Technology Trends
Every building worth tracking something inside already has Wi-Fi. That one fact is why a Wi-Fi positioning system gets shortlisted, and it is also why the shortlist so often ends in disappointment. The network was designed to move data to associated clients. Nobody installed it to work out where a wheelchair is.
The question is not whether Wi-Fi can produce a position. It can, by several methods, and some of them are now written into the IEEE 802.11 standard itself. The question is which accuracy you get for which spend, how long that accuracy survives contact with a real building, and which jobs it is simply too coarse to do.
This page answers those three questions with measured numbers and the sources they came from. Where the honest answer is “that depends on your access point density”, it says so rather than quoting a figure that came from a vendor datasheet.
What is a Wi-Fi positioning system?
A Wi-Fi positioning system works out where a device is by comparing the wireless signals it can hear against known reference data. The device has to listen to access points. It does not have to associate with them.
Four measurement methods sit underneath every product in this category. They are not interchangeable, and the difference between them is the difference between zone-level and metre-level results.
| Method | What it needs | What it returns |
|---|---|---|
| RSSI trilateration | Access point coordinates, signal strength | Coarse position from estimated distances |
| Fingerprinting | A surveyed radio map of each floor | The closest match to stored signal patterns |
| Fine timing ranging | Capable access points and client radios | A time-of-flight distance per access point |
| Angle of arrival | Multi-antenna access points, known orientation | A bearing, combined with others for a fix |
Received signal strength falls off with distance, so strength stands in for range. Walls, lift shafts, stock on shelves and bodies in a corridor all attenuate that signal, which means a weak reading is ambiguous: far away, or close but blocked. Fingerprinting sidesteps the physics by recording what the radio environment looks like at known points and matching live readings against that record. Timing methods measure the round trip directly instead of inferring it.
How accurate is Wi-Fi positioning in practice?
Published error figures for Wi-Fi positioning run from about one metre to about five metres, and the spread is explained by line of sight, channel bandwidth and how recently the radio map was surveyed. The method matters more than the supplier.
| Setup | Error in the cited study |
|---|---|
| RADAR fingerprinting testbed | 2.94 m median |
| Lateration, same testbed | 4.3 m median |
| 802.11mc fine timing, line of sight | about 1 m |
| 802.11mc fine timing, no line of sight | about 5 m |
| IEEE 802.11az design target | under 1 m |
The fingerprinting and lateration figures come from the same early testbed, reported in a review of fingerprint-based indoor positioning (arXiv, 2022): 2.94 m median for fingerprinting against 4.3 m for lateration. That ordering has held up for two decades. A surveyed radio map beats a distance model.
The timing figures are newer. A 2025 comparison of the 802.11mc and 802.11az fine timing protocols (arXiv, November 2025) reports metre-level ranging holding consistently in line of sight on 80 MHz and 160 MHz channels, and roughly 5 m once line of sight is gone. An earlier benchmark of Wi-Fi fine timing measurement (arXiv, 2023) found one to two metres outdoors in ideal line of sight, and reported that ranging error roughly halves each time channel bandwidth doubles across 20, 40, 80 and 160 MHz.
Read that bandwidth result twice, because it is the most actionable number on this page. Accuracy is partly a property of how your wireless network is configured, not only of the positioning software bolted on top. A network tuned for channel reuse in a dense office may be running narrow channels, and narrow channels cost you range precision.
What about an uncalibrated system on existing access points, with no survey and no timing support? It resolves zones. Which floor, which wing, which large room, and even that depends on having access points on more than one side of the user.
How does Wi-Fi positioning compare with GPS indoors?
Satellite positioning is not a competitor indoors, it is the reason the category exists. The official US government figure is about a 4.9 m radius for GPS-enabled smartphones under open sky, from the GPS.gov accuracy page (last reviewed February 2025), which also warns that accuracy worsens near buildings and under cover.
Inside a steel and concrete building the signal is frequently not degraded but absent. Wi-Fi fails differently, and more gracefully: it remains available and loses precision as access point density drops, multipath rises, or the radio map ages.
Regulators have put numbers on how much indoor location error is tolerable in the one case where it is life-critical. For wireless 911 calls, the United States Federal Communications Commission allows carriers to meet the horizontal requirement with either a dispatchable location or coordinates within 50 metres, phased in to 80% of calls over six years, per the Fourth Report and Order as published in the Federal Register (4 March 2015). Vertical location came later: plus or minus 3 metres for 80% of calls from z-axis-capable handsets, with deadlines of 3 April 2021 for the largest markets and 3 April 2023 for the next tier, as set out in the rules published in the Federal Register (16 January 2020).
Those thresholds are a useful calibration for private projects. A national regulator, writing rules for emergency response, settled on 50 metres horizontally and 3 metres vertically. If your internal specification demands half a metre everywhere, ask what decision actually depends on that.
What does IEEE 802.11az change?
IEEE Std 802.11az-2022 adds positioning to the standard as a first-class feature rather than a side effect of ranging timestamps, and it targets sub-metre accuracy. It was approved in December 2022 and published in March 2023, as Amendment 4: Enhancements for Positioning.
| Standard | What it adds | The practical catch |
|---|---|---|
| IEEE Std 802.11-2016 | Fine timing measurement, round-trip ranging | Both ends must support it |
| IEEE Std 802.11az-2022 | Positioning enhancements, sub-metre target | Needs newer access points and clients |
| IEEE 802.11bk | Ranging at wider bandwidth | Still arriving in shipping hardware |
| IEEE Std 802.11bh-2024 | Working with randomised MAC addresses | Changes identity, not accuracy |
A survey of 802.11mc, 802.11az and 802.11bk fine timing research (arXiv, September 2025) describes 802.11az as targeting under one metre and 802.11bk as reaching for under 0.1 m. The measured comparison cited earlier found 802.11az more accurate than 802.11mc even under heavy multipath, which is the condition that breaks signal-strength methods first.
Here is the part that gets skipped in procurement. Standards-based timing needs support at both ends of the link, on the access point and in the client device. An estate of older access points and a mixed fleet of staff and visitor phones will not deliver 802.11az accuracy because the standard exists. It delivers it where the hardware on both sides is capable, and falls back elsewhere. Any accuracy figure quoted for your building should state which fraction of devices it applies to.
IEEE Std 802.11bh-2024, Operation with Randomized and Changing MAC Addresses, addresses a different problem entirely, and the next section covers it.
Why does accuracy drift after launch?
Fingerprinting accuracy decays because the radio map describes a building that no longer exists. Access points get moved, replaced, re-channelled or powered down, partitions go up, and stock levels change. The map is a photograph; the building is not.
This has been measured rather than assumed. In Quantifying the Degradation of Radio Maps in Wi-Fi Fingerprinting (IPIN 2021), researchers surveyed and monitored a single building for more than two years and found that positioning performs best when the radio map is collected on the same day as the test data. Degradation grew over time, but produced large errors mainly when the underlying changes were substantial. The supporting measurements were released as a continuous long-term Wi-Fi fingerprinting dataset (Sensors, 2022) containing 7,435,398 Wi-Fi samples alongside 12 site surveys of the same building.
That is the finding worth carrying into a budget conversation. Radio map maintenance is not a defect of a particular product, it is a property of the method, and someone has to own it after the launch party. Ask who re-surveys, how often, and what triggers an unscheduled re-survey. “After any access point change” is the correct answer and the one nobody has a process for.
Timing-based methods are less exposed to this, since they measure distance rather than recognising a pattern. They are not immune. They still need correct access point coordinates, and a network team that relocates an access point without telling anyone has broken the reference frame either way.
Does MAC randomisation break Wi-Fi positioning?
It breaks passive visitor counting, not device-side positioning. A phone computing its own position from the signals it observes does not care what address it presents to the network. A network counting unique MAC addresses at the edge lost its identifier.
Those two architectures get sold under the same words, so separate them before comparing quotes. Device-side positioning runs in an app on the user’s phone and needs the user to have the app. Network-side analytics runs on the wireless infrastructure, needs no app, and now sees a stream of rotating identifiers instead of returning visitors. IEEE Std 802.11bh-2024 exists to let legitimate network services keep working under randomisation, which tells you how thoroughly the old assumption broke.
Privacy law reaches both. The European Data Protection Board’s Guidelines 2/2023 on the technical scope of Article 5(3) of the ePrivacy Directive, adopted in version 2.0 on 7 October 2024, treat gaining access to information stored in terminal equipment as triggering the consent requirement, whether or not the information is personal data and whether the operator stores or only reads it. Device fingerprinting has been inside that scope since the Article 29 Working Party’s Opinion 9/2014. If a deployment plan treats consent as a later phase, it is not a plan.
Is a Wi-Fi positioning system still worth implementing in 2026?
Yes, for zone-level jobs on networks that already exist, and no for anything that needs reliable metre-level positioning without new investment. The answer follows from the use case rather than from the technology’s reputation.
| Use case | Wi-Fi alone | Why |
|---|---|---|
| Floor and zone occupancy analytics | Good fit | Zone accuracy answers the question |
| Blue-dot wayfinding in a large venue | Workable with fusion | Needs sensors and map matching |
| Finding a specific asset in a hospital | Weak | Metre-level matters at the door |
| Shelf-level retail placement | Wrong tool | Below what Wi-Fi resolves |
| Floor identification for responders | Depends | Vertical is the hard axis |
The trade-offs are real in both directions. Bluetooth Low Energy beacons give tighter proximity at the cost of hardware you buy, mount, and whose batteries you replace on a schedule. Ultra-wideband anchors and tags go tighter again and cost more per square metre of coverage. Wi-Fi starts cheaper because the access points are already paid for, then charges you in surveys, maintenance and compromise.
Sometimes beacons are the right answer. An article that cannot say so is selling, not advising.
The common production design is not any single technology. It is fusion: Wi-Fi for coverage and absolute reference, the phone’s own inertial sensors for continuity between fixes, and a map that constrains impossible positions, such as a pedestrian passing through a wall. Each input compensates for a specific failure in the others.
What should you check before you commit?
Five questions decide whether a Wi-Fi positioning project lands near its promised accuracy. Get the answers in writing before the contract, because every one of them is cheap to ask now and expensive to discover later.
| Question to ask | Why it decides the outcome |
|---|---|
| How many access points per floor, and where? | Density and geometry set the accuracy ceiling |
| What channel width is in use? | Ranging error roughly halves as bandwidth doubles |
| Which devices support fine timing? | Standards-based accuracy applies only to capable hardware |
| Who re-surveys the radio map, and when? | Radio maps degrade as the building changes |
| Device-side or network-side, and under whose consent? | Decides the architecture and the legal basis |
If you are weighing Wi-Fi against beacons, ultra-wideband or a fused approach for a specific building, the useful conversation starts from your floor plans and your access point inventory rather than from a feature list. Bring both and Mapsted’s team will walk through what the existing network can support, where it will fall short, and what the gap costs to close. Book a scoping call.
Frequently asked questions
What is a Wi-Fi positioning system?
A Wi-Fi positioning system locates a device by comparing the signals it hears from wireless access points against known reference data. The device only has to listen, not to join the network. The position comes from a surveyed radio map, from distances estimated out of signal strength, or from timing measurements exchanged with access points that support them.
How accurate is Wi-Fi positioning?
It depends on method and conditions rather than on the supplier. A classic fingerprinting testbed reported 2.94 m median error against 4.3 m for lateration. Measurements of IEEE 802.11mc fine timing show roughly 1 m ranging in line of sight on 80 MHz and 160 MHz channels, and about 5 m without it. Uncalibrated deployments resolve zones, not metres.
Does Wi-Fi positioning really need no extra hardware?
Partly. Existing access points give coarse, zone-level positioning with nothing new bought, which is the honest version of the hardware-free claim. Metre-level results need more: a maintained radio map, extra access points in sparse areas, fine timing support at both ends of the link, or fusion with the phone’s own inertial sensors. The survey work is usually the larger bill.
Is Wi-Fi better than GPS indoors?
Indoors, usually yes. GPS.gov puts smartphone satellite positioning at about a 4.9 m radius under open sky and notes that buildings degrade it; indoors the signal is often unusable rather than merely coarse. Wi-Fi stays available inside and degrades gradually as access point density, line of sight and radio map freshness get worse.
Does MAC address randomisation stop Wi-Fi tracking?
It stops passive device counting far more than it stops positioning. A phone running your own app computes its own position from observed signals, so randomisation changes nothing there. Network-side analytics that recognised returning visitors by MAC address lost its identifier, and IEEE Std 802.11bh-2024 exists to let legitimate services work under randomisation.
