Walking the building to find out where the coverage actually stops

Decide what you are measuring before you start walking

Two complaints get muddled together constantly, and a survey is worth doing partly because it separates them.

Coverage is whether a usable signal reaches a place. Its symptom is a connection that is weak in one room and improves as you walk back towards the radio, and its answers are placement and additional radios.

Throughput is how much gets through once connected. Its symptom is everything being slow at once, including things plugged in with a cable, and its answers lie somewhere else entirely — the line, a slow port, a machine at one end.

Start by establishing which one you have. Plug a laptop into an Ethernet socket and see whether the complaint survives. If it does, the survey below will produce a set of perfectly good readings that solve nothing, and the problem is not in this article.

Draw the plan and mark the fixed points

Sketch each floor, roughly to scale. Mark the position of the router or each access point, the rooms people actually use devices in, and anything you already suspect: a chimney breast, a stone wall, a metal-framed partition, a room with foil-backed insulation, a plant room, a cupboard full of pipework.

Then mark eight to fifteen stations — specific spots where you will take every reading, chosen to include the sofa, the desks, the beds, the kitchen worktop, the far corners, the stairs and the landing, and one or two points outside if the garden matters.

The stations are the whole method. Readings taken while wandering are not comparable to anything; readings taken twice at the same fifteen marks are evidence.

Record four things at each station

At each mark, stand still, wait a few seconds for the device to settle, and write down four things.

The signal level the device reports. The band it is connected on, since a phone that has fallen back to 2.4 GHz in a far room is telling you the 5 GHz radio no longer reaches. The access point it is attached to, where there is more than one. And a plain one-word note about how it feels — fine, hesitant, unusable — because that is what the household will actually judge the result by.

Hold the device the way people hold devices, at the height they hold them. A reading taken with a phone flat on a table near a radiator is not the reading that matters.

Do the whole circuit in one pass without changing anything. Then, if you can, do it again at a different time of day: neighbouring networks and household use both move, and a single pass captures one moment.

Find the cliff, because the cliff is the obstacle

Look down the column of signal levels for the place where the numbers stop declining gradually and fall off sharply between two adjacent stations. That step is not distance — distance produces a gentle slope — it is a thing.

Walk the boundary with the device in your hand and find the exact line. It is nearly always one identifiable object: a wall, a chimney, a steel-framed partition, a mirrored wardrobe, a floor with a mesh in the screed, a large appliance. Mark it on the plan.

Knowing which object it is decides where the next radio goes: on the near side of that obstacle with a clear path past it, rather than in the middle of the bad area. A radio placed inside the room it is meant to rescue is relaying through the same wall that caused the trouble.

Watch the stairs and the corridors, because that is where handover happens

Where a building has more than one access point, the interesting stations are not the rooms but the routes between them. A device generally holds on to the access point it has until the connection becomes genuinely poor, so the quality of a walk from one end of a house to the other is decided by how much the two coverage areas overlap.

Walk the route slowly and record where the device changes access point, and what the signal had fallen to before it did. If it holds on far past the point where the reading was poor, the two units are too far apart and the fix is overlap rather than power.

Two units also need the same network name and password for that handover to be automatic. Separate names turn every handover into a manual act nobody performs.

Change one thing, then walk exactly the same circuit

This is the step that makes the whole exercise worth the hour. Move the router. Raise it onto a shelf. Take it out of the cupboard. Change a channel. Add a node.

Then repeat the identical circuit, the same fifteen marks, in the same order, and put the new column of readings beside the old one.

One change at a time, because two changes produce a result you cannot attribute. And keep the losing configurations written down: a position that was worse in the far bedroom and better on the landing is a genuine trade-off you may want to revisit.

Read the class figure out of the decision entirely

Nothing in this procedure involves the number printed on the box. That figure — the digits after AX or BE — is the arithmetic sum of the theoretical peak rates of every radio in the unit, computed from the channel width and stream count the standard defines. It is a description of the packaging, and it does not vary with where the unit stands, which is precisely why it cannot answer a placement question.

The numbers that do vary with placement are the ones you have just collected in your own building, and there is no substitute for them on any specification sheet.

Keep the sheet, because the next question is a comparison

File the plan and the readings with the rest of the network notes. When somebody adds a room, buys a device that behaves oddly, or wonders whether a new access point is worth it, the survey from last year is what turns that into an answerable question instead of another description.