Most organisations have never had a wireless site survey. A good number of those who believe they have, haven’t — they’ve had a walkthrough with a laptop, or a predictive model built from a floor plan that nobody returned to verify.
The distinction matters, because a wireless network is the one piece of infrastructure you cannot validate by looking at it. A switch either passes traffic or it doesn’t. A wireless network can appear perfectly healthy on a dashboard while a quarter of your staff quietly work around it — tethering to 4G, moving to a different desk to take calls, assuming that video freezing is normal.
This article sets out what a professional WiFi site survey actually involves, what you should receive at the end of it, and the signals that tell you the survey you have been quoted is not one.
Signal strength is not coverage
Almost every misunderstanding about wireless design traces back to a single assumption: that a strong signal means a working network.
Your phone showing full bars tells you one thing — that it can hear an access point. It says nothing about whether the access point can hear your phone, how many other devices are competing for the same airtime, how much interference is present on that channel, or what data rate you will actually achieve when the room fills up.
Three measurements matter more than raw signal strength:
Signal-to-noise ratio (SNR). A strong signal in a noisy environment performs worse than a moderate signal in a clean one. Voice traffic wants an SNR of 25 dB or better. Signal strength alone will never reveal this.
Co-channel interference. In the 2.4 GHz band there are only three genuinely non-overlapping channels. Deploy four access points on a floor without a considered channel plan and they will spend their time waiting for each other rather than serving clients. The symptom is a network that looks well-covered and performs badly.
Channel utilisation. How much of the available airtime is already consumed — by your own devices, by neighbouring tenants, by non-WiFi sources such as microwaves and older wireless headsets. In a shared building in central London this is frequently the binding constraint, and it is invisible without a spectrum analyser.
A survey that reports signal strength and nothing else has measured the least important variable.
Coverage and capacity are different problems
Designing for coverage asks: can every corner of the building hear an access point at a usable level?
Designing for capacity asks: when 200 people are in this space and half of them are on video calls, does it still work?
These pull in opposite directions. Coverage design favours fewer access points at higher power. Capacity design favours more access points at lower power, with tighter cells so that each one serves fewer clients and channel reuse becomes possible.
Getting this wrong is the most common failure in enterprise wireless. A building surveyed for coverage, then filled with people, produces exactly the pattern we are called in to diagnose: full signal bars, unusable performance, and an IT team being told the WiFi is fine because the monitoring dashboard is green.
Which design you need is a business question before it is a technical one. A warehouse with barcode scanners and a trading floor with 300 concurrent video sessions are both “WiFi”, and almost nothing about their designs is the same.
The four types of survey — and when each applies
“Site survey” describes four distinct exercises. A competent vendor will tell you which ones your project needs and why.
Predictive survey. A model built in software from architectural drawings, with wall materials and attenuation values applied. Fast, and the right starting point for a building that doesn’t exist yet or that you cannot access. It is a design hypothesis, not a measurement — and its accuracy depends entirely on whether the material assumptions match reality.
Passive survey. Walking the space collecting data without associating to the network, recording what is actually there: signal levels, noise, channel usage, interference from neighbouring networks. This is how you establish the real RF environment rather than the assumed one.
Active survey. Associating to the network and measuring genuine throughput, retry rates, packet loss and roaming behaviour as you move. This is what tells you whether the network performs, as opposed to whether it exists.
AP-on-a-stick. Temporarily mounting an access point on a pole at candidate positions and measuring the result before anything is installed. Slower and more disruptive, and indispensable in difficult environments — warehousing with high racking, listed buildings, spaces with heavy machinery or unusual construction.
A predictive model alone, presented as a finished survey, is the single most common shortcut in this industry. It is not dishonest — it is genuinely useful work — but it is a prediction, and predictions in RF are wrong often enough that verifying them is the job.
What you should receive
The deliverable is where the difference between a real survey and a superficial one becomes obvious. A professional survey produces documentation an engineer could build from and an auditor could check. At minimum:
- Heat maps for more than one variable. Signal strength, yes — but also SNR, channel overlap, expected data rate and retry rates. A single green-blob signal map is a marketing image, not a design.
- A channel and power plan. Which access point runs on which channel at what transmit power, with the reuse pattern shown. This is the part that determines whether the network scales.
- Access point placement with mounting detail. Position, orientation, antenna type, mounting method, and the reasoning. “Ceiling, roughly here” is not placement.
- A cabling and PoE schedule. Cable routes, containment, patch panel positions, and the switch PoE budget. Modern WiFi 6E and WiFi 7 access points frequently require PoE+ (802.3at, 30 W) or 802.3bt — a design that ignores this arrives on site and stops.
- Requirements documented up front. Device counts and types, applications and their tolerances, density per area, roaming requirements, and the design targets chosen to meet them. If nobody asked you these questions, nothing that follows can be right.
- A validation survey after installation. The step most often omitted and the one that closes the loop: a post-installation survey confirming the built network matches the design, with the evidence attached.
That last item deserves emphasis. A design is a prediction; a validation survey is proof. Without it you have paid for an opinion.
What this looks like in practice
Consider a four-floor office fit-out in central London — the kind of project we see regularly. The predictive model, built from the architect’s drawings, calls for eighteen access points.
The passive survey then finds two things the drawings could not show. First, the neighbouring tenant’s network is saturating several 5 GHz channels through a shared partition wall, removing part of the assumed channel plan. Second, a glazed partition specified late in the fit-out is metal-backed, attenuating far more heavily than the generic “glass” value used in the model.
Neither is exotic. Both are invisible on a floor plan. Together they change access point count and placement across two floors.
Had the predictive model been built as drawn, the outcome would have been coverage that measured correctly on a signal map and performed poorly in the meeting rooms — and the cost of correcting it after ceilings are closed and cabling is terminated is many times the cost of the survey that would have caught it.
Illustrative composite of typical London fit-out conditions, not a specific client engagement.
Eight red flags
If a proposed survey shows several of these, you are not buying a survey:
- No requirements conversation. Nobody asked about device counts, applications, or density before quoting.
- A phone or laptop app as the primary tool. Consumer apps report client-side signal strength. They cannot measure spectrum, and they are not calibrated.
- Signal-strength heat maps only. No SNR, no channel overlap, no data-rate mapping.
- No spectrum analysis. Non-WiFi interference is invisible to WiFi-only tooling, and it is often the actual fault.
- A hardware recommendation before measurement. If the access point model and count were decided before anyone visited, the survey is being used to justify a quote.
- Surveying an empty building for a full one. Human bodies attenuate 2.4 and 5 GHz measurably. A survey conducted in an empty office at the weekend does not describe Tuesday at 10am.
- No validation survey offered. The design is never checked against what was actually built.
- No named engineer or certification. Wireless design is a specialism. Ask who is doing the work and what they are certified in.
Not sure where your network stands?
We have turned the questions above into a one-page checklist. Twelve items, about two minutes, and at the end you will know whether your wireless network has ever genuinely been measured.
Where to start
If you are not sure which category your current network falls into, the honest first question is simply whether anyone has ever measured it. In most organisations we speak to, the answer is no — and that is not negligence, it is that wireless failure is quiet. It shows up as workarounds and lowered expectations rather than outages.
Two practical steps:
Find out when your last survey was, and what it produced. If the answer is “we don’t have one”, or the document is a single signal-strength map, you have no baseline. If it is more than three years old, it predates changes to your device population, your headcount and your neighbours’ networks.
Establish what the network is actually required to do before anyone quotes you hardware. The requirements determine the design. Everything else is sequence.
Seninix is an Ekahau-certified wireless practice. We have spent over twenty years designing, installing and managing enterprise network infrastructure, and we work across Aruba/HPE, Cisco, Meraki and Juniper platforms — which means the recommendation comes from the measurement rather than from a vendor relationship.
We offer a free 30-minute infrastructure review for organisations in London and the South East. We look at what you are currently running, what it is being asked to do, and where the risk sits. No obligation, and no requirement to buy a survey at the end of it.
