WiFi is invisible infrastructure — until it stops working. When it does, everyone in the company knows about it, everyone agrees "something must be done", and then it dies quietly because no one knows where to start. This piece is about where.
In most Polish SMEs the WiFi story looks the same: one router was bought on sale, hung "somewhere in the middle of the office", then a second one was added in the warehouse because there was no signal, then a fifth one on the shop floor. Nobody planned a network — the network just accreted. The result: barcode scanners drop every other aisle, video calls stall at 10:30, and the guest WiFi (if it even exists) puts visitors on the same subnet as accounting.
A well-planned WiFi in a 10–50 person company is not an enterprise project. It doesn't cost 100 000 zł and it doesn't need a dedicated sysadmin. It does require dropping the "home router but at work" mindset.
One network, not a pile of routers
A home router with a single access point is designed to cover a 60 m² flat. If you hang two of them at opposite ends of an office corridor you don't get one network — you get two separate ones fighting for channels, with laptops that don't roam between them. A user walking from one room to the next loses their connection for 20 seconds and wonders why Teams keeps freezing.
The solution isn't a pricier router. The solution is a network built from proper access points that cooperate — same SSID, a common controller, doing real roaming: when you move closer to another AP, the laptop switches over automatically. In SME reality this usually means Ubiquiti UniFi, MikroTik cAP or TP-Link Omada — a single AP costs 400–900 zł, the controller runs on a small mini-PC or for free in the vendor's cloud.
- Measure real coverage (apps like WiFiman, Ekahau) before buying anything — don't guess.
- Count how many APs you actually need: "one AP per 100 m² of open office space" is a safe starting rule.
- Mount them on the ceiling, not on a desk or behind a cabinet. Antennas don't punch through metal.
- Put all APs under one controller — otherwise it's not a network, it's a pile of devices.
Two bands, one SSID
The 2.4 GHz band goes through walls but is congested (shared with Bluetooth, microwaves and the neighbour). The 5 GHz band is faster and cleaner, but has shorter range. Modern APs handle both at once — which is good, provided you don't publish them as two separate networks "Company_2G" and "Company_5G". Publish one name, let the devices choose. I've seen companies where sales sat on 2.4 GHz for a year because "5G didn't reach the meeting room", while 5 GHz worked perfectly well — nobody just switched to it on the laptop.
Guest, IoT, corporate — separate networks
This is the point most SMEs skip and end up with the accounting printer exposed to the internet through the reception's guest WiFi. The guest network must be physically separated from the corporate one (VLAN + different subnet). Cameras, printers, meeting-room TVs — on yet another one. This isn't paranoia, it's basic hygiene: when a visitor plugs in an infected laptop, they don't get into your HR.
Configuring this takes two hours on a decent controller and costs nothing beyond the hardware you already own.
- SSID "Company" — employees, separate password per department or via WPA-Enterprise (802.1X).
- SSID "Company-Guests" — open or with a weekly password, internet-only, client isolation on.
- SSID "Company-IoT" — printers, cameras, scales, payment terminals; no internet beyond what they need.
The warehouse isn't the office
In an office WiFi has to serve a stationary laptop. In a warehouse — a scanner moving between steel racks at walking speed, running a WMS session that mustn't drop. Those two scenarios need different radio plans, different AP density, different transmit-power settings and different attention to roaming.
At one of our clients — a 25-person electrical wholesaler — the warehouse scanners only worked properly next to one of four rack rows. Inventory always ended in arguments, because two warehouse staff had to walk their scanners back to the manager's cubicle to "pick up signal". Measurements showed 900 m² of hall being served by a single home Wi-Fi 5 router mounted next to that cubicle. We replaced it with 6 Ubiquiti U6-Pro access points laid out to match the rack plan, with a dedicated VLAN for the scanners. Hardware cost: 5 400 zł. Install and configuration: 3 800 zł. Result: for the first time in three years inventory finished in one day instead of two, and the WMS stopped throwing "server unreachable" a dozen times a day.
That's the difference we're talking about. Not 100 000 zł, but 9 000 zł and measurement instead of guesswork.
What to monitor so you know it still works
WiFi has this nasty property of degrading slowly. Today it works great, in six months "something's slower", in a year everyone's swearing and nobody can pin down when it broke. So you need minimal monitoring — no SIEM, no 20k dashboard, just three metrics your UniFi/Omada/MikroTik controller gives you for free:
- Clients per AP — if any AP is serving more than 30 clients, it's overloaded.
- Retries / TX-RTS errors — above 15% and someone is stepping on your channel (a neighbour, an analog camera, even an LED lamp).
- Roaming events — if laptops "stick" to one AP instead of switching, transmit power is set too high.
Half an hour of chart-checking once a quarter. That's enough.
What next
SME WiFi doesn't have to be expensive and it doesn't have to be complicated. It has to be planned. Three hours of survey work, an AP placement plan, decent mid-range hardware and a configuration with network segmentation — that's the standard a 10-person and a 50-person company can afford equally.
If you suspect your company suffers from "router patchwork" — instead of buying another range extender, do a measurement. Usually after 30 minutes it's clear where the problem is.
Good WiFi in a company doesn't cost more than bad WiFi. It just costs more thinking before you buy.