A university client called us because Wi-Fi failed every time a lecture hall filled up. Students had four bars. The previous installer's heat map was entirely green. The design had answered the wrong question.
Wi-Fi is a shared medium. Every device in range of the same channel takes turns to transmit, and only one can talk at a time. Coverage tells you whether a device can hear the access point. It says nothing about how long that device waits for its turn. Once a room fills, waiting is the entire user experience.
Why the heat map lied
A predictive survey drops access points into a floor plan and models signal propagation. It is genuinely useful for coverage and for finding walls that will cause trouble. What it usually does not model is:
- Bodies. Three hundred people absorb 5 GHz signal substantially. A room surveyed empty behaves differently full.
- Co-channel contention. Add more access points on the same channel and they compete with each other. Beyond a point, extra access points make throughput worse, not better.
- Client behaviour. Devices decide for themselves when to roam, and many hold onto a distant access point far longer than they should.
- Airtime, not bandwidth. One old device transmitting at a low data rate consumes disproportionate airtime that everyone else waits through.
Design for concurrency instead
Start with a device count, not an area
Ask how many devices will be active in the space simultaneously, then be honest about the multiplier. In education, count roughly 1.8 devices per person: a phone always, plus a laptop or tablet often. In an office, 1.5. In a conference hall with a live-streamed session, closer to 2.5.
Set a realistic per-device budget
Decide what each device actually needs. Browsing and email is around 1 to 2 Mbps. Video conferencing is 2 to 4. Streaming video is 5 or more. Multiply by concurrent devices, and you have an aggregate requirement that is normally far smaller than people fear, and far larger than the existing kit delivers at peak.
Then work out how many radios that needs
A Wi-Fi 6 access point on a clean 80 MHz channel in a busy room realistically delivers a few hundred megabits of usable aggregate throughput, not the number on the box. Divide your requirement by that figure and you have your radio count. In a high-density hall this typically means one access point per 25 to 40 devices, which is a very different answer from one per hundred square metres.
Design the channel plan before the mounting positions
In dense spaces, use 20 or 40 MHz channel widths rather than 80. Narrower channels give you more non-overlapping channels to reuse, and in a contended environment reuse beats raw per-client speed every time. This is the single most common thing we change on an existing deployment, and it often improves things before any new hardware arrives.
Practical settings that make a real difference
- Disable the lowest data rates. Removing 1, 2, 5.5 and 11 Mbps stops distant and legacy clients from consuming enormous airtime, and encourages devices to roam to a nearer access point.
- Turn down transmit power. Counterintuitive but correct. Lower power makes cells smaller, which increases reuse and encourages roaming. Loud access points that shout across a building are the most common cause of self-inflicted contention.
- Limit the number of SSIDs. Every SSID broadcasts management frames on every radio. Four SSIDs is usually two too many. Use one corporate network with role-based policy instead of one per department.
- Enable band steering carefully. Push capable clients to 5 or 6 GHz, but do not make 2.4 GHz unusable for the devices that only support it, such as barcode scanners and building sensors.
- Cable for the future. Run two cables to every access point position. The second one costs very little during a fit-out and saves a ceiling being reopened later.
Validate with a real load. A post-install survey with one laptop proves coverage. It proves nothing about density. Test with the room full, during a genuine lecture or all-hands, and measure retries, airtime utilisation and client counts per radio rather than signal strength.
Where density design matters most
| Environment | The real challenge |
|---|---|
| Lecture halls and auditoriums | Extreme concurrency in short bursts, plus body absorption |
| Hotel conference floors | Load that changes completely week to week |
| Warehouses | Roaming at speed for scanners, with racking that moves and blocks signal |
| Hospital wards | Mixed clinical devices, many legacy, with zero tolerance for dropouts |
| Open-plan offices | Video calls everywhere at once, upstream as well as down |
The outcome
At the university, we re-surveyed on site with the halls occupied, moved to 40 MHz channels, disabled low data rates, reduced transmit power and added access points in the two largest halls only. Capacity complaints stopped. Roughly half the improvement came from configuration changes that cost nothing at all, which is usually the case.
If your Wi-Fi is fine at nine in the morning and unusable at eleven, you do not have a coverage problem. Do not let anyone sell you one more access point until they have measured airtime.