Wi-Fi Indoor Positioning Systems: What They Measure and How Accurate They Are

Every building with Wi-Fi already has a radio sensor on every ceiling. That is why Wi-Fi positioning is usually the first option a facilities team considers: the hardware is bought, powered and cabled, and the question is only what else it can be asked to do. The honest answer depends on which Wi-Fi measurement you use, how wide your channels are, and whether a wall sits between the phone and the access point.

This article covers what a Wi-Fi indoor positioning system actually measures, the accuracy figures published studies report, the conditions that destroy those figures, and how to decide between Wi-Fi and the alternatives before you commission anything.

What is a Wi-Fi indoor positioning system?

A Wi-Fi indoor positioning system estimates a device’s position inside a building from the radio exchange between that device and access points whose coordinates are known. It measures one of three things: signal strength, signal flight time, or signal angle.

Those three inputs behave very differently. Signal strength is cheap and available on every device made in the last twenty years, and it is also the noisiest, because a trolley, a fire door or a crowd changes it. Flight time is a physical distance measurement and much steadier, but both ends of the link have to support the protocol. Angle needs an access point with multiple antennas and the software to read phase differences across them.

A positioning system is not the same thing as a wayfinding system. Positioning answers “where is this device”. Wayfinding adds a routable map, turn instructions and a user interface on top of it. You can buy good positioning and still ship a product nobody can follow.

Why does GPS stop working indoors?

Because satellite signals arrive at the ground weakly and a building either blocks them or reflects them into multipath. GPS.gov, the United States government’s official GPS information service, puts a GPS-enabled smartphone at roughly a 4.9 m radius under open sky, and states accuracy worsens near buildings, bridges and trees.

A 4.9 m radius outdoors is already too coarse to tell a hospital visitor which of two corridors to take. Indoors it degrades further, and the vertical axis fails completely. Regulators treat that gap as a public safety problem. The FCC’s Wireless E911 Location Accuracy Requirements final rule, published in the Federal Register on 16 January 2020, adopted a vertical accuracy metric of plus or minus 3 metres for 80 percent of indoor wireless E911 calls from z-axis capable handsets, with deployment required in the top 25 markets by 3 April 2021 and the top 50 markets by 3 April 2023.

Three metres vertically is about one storey. That is the floor-level question, and it is the question GPS cannot answer.

How does Wi-Fi work out where a phone is?

By comparing a live radio measurement against a model of the building. The model is either a map of expected signal strength built by walking the site, or a set of access point coordinates against which distances and angles are solved geometrically.

Fingerprinting is the oldest approach and still the most widely deployed, because it needs nothing from the access point beyond existing beacon frames. The cost sits in the survey. Someone walks the building recording the received signal strength indicator (RSSI) from every audible radio at known points, and that map has to be refreshed when access points are moved, replaced or retuned, or when the building is refitted.

Round-trip time is different in kind. Fine Timing Measurement, introduced in IEEE 802.11mc and exposed on Android as Wi-Fi RTT, times a frame’s journey out and back and converts it to a distance. There is no survey to keep current, and no assumption that signal strength maps cleanly onto distance. There is, instead, a hardware dependency: the access points and the client both have to implement it.

How accurate is Wi-Fi indoor positioning?

Metre-level, in good conditions, with wide channels. A 2023 Sensors study, Testing and Evaluation of Wi-Fi RTT Ranging Technology for Personal Mobility Applications, reports a mean absolute ranging error of 0.85 m in line of sight and 1.24 m without it, and 1.1 m RMSE for two-dimensional positioning using five access points.

The last two rows come from Performance comparison of 802.11mc and 802.11az Wi-Fi Fine Time Measurement protocols, submitted 22 November 2025, which measured consistent metre-level ranging in line of sight on 80 MHz and 160 MHz channels and about 5 m once line of sight was lost. Channel width is not a detail. Benchmarking and Security Considerations of Wi-Fi FTM for Ranging in IoT Devices, submitted 7 March 2023, found one to two metre accuracy in ideal outdoor line of sight, and that similar indoor short-range accuracy was only reachable on chipsets running FTM over wider VHT80 channels rather than narrow 20 MHz ones.

Read those numbers as a ceiling, not a promise. They were produced with calibrated offsets, known access point positions and controlled geometry. Your building has none of that on day one, which is why accuracy has to be measured on site against surveyed ground truth rather than quoted from a datasheet. NIST makes the same point with capital equipment: its Z-Axis Test Facility, announced in May 2026, uses a 50-camera motion capture system with sub-centimetre tracking purely to generate the ground truth that indoor localisation claims are judged against.

What changed with IEEE 802.11az?

It made positioning a first-class function of the protocol instead of a side effect. The IEEE Standards Association, writing on the release of IEEE 802.11az on 31 May 2023, describes accuracy moving from 1 to 2 metres under IEEE 802.11-2016 to sub-metre, and into the region of less than 0.1 m, or about 4 inches.

The mechanism is bandwidth and better exchange design. Timing resolution improves as the channel gets wider, which is why spectrum policy shows up in an accuracy conversation. The FCC’s Unlicensed Use of the 6 GHz Band rule, published 26 May 2020, authorised low-power indoor access points and standard-power operation under automated frequency coordination, with channels up to 320 MHz. Wide, quiet channels indoors are exactly the conditions the ranging studies found necessary.

Two cautions before that figure enters a business case. The 2025 comparison found 802.11az ahead of 802.11mc mainly in multipath-rich environments with interference, which is the realistic case, not the best case. And a protocol capability only becomes a product capability when the phones in your visitors’ pockets implement it, which is a fleet question you cannot answer from the ceiling down.

Where does Wi-Fi positioning fail?

In the places where the radio path is blocked, the geometry is poor, or the infrastructure was designed for throughput rather than location. Access points are normally sited for coverage and capacity, so they end up spread evenly and sometimes in a near-straight line down a corridor, which is the worst arrangement for solving a position.

Four failure modes recur:

• No line of sight. Both studies above show error roughly tripling to quadrupling once a wall intervenes.
• Thin geometry. A position solved from radios that are nearly collinear has a large error ellipse even when every distance is accurate.
• Fleet fragmentation. Fingerprinting works on almost any device. FTM does not, and a visitor-facing app has to handle both.
• Change. A moved access point invalidates a fingerprint map silently. Nothing breaks. The map just gets quietly worse.

Beacons are sometimes the right answer. If your building has sparse access points, a Wi-Fi refresh several years away, and a need for aisle-level accuracy in one retail wing, Bluetooth Low Energy beacons in that wing will beat a reworked Wi-Fi design on both cost and schedule. Most working systems fuse two or more of these inputs rather than betting on one.

What does a Wi-Fi positioning deployment need from the building?

Known access point coordinates, usable geometry, firmware that exposes the measurement you intend to use, and a change process that keeps all three current. Most failed pilots fail on the fourth item rather than the first three.

Coordinates mean real positions on a real floor plan, including height, not a label like “second floor west”. Geometry means enough radios audible from each serviceable point, spread in two dimensions. Firmware means a vendor release that reports FTM or angle data through an interface your platform can read, on the access points you own today. Change control means someone owns the map when the estates team relocates a radio above a new partition wall.

What are the privacy rules for Wi-Fi positioning?

Device identifiers used to locate people are personal data in the European Union, and systems have to be designed on that basis. Regulation (EU) 2016/679 defines personal data in Article 4(1) to include any identifier a person can be identified by, and Recital 30 names device-supplied identifiers such as IP addresses, cookie identifiers and radio frequency identification tags.

The practical consequence is a split between two architectures. Device-side positioning computes the position on the phone, inside an app the user installed, with a lawful basis you can explain in one sentence. Infrastructure-side positioning observes devices passively from the network, which is powerful for analytics and much harder to justify, especially as MAC address randomisation now breaks naive device-level tracking by design. Regulators also attach privacy conditions to location data in safety contexts: the same 2020 FCC rule extended existing privacy protections to z-axis data conveyed with 911 calls.

Decide which architecture you are buying before you specify accuracy. It changes the answer.

How do you choose between Wi-Fi, Bluetooth Low Energy and ultra-wideband?

By writing down the accuracy you need, where you need it, on which devices, and then testing in the building rather than in a datasheet. The decision usually resolves itself once those four answers are on one page.

Run a short measured pilot in the hardest part of the building, not the easiest. Pick twenty surveyed points behind walls, in stairwells, near metal shelving and in a crowd, and record the error distribution rather than an average. If the result holds the threshold you wrote down, the technology choice has made itself. If it misses by half a metre in two zones, you now know exactly where to add anchors instead of redesigning the whole network.

Frequently asked questions

Can Wi-Fi indoor positioning work without an app?

Only in the passive, infrastructure-side form, where the network observes devices and no app is involved. That mode supports footfall analytics, not turn-by-turn guidance, because there is nothing on the device to show a route. It is also the mode most affected by MAC address randomisation and by data protection obligations. Any product that guides a person needs software on the device the person is holding.

How many access points are needed for metre-level accuracy?

The published metre-level figures were achieved with a handful of radios in favourable geometry. The Sensors study reached 1.1 m RMSE in two dimensions using five access points. Count how many are audible at each point you care about rather than how many exist per floor, and check that they are not nearly collinear, because spread matters as much as density.

Is Wi-Fi positioning accurate enough for floor-level detection?

Usually yes, and it is one of Wi-Fi’s genuine advantages, because access points sit on a known floor and a device hears its own floor loudest. Floor detection is a classification problem rather than a distance problem, so it tolerates noisier measurements. Stairwells, atria and lift shafts are the recurring exceptions, and they are where barometric or inertial inputs help.

Does Wi-Fi 6E or Wi-Fi 7 improve indoor positioning?

Indirectly, through bandwidth. Timing accuracy improves on wider channels, and the FCC’s 2020 rule opened 6 GHz spectrum for indoor use with channels up to 320 MHz. The measured studies both found that 80 MHz and above is where metre-level ranging becomes consistent. The gain arrives when access points, client devices and the positioning platform all support the wide-channel measurement, not when the access point ships.

What should a buyer do next?

Write down the accuracy threshold first, test it in the worst part of the building, and only then pick a technology. That sequence matters more than the shortlist, because a number measured on site survives a procurement argument and a datasheet figure does not.

• Write the threshold as an error distribution in named zones, not a single average for the site.
• Inventory what is already on the ceiling, including firmware versions and whether FTM is exposed.
• Name the person who owns the floor plan and the access point map after go-live.

If you are weighing Wi-Fi against beacons or ultra-wideband for a specific building, the useful next step is a conversation about that building: its floor plans, its access point inventory, and the accuracy you actually need in each zone. Mapsted builds indoor positioning, mapping and wayfinding systems, and this scoping work is where a deployment is won or lost. Book a scoping call and bring your floor plan.