Skip to main content
Back to the guide
Networking13 min read

Planning UniFi Wi-Fi: coverage, channels and roaming

Good Wi-Fi is not defined by full signal bars in the corridor. It is defined by the intended applications working with the actual endpoints in every important location. This guide explains how to plan, survey and accept a UniFi Wi-Fi network for SMEs around capacity, the building, frequency bands, channels and roaming. As of 8 August 2026.

Published on August 8, 2026 · Daniel Gläser

Planning UniFi Wi-Fi: coverage, channels and roaming

1. Plan capacity, not just floor area

Floor area is only one input. A nearly empty warehouse aisle and a full training room can be the same size but impose entirely different requirements. Wi-Fi shares airtime: the more active devices, slow transmissions, broadcasts and interferers occupy a channel, the less airtime remains for each participant. Planning therefore starts with usage zones and applications.

Inputs required for capacity planning
ZoneRecordWhy it matters
Work areasConcurrent laptops, phones, docks, cloud and file useDetermines capacity, band choice and uplink demand
Meeting and trainingMaximum occupancy, video meetings, presentation and guest devicesMany active clients coincide with real-time traffic
Warehouse and productionScanners, tablets, machinery, movement paths, metal and changing stockRobust connectivity and roaming often matter more than peak throughput
Point of sale and voiceBusiness-critical workflows, latency and interruption toleranceRequires dedicated, testable acceptance criteria
Outdoor and secondary areasPurpose, weather, mounting, neighbouring networks and regulatory limitsNeeds suitable hardware and a separate radio assessment
The number of concurrently active devices matters more than the number of all registered devices. Design and acceptance use representative workloads rather than theoretical maxima.

No credible metre or square-metre rule

A fixed range per access point ignores construction materials, mounting, antenna behaviour, interferers and widely varying client radios. A defensible AP count comes from prediction and measurement, not a universal area formula.

2. Materials and obstacles determine the radio path

Radio does not spread through an office as a neat circle. Concrete, fire walls, coated glass, lifts, steel shelving, machinery, water and even changing stock attenuate or reflect signals. Openings and corridors can carry radio unexpectedly far. Higher frequencies behave differently from lower ones, so a floor plan alone is not yet a radio model.

  • Draw wall and ceiling types, and inspect them on site where the construction is uncertain.
  • Mark metal surfaces, cold rooms, lift cores, service shafts and movable shelving as special radio zones.
  • Measure neighbouring networks and non-Wi-Fi interferers during real operating hours, not only in an empty building at the weekend.
  • Include weaker or business-critical client types. A high-end survey laptop does not represent a scanner with a small antenna.

Ubiquiti publishes antenna radiation patterns for its access points. They show the direction and shape of radiation under laboratory conditions. The vendor itself notes that materials, geometry and interference alter real installations, and recommends measuring before, during and after installation.

3. Place access points around the intended radio cell

Mount an access point where its antenna behaviour serves the intended usage zone and a sound cable route is possible. The geometric centre of a floor is not automatically correct. With several APs, cells should overlap in a controlled way: enough to support transition, but not so much that many clients and APs unnecessarily share the same airtime.

  • Install ceiling and wall models in their intended orientation. Arbitrary rotation changes the radiation pattern.
  • Keep APs clear of metal, service channels, lights, cupboards and other direct obstructions where possible.
  • Provide capacity locally in busy rooms instead of covering them from a distant cell at high transmit power.
  • Do not raise transmit power as a substitute for a missing cell. The client still has to complete the return path to the AP.
  • Make mounting points accessible, clearly labelled and photographed so replacement does not require redesign.

Mock up before drilling

A temporary mount using the intended AP model is worthwhile in difficult areas. A few metres or the opposite side of a fire wall can change the result considerably. Measuring a mock-up is cheaper than rerouting cable after installation.

4. 2.4, 5 and 6 GHz serve different purposes

Frequency bands in business Wi-Fi planning
BandPlanning benefitConsiderations
2.4 GHzBroad client support and better penetration than higher bandsLittle non-overlapping channel space, many third-party and IoT devices, high interference exposure
5 GHzMore usable channel space and the main performance band for many clientsLess reach and material penetration than 2.4 GHz; include DFS behaviour and channel planning
6 GHzAdditional, often less occupied spectrum for Wi-Fi 6E and Wi-Fi 7 clientsCompatible clients only, shorter radio paths, regional rules, and WPA3 and PMF requirements
Band choice is not a ranking. A mixed endpoint estate often needs several bands, with cells and settings planned separately.

6 GHz is not an automatic range booster. It provides additional spectrum for suitable clients but requires closer attention to usage zones and compatible security configuration. Some IoT and legacy endpoints still need 2.4 GHz, but it should not carry the entire high-performance fleet. In many SMEs, 5 GHz is the practical main band, yet it still needs planning around actual interference.

5. Channel plan and width: distribute airtime deliberately

Wider channels can deliver more peak throughput in good conditions, but they also occupy more spectrum. In a dense environment this reduces the number of independently usable channels and makes neighbouring cells compete sooner. The correct width follows density, interference, client capabilities and application targets, not the largest selectable value.

  • Start 2.4 GHz planning at 20 MHz and avoid overlapping channel use. Wider settings aggravate an already constrained spectrum.
  • On 5 GHz, balance channel width against cell density and neighbouring occupation. More width per cell leaves fewer separate channels for adjacent cells.
  • Wide channels are possible on 6 GHz, but make sense only when clients, capacity targets and channel reuse support them.
  • Avoid assigning the same channel to directly adjacent APs where possible. Your own Wi-Fi can be the strongest source of interference.
  • DFS channels can expand the 5 GHz options, but must vacate in response to radar under regulatory rules. Include that behaviour in acceptance testing.
  • Use Channel AI and automated optimisation as tools, but review recommendations against the floor plan, measurements and critical applications.

Automation needs an engineering frame

Automatic channel selection sees the measured radio environment, but it does not automatically know your business processes, planned event times or critical zones. Channel width and transmit power remain design decisions and require validation after changes.

6. Site survey: connect prediction, measurement and acceptance

A professional workflow combines three perspectives. Predictive design models the floor plan, materials, AP models and expected cells. An on-site survey checks assumptions and captures neighbouring use or interferers. Post-installation validation shows whether the built network meets the agreed objectives under real conditions.

The three phases of a Wi-Fi survey
PhaseTaskDeliverable
PredictiveModel floor plan, materials, capacity zones, antennas and cable routesProposal for AP quantity, positions, bands and initial channel design
On siteMeasure temporary APs, spectrum and representative clients at critical pointsConfirmed or corrected mounting points and identified interferers
Post-installationTest coverage, SNR, airtime, packet behaviour, latency, throughput and roamingAcceptance record, remaining issues and documented tuning
Thresholds are derived from applications and client classes before the project starts. One RSSI value is not a complete quality proof.

Measure at different times with representative devices. Alongside signal level, capture signal-to-noise ratio, channel and airtime utilisation, retries, packet loss, latency and usable throughput. A green heatmap cell does not prove a stable video meeting or a clean scanner handoff between warehouse areas.

7. Cabling and PoE are part of radio design

Every AP is limited by its uplink and power supply. The intended model determines port speed, IEEE PoE class and maximum draw. The switch must meet those requirements per port and in total. Cable run, patching and certification must support the expected data rate.

  • Cable every AP in normal business installations. Mesh is a justified exception, not the default substitute for data cabling.
  • Assess AP port, switch port and uplink speeds together so the bottleneck does not sit immediately behind the radio.
  • Check PoE mode and maximum draw in current data sheets instead of relying on the generic PoE label.
  • Plan total PoE availability with headroom, and assess the effect of switch or power loss on critical radio cells.
  • Test and label cables and outlets after installation. A faulty pair can present as an apparent Wi-Fi fault.

8. Roaming is assisted, but the client decides

The client fundamentally decides when to leave an access point and which one to join. UniFi can assist through neighbour information, BSS transition and fast roaming. This is reliable only when cells overlap sensibly, security settings are compatible and the endpoints implement the mechanisms correctly.

  • Test roaming along real movement paths: corridors, stairs, warehouse aisles, outdoor transitions and lift lobbies.
  • For voice and real-time applications, record packet loss, latency spikes and audible interruption while moving.
  • Respect client differences. Operating system, driver, power saving and antenna affect the decision.
  • Do not maximise transmit power everywhere. Oversized cells can keep clients attached to a poor AP for too long.
  • Use Minimum RSSI only per AP after measurement, not as a global supposed roaming switch.

Minimum RSSI is not a cure-all

When a client falls below the configured value, the AP can disconnect it. The client must then select a better AP itself and may even reconnect to the same one. Ubiquiti therefore warns against universal values: poor tuning can create repeated disconnections and instability.

9. Validation: accept applications, not just measurements

Agree acceptance criteria before construction. They apply to defined areas, operating conditions and client classes. Measurements provide diagnosis, while acceptance also follows the business workflow: the scanner books stock, the call remains intelligible while walking, and the video meeting works in a full room.

Examples of testable acceptance points
TestMethodAssessment
CoverageMeasurement points in all agreed usage zones with defined client classesProject targets for signal, SNR and band are met
CapacityRepresentative concurrent load in high-occupancy areasApplication, latency and usable throughput remain within the agreed range
RoamingMovement along defined paths with real-time trafficTransitions meet the agreed interruption and quality limit
InterferenceInspect airtime and spectrum during normal and peak operationNo unassessed persistent interferer or avoidable co-channel use
FailureTest agreed switch, WAN or power scenarios in a controlled mannerExpected behaviour, alerting and recovery are documented
Exact thresholds depend on applications, client hardware and risk profile. Define them before measuring rather than adapting them to the result afterwards.

10. Common mistakes that are avoidable

  • Buying access points from an area estimate and considering capacity or cable routes only after mounting.
  • Running every radio at maximum transmit power and maximum channel width, causing your own cells to compete more heavily.
  • Surveying only with a high-end laptop and never testing scanners, phones or older IoT devices.
  • Broadcasting too many SSIDs. Every extra SSID adds management traffic and complexity to VLANs and policies.
  • Accepting automatic values without review, or changing several settings at once whenever a problem appears.
  • Using mesh as a convenient substitute for missing cabling and overlooking the capacity cost of the wireless uplink.
  • Treating a one-off heatmap as permanent even though neighbouring networks, furniture, stock and endpoints change.

11. Connect Wi-Fi to the wider architecture and operations

SSID, VLAN, firewall, switch port, PoE and radio cell form one chain, so Wi-Fi is not accepted in isolation. Guests, employees and IoT devices must land in the right networks, reach only intended services and remain traceable when changes occur. After go-live, monitor utilisation, interference, abnormal clients and channel changes. Make tuning changes individually, document them and measure again.

The companion guide Planning a UniFi network: gateway, switches, VLANs and Wi-Fi for SMEs covers gateway sizing, switching, PoE budget and segmentation. For a broader decision framework, read UniFi in business. If cameras share the infrastructure, also see UniFi Protect and workplace video surveillance. Design, surveying and installation are available through my UniFi networks for businesses service.

Sources

This article is carefully researched guidance, not legal or tax advice. For binding information, please consult your tax advisor or lawyer.

Frequently asked questions

How many square metres does one UniFi access point cover?+

There is no defensible universal figure. Construction materials, mounting height and orientation, antenna behaviour, interference, frequency band, client radios and capacity demand all change the usable cell. Derive quantity from a predictive design and verify it with a mock-up or site survey.

Does a small office really need a Wi-Fi survey?+

The more critical Wi-Fi is to voice, point of sale, production or daily work, the more important measurable design becomes. The scope can be smaller for simple spaces: inspect the plan, mock up positions and perform acceptance measurements. Skipping validation only moves the risk into operations.

Should I disable 2.4 GHz?+

Not universally. Some IoT and legacy devices require it, and the band propagates differently. The goal is to move capable clients sensibly to 5 or 6 GHz while retaining 2.4 GHz under a controlled channel plan for endpoints that actually need it.

Which channel width is best?+

That depends on cell density, neighbouring occupation, band, client capabilities and performance goals. On 2.4 GHz, 20 MHz is the sound planning starting point. On 5 and 6 GHz, balance wider channels against channel reuse and interference, then measure the result.

Does fast roaming guarantee interruption-free handoffs?+

No. Roaming remains a client decision. 802.11k, 802.11v and 802.11r can assist, but they require suitable cell overlap, compatible security and capable endpoints. Test critical movement paths with the real application.

Can Channel AI replace a site survey?+

No. Channel AI can analyse the observed radio environment and recommend channels, but it does not know every building assumption, business workflow, client limitation or acceptance criterion. It complements design and measurement, and changes still require validation.

Design, measure and accept Wi-Fi with confidence

I design UniFi Wi-Fi around the building, applications and actual clients, not a square-metre rule. From the on-site assessment and AP and channel design through to documented acceptance, the radio network has to prove itself in daily work.

Daniel Gläser

Daniel Gläser

Owner of Gläser IT-Solutions, Chemnitz

I build software and run IT infrastructure for small and medium businesses, from the first analysis to day-to-day operations. Everything here comes from real projects and is backed by sources.

More articles