Access point placement is the difference between Wi-Fi that quietly works and Wi-Fi that generates tickets forever. It is also the part most often decided by whoever was holding the drill.
This is how we actually plan placement on commercial installs — the rules we use, the numbers we aim for, and the mistakes we get called in to fix.
Coverage is not the goal. Capacity is.
The instinct is to place access points so that signal reaches everywhere. That gets you a network where every device shows full bars and nothing works at 11am.
Wi-Fi is a shared medium. Every device on a channel waits its turn. One access point serving eighty people in an open-plan floor will show beautiful signal strength and terrible performance, because those eighty devices are queueing.
So the real question is never “does signal reach here?” It is “how many devices will be in this area, and how much airtime will they need?”
How many access points do you need?
Start from device count and use case, then sanity-check against area:
General office, light use: roughly one access point per 1,500–2,500 sq ft
Dense open-plan, video calls: one per 1,000–1,500 sq ft, or one per 25–30 concurrent devices
Conference and training rooms: one per 15–25 seats, planned on capacity not coverage
Warehouse and back-of-house: coverage-driven, but watch ceiling height — a 30 ft ceiling changes everything
Walls matter more than square footage. A 3,000 sq ft open loft may need two access points. A 3,000 sq ft medical suite chopped into fifteen exam rooms with tile, plumbing and lead-lined walls may need five.
Where to physically mount them
⚠️ The single most common mistake we see is access points mounted on a wall at head height, or worse, sitting on a shelf in the IT closet with the door shut.
✅ What works:
Ceiling, not wall. Ceiling mount, face down, 8–12 ft. Above about 20 ft the useful cell gets wide and shallow and the client’s weak transmitter cannot answer back.
In the open, over the people. Put access points where people sit, not in corridors. Corridors act as waveguides and carry signal much further than you want, which creates interference two rooms away.
Away from metal. Not above a metal duct, not inside a metal cabinet, not behind a projector screen, not tight against a concrete beam.
Away from water and glass. Mirrors, fish tanks, elevator shafts and full-height filing are all serious attenuators. Water absorbs 2.4 and 5 GHz effectively.
One home run each. Every AP should be at the end of its own Cat6 or Cat6a run back to the rack. Daisy-chaining through a desk switch will hurt you eventually.
Outdoor and semi-outdoor areas — patios, loading docks, courtyards — need hardware rated for it, not an indoor unit pointed out a window.
2.4 GHz vs 5 GHz vs 6 GHz
These bands are not interchangeable, and treating them as one network is where a lot of performance goes.
2.4 GHz
Travels furthest and penetrates walls best, and is almost always the most congested band in a city. It has only three non-overlapping channels in the US — 1, 6 and 11. Everything else overlaps.
It is also shared with Bluetooth, microwaves, wireless cameras, cordless phones and your neighbours. In a dense Manhattan building, 2.4 GHz is usually a sewer. We often run it deliberately at low power, or disable it on most access points and keep it alive only where legacy devices need it — older barcode scanners, some building controls, some IoT.
5 GHz
The workhorse. Far more channels, far less congestion, shorter range, worse wall penetration. For almost every office, 5 GHz should be carrying your traffic and everything else is a supporting act.
Use 40 MHz channel width in most offices. 80 MHz sounds faster and gives you fewer non-overlapping channels, which in a dense deployment costs more than it gains.
6 GHz (Wi-Fi 6E and Wi-Fi 7)
Enormous clean spectrum, shortest range, stopped by almost any wall. It is excellent for high-density rooms where you can put an access point in the room itself. It is not a whole-floor solution. We cover the generational differences in our Wi-Fi 6 vs 6E vs Wi-Fi 7 guide.
What signal strength should you actually aim for?
Signal is measured in dBm, and it is negative — closer to zero is stronger.
−30 to −50 dBm: excellent, right under the access point
−50 to −65 dBm: solid. This is your design target throughout the usable space.
−65 to −70 dBm: usable for email and browsing, marginal for video calls
Below −70 dBm: expect problems
Design for −65 dBm or better at the edge of each cell for a voice and video environment. And remember that signal strength is only half the story — signal-to-noise ratio matters just as much. Strong signal in a noisy room still performs badly.
Channel planning and co-channel interference
If two access points within earshot of each other use the same channel, they do not run in parallel — they take turns. That is co-channel interference, and it is the quiet killer of dense deployments.
The fix is a channel plan that treats the floor like a honeycomb, so that no two neighbouring cells share a channel, including diagonally and vertically between floors. In a multi-tenant building you also have to plan around everyone else’s networks, which you do not control.
Turning transmit power to maximum on every access point feels like it should help. It does the opposite: it makes every access point audible to every other one, guarantees co-channel interference, and creates cells that clients can hear but cannot talk back to. Lower power is usually better than you think.
Access point vs range extender vs mesh
A range extender receives and rebroadcasts on the same radio, which roughly halves throughput on every hop and adds latency. They are a domestic sticking plaster and have no place in a commercial install.
Mesh is better — purpose-built, self-organising — but every wireless hop still consumes airtime that could have carried client traffic. Mesh is the right answer when you genuinely cannot run a cable: a landmarked facade, a detached structure, a temporary fit-out.
Everywhere else, run the cable. A cabled access point is faster, more predictable and easier to troubleshoot than any wireless backhaul. This is exactly why cabling gets designed before Wi-Fi.
Roaming: the part people forget
A laptop carried from a desk to a conference room should hand off cleanly. Most devices are stubborn — they cling to the original access point until the signal is genuinely bad, then drop and rejoin, which is what kills the call.
Good roaming needs overlapping cells (about 15–20% overlap at your target signal level), the same SSID and security settings across all access points, sensible transmit power so cells stay small, and the 802.11k/v/r fast-roaming features enabled and actually supported by your hardware. Both UniFi and Cisco Meraki handle this well when configured properly — and badly when left on defaults.
Validate, do not assume
A predictive design from a floor plan is a starting point, not an answer. Drywall thickness, unexpected steel, glass partitions and full filing cabinets all change the picture.
After install, walk the space and measure: signal and noise at the edge of every cell, throughput at the furthest desk, a real video call while walking the roaming path, and a check of what the neighbouring networks are doing to your channels.
We include a post-install survey on every enterprise Wi-Fi deployment, because the difference between a design and reality is always a few access points moved by a metre.
Need it done properly?
Super-G Intelligence designs and installs commercial Wi-Fi across New York City and Long Island — survey, cabling, access points, channel plan and documentation. Book a free site survey or call (718) 516-3554.
