Wi-Fi problems are rarely a Wi-Fi equipment problem.
Almost everyone has lived with a wireless network that is fine most of the time and humiliating at the worst possible moment. The usual response is to buy another access point. That often makes it worse, because the thing that was wrong was not coverage — and nobody measured it before spending.
What this means for you
The technology is the same whoever you are. The consequence is not.
The equipment is probably fine
The router your provider supplied is not the reason the video call broke. In a building with concrete, foil-backed insulation or steel, the signal is being absorbed long before it reaches the far room — and adding a second box on the same channel as the first can make both worse. The cheap diagnosis is to measure what the devices actually experience where people actually sit, which costs a visit rather than new hardware.
Start hereBefore buying anything, write down where and when it fails. A pattern in time or place is worth more than any speed test.
Coverage is solved; quality is not
You have controllers, dashboards and a survey from the installation. What they report is signal strength, which is the least interesting number. Users experience roaming failures, retries, interference and airtime contention — none of which appear as an outage, all of which appear as "the wireless is bad". The gap between a green dashboard and an unhappy floor is where the real work is.
Start hereAsk for the retry and roaming statistics rather than the coverage map.
Guests, staff and things, on one estate
Public-facing sites carry visitors, staff and a growing population of devices nobody logs in to — cameras, sensors, displays. They share the same air whether or not they share a network. Separation policy that exists only in configuration, without measurement of what the radios are actually doing, tends not to survive contact with a busy day.
Start hereEstablish what is on the air now, before writing the policy for it.
The number everyone quotes is the wrong one
Wireless is almost always discussed in megabits. Peak throughput is easy to measure, easy to advertise and nearly irrelevant to how a network feels to use.
What people actually notice is smoothness: whether the call holds, whether the video stalls for a second at the same point in the hallway every time, whether the payment terminal hesitates. Those are questions about latency, retries, interference and hand-off — not about the maximum speed achievable by one device standing next to the antenna at midnight.
A network can test beautifully and behave badly. Optimising for the test is how that happens.
Why adding an access point often makes things worse
Radio is a shared medium. Every device within earshot of a transmission has to wait its turn, whether or not the transmission was meant for it. Two access points that can hear each other on the same channel do not double capacity — they halve each other’s airtime while both report excellent signal strength.
The same logic applies to power. Turning an access point up extends how far it can be heard, which sounds helpful. But a device at the far edge, just barely in range, transmits slowly and holds the channel for longer while doing so — degrading service for everyone closer in. There is a real case for cranking the power at the edge of a property and a real case for turning it down in the middle, and the difference is not intuitive.
None of this is visible from a coverage map, because a coverage map answers a question nobody was asking.
Buildings are the variable nobody accounts for
Signal loss depends on what is between the radio and the device, and buildings differ far more than product datasheets suggest. Concrete, foil-backed insulation, tiled roofs, mirrors and steel-framed glass all attenuate strongly and unevenly. Two rooms the same distance from the same access point can differ by an order of magnitude in usable signal.
This is measurable, and measuring it takes an afternoon. It is also the difference between placing equipment where it helps and placing it where there happened to be a power socket.
Roaming: the failure that looks like everything else
A device that has associated with one access point tends to stay with it, sometimes far past the point where a nearer one would serve it better — because the decision belongs to the device, not the network. The result is a laptop clinging to a distant radio at a crawl while a perfectly good access point sits idle two metres away.
Networks can help. Standards exist to hand devices between access points quickly, and to nudge a device towards a better radio or a less congested band. They work unevenly across manufacturers, and enabling them carelessly can lock older equipment out entirely. This is the most common source of “it worked yesterday” — and it is only diagnosable from records showing what each device did over time.
What good practice actually looks like
Measure before spending. The failure is almost never where the complaint says it is, and hardware bought on a guess usually adds interference.
Design for the quality of experience, not the peak rate. If the calls are smooth and the hand-offs are invisible, nobody will ever ask what the throughput was.
Change one thing at a time, with numbers either side. Wireless is full of interacting settings; a batch of simultaneous changes cannot be attributed, and an improvement you cannot attribute is one you cannot repeat.
Keep the records. Wireless problems are intermittent by nature. Without history, every incident is investigated from scratch and the same causes are rediscovered repeatedly. With history, the pattern is usually visible before anyone has finished describing the symptom.
Treat the devices you cannot control as part of the design. Printers, televisions, cameras, sensors and guest phones will not be reconfigured to suit you. They are part of the environment, and a design that assumes cooperation from them is a design that will disappoint.
Where to start
We work through this with boards, owners and technical teams — and we are happy to give this material as a talk rather than a proposal.
Reviewed 22 July 2026. This is a summary of public programmes and moves quickly — treat it as orientation, not as advice on a specific obligation. Drawn from material first presented as: Written from live measurement of a multi-building property: RF surveys, controller telemetry and per-client roaming records.