Wi-Fi in a building made of metal, water and moving forklifts
Office wireless design does not survive a packhouse, a coolstore or a wharf shed. Most industrial Wi-Fi problems are a survey that was never done.
Wireless in an office is close to a solved problem. Put enough access points in the ceiling, keep them apart, done.
Then the same approach gets applied to a packhouse or a coolstore and scanners drop mid-shift, tablets lose their session halfway down an aisle, and everybody concludes the hardware is bad. It usually is not.
Why these buildings are hostile
Metal reflects. Racking, cladding, conveyor frames and roller doors all bounce signal around, so coverage is not a smooth circle around an access point. It is a pattern with strong spots and dead ones a metre apart, and it changes when the racking is reconfigured.
Water absorbs. Product is mostly water, and so are people. A coolstore full of stock attenuates far more than the same coolstore empty, which is why a network that worked in September fails in March. Signal readings taken in an empty building are close to meaningless.
The environment moves. Forklifts, pallets stacked to the roof, a door that is open in summer and shut in winter. The radio environment on Tuesday is not the radio environment on Friday.
Cold and condensation. Equipment rated for an office fails in a chiller, and moving a device between a coolstore and a loading bay repeatedly is hard on it.
Interference you do not control. Industrial machinery, older cordless equipment, and the neighbour’s network coming through a shared wall.
The two failures that cause most complaints
Coverage was designed and roaming was not. A device standing still has signal everywhere. The problem appears when someone walks: the device clings to the access point it first joined until the connection is unusable, then drops the session rather than handing over cleanly. A scanner that loses its session mid-transaction is the single most common industrial wireless complaint, and it is a roaming and channel design problem rather than a coverage one.
Too many access points, turned up too loud. The instinct when coverage is poor is to add more and increase power. On the 2.4GHz band there are only three non-overlapping channels, so past a certain density the access points are mostly interfering with each other. Fewer, placed properly, at lower power, frequently performs better than more.
What a proper job involves
A survey in the building, loaded, with the equipment you will actually use. Not a prediction from a floor plan. Walk it with the scanner model your staff use, because a handheld with a small antenna sees a very different network from a laptop.
Design for the device, not the building. The weakest client determines the design. If the fleet is rugged handhelds, the network has to work for them, and they are usually much less capable than a phone.
Separate networks for separate purposes. Scanners and operational equipment on their own SSID and VLAN, staff devices elsewhere, guests nowhere near either. This is a security requirement and it also stops a phone streaming video from degrading the line.
Cabling and mounting suited to the space. Enclosures where there is washdown, mounting that survives a knock, and cable runs that were planned rather than improvised.
Re-survey after a change. New racking, a new chiller, a mezzanine. Any of those can undo a good design, and re-checking is a morning.
The cheap diagnostic
If wireless is unreliable, ask two questions before buying anything.
Does it fail while people are moving, or while they are standing still? Moving points at roaming, standing still points at coverage, and they have different fixes.
And does it fail worse when the building is full? If yes, the design was validated empty, and that is the whole answer.
Neither question costs anything, and between them they explain most of these problems.