Why mobile networks fail at events

This matters to organisers for a reason that goes beyond irritated attendees. The same effect is what takes down trader card machines, ticket scanners at the gate and the production team’s uplink, usually within the same twenty-minute window. If you understand why it happens, the fix stops looking like an expense and starts looking like risk management.

Bars measure the wrong thing

Signal bars tell you how loudly your phone can hear the mast. They tell you nothing about how many other people are talking to it.

A mast has a fixed amount of capacity, shared between everyone connected to that cell. On a normal Tuesday in a village, that is a few hundred people spread across a wide area. On the Saturday of your show, it is several thousand people standing in a field that sits inside one cell, all of them wanting to upload the same photograph at the same time.

The data on this is unambiguous

Ookla measured peak-time congestion across ten European markets with high 5G adoption in early 2026. 4G speeds drop by an average of 32% at peak. 5G drops by 27%. The UK scored 30.7 on their congestion scale, where zero is uncongested and Spain, the worst performer, sat at 62.1.

That is the drop on an ordinary weekday evening, from ordinary domestic demand spread across a whole town.

An event is not an ordinary evening. An event takes a town’s worth of people, compresses them into a few hectares, and points them all at one or two masts for twelve hours. Then it does it again the next day.

The starting point matters too. Opensignal’s UK measurements from January 2026 give overall download speeds, blending 4G and 5G, of 53.2 Mbps on EE, 51.0 on Three, 37.5 on Vodafone and 32.8 on O2. Upload sits between 6.4 and 10.4 Mbps. Those are the real-world national averages, not the headline 5G numbers, and they are the honest place to start when you are sizing a field.

Take the O2 figure of 32.8 Mbps down, knock 32% off it for peak congestion, and you are at roughly 22 Mbps before you have accounted for the fact that everyone at your event is on the same cell rather than spread across the country.

The coverage problem underneath it

There is a second issue that only shows up outside cities, and Ofcom publishes the number for it.

That twelve-point gap is where a lot of events happen. Showgrounds, farm sites, country parks, estate grounds. In that gap, whether you have usable data depends entirely on which network you happen to be on. Your ticketing supplier might be fine on EE while every trader on O2 is dead, and neither of them will understand why.

It is also why a single 4G router is a poor answer for anything that matters. You are betting the whole event on one carrier’s coverage at one spot, and you find out whether you were right on the morning of the show.

What breaks, and in what order

When a cell saturates, things fail in a fairly predictable sequence, and it is rarely the thing you expect.

Card machines go first. Not because they use much data, but because they are latency-sensitive and they retry. A payment terminal that cannot reach the acquirer within a few seconds throws an error, the trader tries again, and now there are two transactions queued on a network that could not carry one. Multiply that across forty stalls at half past two on Saturday. More on that in cashless events and what traders need from your network.

Ticket scanning goes next, usually at the worst moment, because gate rush and network peak are the same event. A queue that should move at four people a minute drops to one, and forty minutes later you have a crowd management problem rather than a connectivity problem.

Upload dies before download. This surprises people. Mobile upload is a fraction of download at the best of times, 6 to 10 Mbps nationally, and it degrades first under load. Anyone streaming, sending files, or running a broadcast position discovers this at exactly the point they cannot afford to.

Public browsing is last, and it is the only one your attendees will actually complain about on social media. By the time they do, the operational failures have already happened.

What actually fixes it

There is no version of this where you make the public mobile network work better. It is not yours and it is not going to change. What you can do is stop depending on it for the things that matter.

Bring your own backhaul, and more than one route. Where a site can take a fixed line, that is the strongest answer and the one we would recommend at scale. Where it cannot, bonding several carriers means the coverage gaps do not line up, and if one network saturates the others carry the traffic. Where there is no useful mobile signal at all, Starlink gives an independent path that does not care what the local masts are doing. Its UK median download was measured at 106 Mbps in Ookla’s most recent published UK figures.

One caveat worth knowing if you are comparing quotes: when traffic is bonded it runs inside a tunnel, so the router’s VPN throughput sets the real ceiling, not the firewall figure printed on the box. Ask which number you are being quoted.

Separate the networks that matter from the ones that do not. Card payments, ticketing, production and welfare each get their own network, isolated, with their own share of the bandwidth. Public WiFi gets what is left. Done properly, a field full of people streaming video cannot slow down a card terminal, because they are not sharing anything.

Design for the crowd, not the datasheet. Enterprise access point datasheets routinely quote 300 or more connected devices. That is an association ceiling, not a working figure. Realistic high-density design sits nearer 25 clients per radio, or around 50 per access point. A network designed on the datasheet number is roughly six times oversubscribed and will fall over at exactly the moment it is needed.

Use 20 MHz channels. In a dense deployment, narrower channels mean less interference between your own access points. It is one of the clearest signs that a network has been designed rather than deployed.

Plan for the take-up you actually cause. A network nobody knows about sees a fraction of the crowd. Advertise free WiFi on the wristband and take-up can multiply, anywhere from 20% to 100% of the crowd connecting. If you are going to promote it, say so early, because it changes the access point count.

What good looks like on the day

If it has been done properly, nobody notices. Traders take payments all afternoon without mentioning it. The gate moves. The production team uploads without asking anyone to get off the WiFi. That is the entire deliverable, and the measure of it is an absence of phone calls.

If you want a sense of what that costs, we have put indicative UK price bands in one place along with what actually moves the number: how much event WiFi costs. Or see how we approach festival WiFi and event WiFi generally.