Active Wireless Site Survey for Hong Kong Offices: What It Measures and Fixes

TL;DR: A coverage heatmap that is green on every floor says only that a signal is present, not that the network holds up when 110 people join video calls at 2 p.m. An active wireless site survey connects to the network, pushes real traffic and measures throughput, retries and roaming. Fix configuration first, re-test the same way, then hand the result to someone who watches it.

Picture a 110-person office on one and a half floors of a Kowloon East tower. Open-plan desks, a boardroom, a pantry, a finance corner by the window. Six months ago somebody ran a coverage survey and the heatmap came back green everywhere. The access points are three years old and were installed by the fit-out contractor. Nobody has touched the configuration since.

This office is an illustrative composite, not a named client. We use it because the pattern is common in Hong Kong towers: the survey passed, the complaints did not stop. Every number we attach to this office later in the article is this composite office's own reading, offered to show what an active survey looks like on paper, not as a market benchmark.

The office and the complaint: why is the Wi-Fi green on the heatmap but bad in real life?

The complaints in this office have a shape. Video calls are fine in the morning. Somewhere after 2 p.m. they begin to freeze, the voice turns robotic, and people drop out of Teams meetings and rejoin. Large file uploads to the shared drive stall part-way through. The finance corner, which sits nearest the window and has the densest cluster of desks, complains loudest. People have started tethering to their phones, which is a quiet vote of no confidence.

The helpdesk ticket history says "Wi-Fi slow" or "call dropped". It never says why. And the last survey result is on file: green everywhere. So the IT lead is caught between a document that says the network is fine and a floor that says it is not.

Why "add more access points" is the first wrong instinct

The reflex is to buy more access points. It feels logical: more radios, more coverage. But coverage was never the problem. The heatmap already says there is signal in every corner. Adding access points in a space where the radio environment is already crowded usually makes things worse, because every new radio is another transmitter competing for the same few channels. In a dense office, the scarce resource is not signal. It is airtime, the slice of each second in which a channel is free for someone to speak.

So the first job is not to buy anything. It is to find out what the network is doing under real load. That is what an active survey is for.

Predictive, passive, active and spectrum: what does each survey measure?

The word "survey" covers four different activities, and confusion between them is why so many offices end up with a green heatmap and a bad network. Our published guide to survey types covers all four in one place. Here we only need the differences that matter for this complaint.

Predictive survey

A predictive survey is a model. The engineer imports a floor plan, marks wall and glass materials, places virtual access points and the software estimates coverage. It is useful before installation, when there is no network to measure. It cannot know what your neighbours broadcast, how your staff actually behave, or what a pantry microwave does to the band. It is a plan, not a measurement.

Passive survey

In a passive survey the survey device listens. It walks the floor and records signal strength, noise and which access points it can hear, without connecting to anything. Brocent's wireless site survey page describes it exactly this way: the device passively listens for signal strength, noise and AP presence, without connecting. The output is the familiar heatmap, green where signal is strong and noise is low.

A passive survey is a good tool for what it measures. The weakness is what it does not measure. It never joins the network, so it never finds out what happens when a laptop actually tries to send data through that access point.

Active survey

In an active survey the device does connect. It associates to the network like a real client, then pushes real traffic and measures what happens: throughput, latency, packet retries, how it roams between access points. Brocent's page describes the active survey as the device actively connecting and simulating real user data transmission, to give a picture of throughput, latency and actual user experience.

This is the experience-level view. Where the passive survey asks "can a device hear the network here?", the active survey asks "if a device uses the network here, what does it get?".

Spectrum analysis

Spectrum analysis looks at the radio band itself rather than at Wi-Fi. It identifies non-Wi-Fi interference sources in the 2.4 and 5 GHz bands. A proprietary wireless video link, an older 2.4 GHz cordless phone base, a microwave oven: none of them speak Wi-Fi, so a Wi-Fi view never names them, but they still occupy the channel.

With Ekahau Sidekick 2, which Brocent lists among its survey tools, a single site walk captures passive, active and spectrum data together. That matters for cost and for comparison: the three views come from the same walk, at the same time, in the same radio conditions.

Why a passive heatmap can be green while the network fails under load

Here is the mechanism, in plain terms. A passive heatmap reports whether signal is strong enough and noise is low enough. It says nothing about how many other devices are talking on that channel at that moment. A Wi-Fi channel is shared. Only one device can transmit at a time on a given channel in a given area, and everything else waits. When access points on the same channel, including your neighbours' access points, can hear each other, they take turns, and each turn spends airtime.

At 9 a.m. with half the staff in, there is plenty of airtime and everything works. At 2 p.m., with most people at their desks, in calls, and with phones and laptops both connected, airtime is used up. The signal is still strong, so the heatmap is still green. But each transmission waits longer, retries more, and the experience degrades. The survey was not wrong. It was measuring something that was never the bottleneck.

What did the active survey find in this office?

The engineer walked the same floor plan with a device that connected and pushed traffic, and did it in the afternoon, when the complaints begin. Timing matters: an active survey run on a quiet morning would have looked as healthy as the passive one. The findings below are this composite office's own readings, labelled as such, to show what each problem looks like in a report.

Co-channel contention

Co-channel contention means too many access points are sharing the same channel and therefore the same airtime. In this office the survey found that the contractor had left several of the office's own access points on the same few channels, and that the landlord's and neighbouring tenants' networks were visible on every channel too. Access points on the same channel do not interfere in the sense of corrupting each other's signals. They politely wait for one another, which is worse for throughput than it sounds, because the waiting is spent doing nothing useful.

Channels too wide for the density

On the 5 GHz band, wider channels give higher peak speeds but leave room for fewer non-overlapping channels. The contractor had configured wide channels throughout, which suits a sparse install and a speed test in an empty room. In a dense office with neighbours, wide channels mean more overlap and more contention. In this composite office, a narrower channel width on the 5 GHz band allowed a better channel plan with fewer overlaps.

Sticky clients that will not roam

A sticky client is a device that stays attached to its first access point even after it has moved somewhere another access point would serve it far better. The survey device, walking from the finance corner toward the boardroom, logged the behaviour directly: it held on to a distant access point at a weak signal and low data rate instead of moving on. Real laptops do this all the time. A sticky client at a low data rate takes far longer to send the same data, and so consumes more airtime than a well-connected one, hurting everyone sharing the channel.

High retry rates in the finance corner

A retry is a frame that had to be sent again because it was not acknowledged. High retry rates mean wasted airtime. In this composite office's own reading, the finance corner showed a retry rate of roughly 28 per cent in the afternoon walk, compared with single digits near the reception end of the floor. The exact figure is not a standard. What matters is the contrast between two parts of the same floor in the same hour, and the fact that the passive heatmap showed the finance corner as green.

Legacy low data rates still enabled

Older Wi-Fi standards used very low data rates, in the range of 1 to 11 Mb/s. Many access points ship with these rates enabled for compatibility with ancient devices. The cost is that management and broadcast frames get sent at the lowest enabled rate, and a frame sent slowly occupies the channel for a long time. A few old rates left switched on can quietly eat a large share of a busy channel. In this office they were all still enabled, because nobody had changed the contractor's defaults.

Airtime utilisation saturating in the afternoon

Airtime utilisation is the percentage of time a channel is busy, whether with your traffic, neighbours' traffic, retries or interference. In this composite office the afternoon walk read about 85 per cent busy on the busiest 2.4 GHz channel, against about 40 per cent in the morning. Those are this office's own readings. A common engineering target is to keep a channel well below saturation so there is headroom for bursts, and a channel that is mostly full cannot absorb a few more video calls without everyone feeling it.

This is the number most worth explaining to a non-technical director. It converts the complaint "calls drop after 2 p.m." into a statement the business can understand: the channel is full by then.

A 2.4 GHz band doing most of the work

Finally, the survey showed that a large share of client devices were attached to the 2.4 GHz band. That band has only a few non-overlapping channels and is the most crowded, because it carries neighbours' networks, older devices and non-Wi-Fi interference. The 5 GHz band was underused. In this composite office, devices that could have used 5 GHz were still on 2.4 GHz because both bands shared one network name with no steering, and the clients simply chose the stronger signal, which was usually 2.4 GHz.

What does the spectrum view show that the Wi-Fi view cannot? Neighbours and non-Wi-Fi interference

Coexistence is the part of an office survey that the occupants cannot see from inside. You do not control the building. You share the air with the landlord, with the tenant on the next floor and with whatever they have installed.

Neighbouring and landlord networks

A Wi-Fi view lists these networks by name and shows how strong they are on each channel. This is useful, because it tells you which channels are least contested and which are hopeless. In the composite office the landlord's guest network and several neighbouring tenants' networks were visible on every channel. You cannot switch them off. You can only plan around them, and the plan has to be based on measurement rather than guesswork.

Non-Wi-Fi interference

Here the spectrum view earns its place. In the composite scenario the engineer found three kinds of sources a Wi-Fi view would never name:

• Wireless presentation system — a meeting-room wireless video link using a proprietary, non-Wi-Fi 5 GHz signal, transmitting continuously while people mirrored slides, occupying a channel the Wi-Fi view reported as quiet.
• Cordless phone and Bluetooth devices — Bluetooth headsets and peripherals, plus an older cordless phone base operating at 2.4 GHz, a steady background load on that band. DECT phones use a separate band around 1.9 GHz and are not part of this.
• A microwave oven in the pantry — an intermittent source that wipes out a slice of the 2.4 GHz band around lunch, which is exactly when people drift to the pantry end of the floor.

None of these appears in a list of networks. A spectrum analyser, such as the Chanalyzer Wi-Spy Dbx that Brocent's page names, or the spectrum capability of the Sidekick 2, shows the energy directly and lets the engineer identify the source. Without it, the fix list would have been blind to part of the problem.

How the three views become one conclusion

With co-channel contention, retries, legacy rates and interference sources on the table, the next step is not to patch them one by one but to look at them on one map. The method is to put the readings for the same spot side by side: signal strength, airtime utilisation on that channel, retry rate, and whether there is non-Wi-Fi energy nearby. The finance corner is usually worst not because of one cause but because several stack up in the same place: a channel shared with neighbours, access points waiting on each other, clients stuck to a distant access point, and a presentation device transmitting continuously beside them.

Seen together, the order of the fixes becomes clear. Whatever configuration can solve several items at once goes first; anything that needs cabling or ceiling work goes later. A good report lets its reader answer four questions: which area has the problem, what causes it, what change fixes it, and which number proves the fix afterwards.

What the survey day needs from the business

An active survey does not need staff to leave their desks, but three things need agreeing in advance. First, timing: pick the hours when the complaints happen, not the hours that suit the engineer. Second, scope: which parts of the one and a half floors are to be measured, which are off limits, and what happens when the boardroom is in use. Third, a reference point: export the current access point positions, channels and power settings before the survey, so the engineer can match the running configuration to the measured results rather than seeing only symptoms.

Agreed in advance, the survey day is a quiet walk rather than an interruption. More importantly, it makes the baseline repeatable: the re-test a few weeks later can walk the same route under the same conditions.

What is the fix list, in order of cost?

The fix list follows the order in which things should be tried, which is the order of cost. Most of the fixes cost an engineer's time and nothing else.

Configuration first

• Channel plan — assign channels so neighbouring access points do not share one and so the plan avoids the channels where the landlord's and neighbours' networks are strongest.
• 20 MHz on 2.4 GHz — in a dense space, narrower channels give more non-overlapping channels and less contention. The 2.4 GHz band is simply too small for anything wider.
• Disable legacy rates — remove the lowest data rates so that management and broadcast frames stop occupying the channel for so long, and so that devices which cannot connect at a decent rate are pushed to connect better or roam.
• Minimum RSSI and band steering, where supported — encourage sticky clients to let go of a poor access point and encourage dual-band devices onto 5 GHz. Support varies by platform and client, so these settings need testing, not assumption.
• Power adjustments — reduce transmit power so that access points hear each other less and each one serves a smaller, cleaner cell. This is the opposite of the "more power" instinct, and it often helps.
• SSID reduction — every extra network name multiplies management traffic. Consolidate to the few the business actually needs.

Placement moves second

If configuration cannot close the gap, the next step is to move an access point, perhaps from a corridor ceiling to the middle of the finance corner, or to relocate a ceiling-mounted unit that is serving an area it does not need to. This costs cabling or a few hours of an installer's time. The survey report should recommend positions, with each area measured against its own acceptance threshold, which is how Brocent's deliverable is organised.

Replacing or adding hardware last

Only after configuration and placement have been tried is it reasonable to replace or add hardware. Three-year-old access points are not necessarily obsolete, and the survey tells you which of the symptoms are due to age and which are due to configuration. In this composite office, most of the problem was configuration. If the re-test had shown otherwise, the report would have said so, with evidence.

For the supporting detail on why a survey before purchase changes the outcome, see our article on a survey before buying hardware for a Singapore warehouse.

What should the re-test show, and how should it be run?

A fix is a hypothesis until it has been measured. The client should demand a re-test, and the re-test should have these properties.

The before and after numbers to demand

• Throughput per area — measured at the same points on the same floor plan, so that the finance corner, the boardroom and the open floor can each be compared to their own earlier reading.
• Retry rate — the share of frames that needed to be resent, per area. The finance corner is the one to watch.
• Airtime utilisation — per channel, at the same time of day as the baseline. This is the number that should drop most.
• Roaming behaviour — whether the survey device now moves to a closer access point as it walks, rather than clinging to a distant one.

Same method, same time of day

The re-test must use the same method as the baseline: the same walk path, the same device, the same traffic profile, and the same hours. Comparing a quiet-morning re-test to an afternoon baseline would produce a flattering improvement that means nothing. In this composite office the baseline was an afternoon walk, so the re-test was too.

This is also where acceptance thresholds matter. Each area should be measured against its own acceptance threshold, agreed before the work, so the report says "passed" or "did not pass" against a number the client already approved. We will not claim an industry standard here. The thresholds in this article's composite office are this office's own.

Who watches it after the survey?

A survey is a measurement on one day. Afternoon load changes with headcount, with a new team moving in, with a neighbour installing a new system, and with firmware that changes behaviour after an update. The fix is true on the day of the re-test. Whether it is still true in March depends on someone watching.

That is the job of a managed controller. Brocent's Managed Wireless Network service runs a Brocent-hosted, Brocent-operated UniFi Network controller on a shared platform with tenant-scoped data. It is offered in two models: bring your own UniFi access points, or a subscription with hardware included, quoted individually, with hardware maintenance and replacement within SLA. What the service provides is the operating discipline around the configuration:

• 24×7 monitoring with alert-to-ticket — an access point that goes quiet at 3 a.m. becomes a ticket, not a Monday surprise.
• Firmware upgrades inside a maintenance window — changes happen on a schedule, not mid-afternoon.
• Daily configuration backup, retained for three years — the channel plan and the survey-driven settings can be restored if something is reset.
• A monthly availability report — a record of what happened, against a 99 per cent controller availability commitment.
• Priced per access point per month — the figure is published on the Managed Wireless Network service page.

UniFi is mentioned here only because it is the platform that service runs on. The survey findings in this article are vendor-neutral: co-channel contention, retries and sticky clients look the same on any brand.

Passive, active and spectrum: which one is for what?

Comparison of the three survey views

• Passive survey — measures signal strength, noise and which access points are heard, without connecting to the network. It cannot see throughput, latency, retries or roaming, because it never uses the network. Use it for coverage and for a quick view of what is on the air.
• Active survey — connects to the network and simulates real user traffic, measuring throughput, latency, retries and roaming. It cannot name non-Wi-Fi interference on its own, and its result depends on when it is run, so it must be done at the load that causes the complaint. Use it when coverage looks fine but the experience is not, and for before-and-after validation.
• Spectrum analysis — measures energy in the 2.4 and 5 GHz bands and identifies non-Wi-Fi sources. It cannot tell you how the network is performing for users. Use it when you suspect interference that no network list explains, or before you finalise a channel plan.

In practice the three are not rivals. They answer different questions, and with the right tool a single walk collects all three. The survey types Brocent offers include predictive, passive, active, AP-on-a-stick, post-installation validation and health-check surveys. The wireless site survey page also publishes an effort example: a passive and active survey of existing access points over about 5,000 m² is about one man-day. That is an effort figure published on Brocent's wireless site survey page, not a quote for your office; survey work is scoped and quoted per project.

On the question of certification and tooling, our separate article on Ekahau WiFi survey certification and tooling in Hong Kong covers it. The survey itself is described on the wireless site survey service page.

What does a wireless network teach about what "managed" means?

The wireless control plane is the cheapest place for a company to learn what managed IT means in practice. It has a few clear properties. Someone is watching at 3 a.m. A change goes through a window instead of in the middle of a call. A monthly report says what happened, in language that an executive can read. If any of those is missing, you do not have a managed service. You have a device with a login.

The per-user managed IT plan, covering the whole staff estate, runs on the same engine: a service desk, monitored systems, scheduled change and a report. Wireless is simply the most visible place to see it, because when it fails, everyone notices at once. If your office is dealing with a green heatmap and a bad afternoon, the survey is the first step. The plan is what keeps the result true.

Frequently asked questions

Active or passive: which survey do we need?

If your complaint is that there is no signal in some places, a passive survey will show it. If your complaint is that there is signal but the experience is bad, particularly at busy hours, you need an active survey. Most offices with a "green but slow" problem need both, which is why a single walk with a device that captures passive, active and spectrum data together is the efficient route.

How long does an active survey take?

It depends on floor area and how much of it has to be walked. Brocent's wireless site survey page gives one published example: a passive and active survey of existing access points over about 5,000 m² takes about one man-day. A one-and-a-half-floor office will not be the same as a warehouse, and the engineer will confirm the effort once the floor plan and the hours are known. Reporting and the re-test come on top of the walk itself.

Do users need to leave the floor?

No. The survey is done by an engineer walking with a survey device, with staff working as normal. In fact that is the point, because an afternoon survey is meant to capture the network under real load. Staff may notice an engineer carrying a device, but nobody needs to move or stop work.

What is airtime utilisation?

It is the proportion of time a Wi-Fi channel is busy, whether with data, retries, other networks' traffic or interference. When airtime utilisation is high, every device waits longer for its turn, so calls stutter even though signal strength looks fine. It is the single most useful number for explaining why a network with good coverage can still feel slow at busy hours.

Can we do an active survey on our existing access points?

Yes. An active survey of existing access points is a standard case and is called out on Brocent's wireless site survey page. In fact it is often the best time, because the survey shows how the network you actually run behaves, rather than how a model says a new one would. The findings then tell you whether configuration changes are enough or whether hardware needs to change.

What should the survey report contain?

At a minimum: heat maps for signal strength and co-channel interference, recommended access point placement where relevant, and each area measured against its own acceptance threshold. For an active survey, ask also for throughput, latency, retry rate and roaming results by area, with the time of day recorded, and for a clear list of recommended changes ordered by cost. A report that is only a set of coloured maps is a passive report.

Why did the first survey miss this?

Most likely because it was a coverage survey, passive or predictive, and was run before the office was full. It answered the question it was asked: is there signal? It was never asked whether the network holds up when most of the floor is in a call. A green result and a bad experience can both be true at once, which is the whole reason for the active survey.

Does an active survey need special hardware?

It needs a survey device and software able to connect as a client, push traffic and record the results alongside location. Brocent's page names Ekahau AI Pro and Ekahau Sidekick 2, AirMagnet and Chanalyzer Wi-Spy Dbx among its tools. The point is not the brand but the method: connect, push real traffic, record location, and run at the time of day that matters.

Making the fix last: from survey to managed service

The sensible sequence is a baseline active survey, a configuration-first fix, a re-test by the same method, and then a decision about who watches the network. If you want that last part handled for the whole office rather than only the wireless layer, the primary route is the per-user managed IT plan, with wireless monitoring included in or added on to that plan. The pricing page shows how the plan is structured, and contact us when you are ready to describe your floors and your afternoon.