Check every operator, not just one. A field may be silent on one network and fine on another; the collar's SIM decides which you need.
- Repeat in summer if you surveyed in winter. Leaf cover can take several decibels off the margin.
The output is a map of fields that will report reliably, fields that will report intermittently, and fields that will go silent. If the silent fields are the ones you most want to graze with collars, that map is also the brief for a farm cell: where it goes, how high it needs to be and which networks it must carry. Our earlier piece on RTK correction signals covers the same survey approach for tractors, and the two surveys are often worth doing on the same day.
Where does virtual fencing fit in a connected farm?
Virtual fencing is often the first device on a livestock farm that is genuinely useless without coverage across the whole holding, rather than just around the yard. That makes it a useful test case. Once a farm has the coverage map from a collar survey, the same data tells it whether a trough sensor on the top field will report, whether a lone worker can call for help from the far boundary, and where a camera on the lambing shed can send pictures.
If you are planning a collar roll-out on land where the signal is uncertain, contact us with the holding boundary and the fields you want to graze, and we will run a coverage prediction for a farm cell before you buy. More on our work with farms is on the agriculture sector page.