Indoor Navigation Solutions and the Indoor-Outdoor Handover

A visitor crosses a car park, pushes through a revolving door, and walks into a mall. For the first half of that walk their phone knows where they are. For the second half it guesses. Indoor navigation solutions exist to close that gap, and the hardest part of the work is neither the outdoor half nor the indoor half. It is the few seconds at the threshold, where one positioning method stops being trustworthy and another has not yet produced a fix.

This article is written for someone who has to buy one of these systems and defend the choice. It covers what the technologies actually measure, what independent testing has found, what two regulators already require, and the questions that separate a demo from a deployment.

What is an indoor navigation solution?

An indoor navigation solution is a positioning engine, an indoor map and a routing layer working together to tell a person on a phone where they are inside a building and how to walk to somewhere else in it.

Those three parts fail in different ways, and conflating them is the most common mistake in a procurement document. A positioning engine can be accurate while the map is eighteen months out of date, in which case the blue dot is right and the route is wrong. A map can be perfect while the engine puts the visitor one floor up. Routing can be correct and still useless, if it sends a wheelchair user to a staircase.

The outdoor half of the problem is largely solved and has been for two decades. Satellite positioning handles roads, car parks and campus paths well enough that nobody builds it themselves. What a vendor is really selling you is the indoor engine and the handover between the two.

Why does GPS stop working when you walk indoors?

Because satellite positioning needs line of sight to satellites, and a roof removes it. GPS.gov’s accuracy page puts a GPS-enabled smartphone at about 4.9 m under open sky, attributing the figure to the Institute of Navigation, and states plainly that accuracy worsens near buildings, bridges and trees.

Two numbers on that page are often confused, and the difference matters. The U.S. government commits to a daily global average user range error of 2.0 m or better with 95% probability, and on 20 April 2021 the measured figure across all satellites was 0.643 m or better, 95% of the time. That is a property of the signal in space. It is not your accuracy. Your accuracy depends on satellite geometry, the receiver, and whatever is between the two.

Indoors, the same page lists the mechanisms that wreck a fix: signal blockage from the structure itself, and multipath, where the signal arrives after bouncing off a wall. The receiver still returns a position. It simply returns a bad one, usually with a large reported uncertainty attached.

That reported uncertainty is more useful than it looks. A fix with a 60 m error estimate is not a location, but it is strong evidence about which side of the wall the phone is on. Good indoor-outdoor engines treat the quality of the satellite fix as an input to transition detection rather than throwing it away.

What has to happen at the door?

Four things, in under about two seconds: detect that the transition is happening, load the right map for the right building, carry the person’s heading across the boundary without resetting it, and keep the active route unbroken so the instruction on screen never goes blank.

Detection is the part most systems get visibly wrong. Trigger too early and the app snaps indoors while the visitor is still twenty metres from the entrance. Trigger too late and they stand inside the lobby looking at a map of the car park. Motion context helps here, because a person approaching a door walks differently from a person crossing open ground, and a building with one entrance gives you a strong prior about where the crossing will happen.

Map loading is an engineering constraint, not a positioning one. If the floor plan is fetched when the visitor opens the door, the first fix waits on the network, and a crowded lobby is exactly where the network is worst. Pre-fetching the surrounding venue while the person is still outside moves that cost to a moment when nobody is watching.

Heading continuity is the quiet one. The magnetometer in a phone is reliable in a field and erratic inside a steel-framed building, so an engine that leans on the compass alone will spin the arrow on entry. Carrying a heading estimate derived from motion across the boundary, and correcting it slowly, is what makes the arrow hold still.

Which indoor positioning technology should you choose?

There is no single answer, and any vendor who gives you one without asking about your building is selling, not advising. The honest comparison is what each method measures and what independent sources report it achieving.

Wi-Fi ranging has the clearest public evidence. IEEE 802.11az-2022, the Next Generation Positioning amendment, standardises higher-accuracy ranging, and a November 2025 measurement study on arXiv comparing it with the older 802.11mc protocol found meter-level accuracy consistently attainable in line of sight on 80 MHz and 160 MHz channels, and about 5 m in non-line-of-sight conditions. Read that second figure carefully. Most of a hospital corridor is non-line-of-sight.

Ultra-wideband is standardised too. IEEE 802.15.4z-2020, published on 25 August 2020 and since folded into the consolidated IEEE 802.15.4-2024 base standard, adds coding options and encrypted ranging waveforms specifically to improve the integrity and accuracy of ranging measurements. What the standard does not give you is comparability between products: a 2022 arXiv survey of UWB standards and organisations notes that the localisation technique layered on top of the standardised ranging is mostly proprietary. Two compliant systems can perform very differently.

Bluetooth Low Energy beacons deserve a straight answer rather than a number. Signal-strength positioning works, it is widely deployed, and its accuracy is a function of how densely you install and maintain the hardware. The density figures in circulation come almost entirely from the companies selling the beacons, so there is no figure here. What you should take from that absence is the cost driver: beacons are an estate of batteries and adhesive mounts, and someone has to own them after launch.

Inertial sensing is the glue rather than a solution on its own. Step counting, gyroscope heading and barometric pressure are available on every phone and need no infrastructure at all, which is precisely why they carry the transition at the door. They also drift, so they have to be corrected against something external at intervals.

How accurate is “accurate”, and who measured it?

Independent testing of indoor positioning is rare, which makes the few public results worth more than any datasheet. The most useful is NIST Technical Note 1962, published April 2017, which evaluated a Wi-Fi-based emergency calling app in a four-storey building on the NIST campus in Gaithersburg, Maryland.

Those numbers are a healthy corrective. A system can average 7.45 m and still be 15 m out across a twentieth of the building, and it can pick the right floor nearly nine times in ten, which means it picks the wrong floor more than one time in ten. For a shopping centre that is an annoyance. For a responder entering a tower it is a different conversation, which is why NIST’s own public safety programme sets a target of better than 3 m at 95% probability without beacons, and why its Z-Axis Test Facility, announced in May 2026, was built to generate sub-centimetre ground truth for vertical positioning.

The protocol matters as much as the result. ISO/IEC 18305:2016 treats a localisation system as a black box, tests it across a whole building rather than at chosen points, and defines the building types and movement modes the test has to cover. A revision is in progress. If a vendor quotes an accuracy figure, the only follow-up question worth asking is which edition of which standard produced it, and over how much of the floor area.

One more note on numbers. The market-size projections that circulate on this topic almost always trace back to a press release summarising a paid report, with the original methodology unpublished, so none appear here. If a figure cannot be traced to a regulator, a standards body or a named study, it has no place in a business case either.

What do regulators already require?

Two separate regimes bear on indoor navigation in the United States, and both now have dates attached.

The vertical accuracy metric comes from the FCC’s Sixth Report and Order on wireless E911 location accuracy, adopted in 2020, which set a z-axis metric of within 3 m above or below the handset for 80% of calls from capable devices, and required nationwide carriers to deploy such technology nationwide by 3 April 2025, with smaller carriers given until 3 April 2026. The horizontal benchmark is separate: 50 m, or a dispatchable location, for 80% of calls. This is a carrier obligation rather than a building owner’s, but it sets the accuracy bar a 911 dispatcher is accustomed to, and it is the number your safety team will quote back at you.

The accessibility requirement is the one most buyers miss. The Department of Justice’s ADA Title II web and mobile app rule, published 24 April 2024, makes WCAG 2.1 Level AA the technical standard for state and local government web content and mobile applications. That covers a wayfinding app for a public university, a transit hub or a civic hospital.

The dates moved this year. An interim final rule published on 20 April 2026 extended compliance to 26 April 2027 for entities serving 50,000 or more people, and to 26 April 2028 for smaller entities and special district governments, as ada.gov’s fact sheet now records. The standard itself did not change. If you are shortlisting vendors for a public venue, a conformance statement against WCAG 2.1 Level AA is now a procurement item, not a nice-to-have.

How do you evaluate an indoor navigation solution before you sign?

Ask for measurements in your own building, against a named standard, at the 95th percentile rather than the average. Everything else in a pilot is theatre.

Three further things are worth writing into the pilot itself rather than discussing in a meeting:

• Walk the worst route in the building, not the showcase one. Underground parking, a lift lobby, a glass atrium and a stairwell will each break something different.
• Test on a mid-range Android handset that is three years old, because that is what a large share of your visitors are carrying, and sensor quality varies more across phones than across vendors.
• Measure battery drain over a thirty-minute route with the screen on. Continuous positioning, sensor processing and radio scanning all draw power, and a visitor whose phone dies at 40% has had a worse experience than one who never opened the app.

When is indoor navigation the wrong answer?

Sometimes it is, and a vendor who will say so is the one to trust. A building with a simple layout, clear sightlines and good static signage does not need a positioning engine. A visitor who comes once and knows the destination name may be better served by a search box and a static floor plan than by turn-by-turn guidance.

Beacons are occasionally the right answer too. If you need centimetre-grade tracking of equipment in a single room, or positioning for devices that are not phones, a dedicated anchor network built on ultra-wideband is a reasonable engineering choice, and the fact that it carries hardware cost does not make it wrong. The trade-off is genuine: infrastructure buys precision and charges rent for it in maintenance.

The case for an infrastructure-light indoor navigation solution is strongest where the building is large, the layout is confusing, the visitor is new, and the estate spans several structures that a person walks between. Campuses, hospital complexes, airports and transit interchanges all have that shape. So does a mall with a car park, which is where this article started.

Frequently asked questions

What is an indoor-outdoor navigation system?

It is a single navigation experience that works on both sides of a building’s entrance. Outdoors it relies on satellite positioning. Indoors it switches to a method that does not need satellites, such as Wi-Fi ranging, Bluetooth Low Energy, ultra-wideband or phone inertial sensors, and it keeps the map, the heading and the active route continuous while that switch happens so the user never sees the handover.

Can indoor navigation work without GPS?

Yes. Indoor positioning does not use satellite signals for the indoor portion at all, because those signals are blocked or reflected by the structure. Positioning indoors comes from radio ranging to known transmitters, signal-strength patterns, or the phone’s own accelerometer, gyroscope and barometer. Satellite data still helps, mainly as a cue for detecting when someone has crossed the threshold.

Do I need beacons for indoor navigation?

Not necessarily. Beacons are one option among several, and they trade hardware cost and ongoing maintenance for positioning stability. Infrastructure-light approaches use the phone’s sensors and existing radio environment instead, which removes the battery-replacement estate. For centimetre-grade tracking of equipment in a confined area, a dedicated ultra-wideband anchor network is still usually the better engineering choice.

How accurate does indoor navigation need to be for wayfinding?

Accurate enough to put the user on the correct floor and the correct side of a wall. For pedestrian guidance, picking the right corridor matters far more than a metre of lateral error, because a person reads the route against what they can see. Floor-level correctness is the figure to interrogate: NIST’s 2017 test of one Wi-Fi app found the floor correct across 87.6% of a building’s area.

What drives the cost of an indoor navigation solution?

Three things: whether hardware has to be installed and maintained, how many square metres of floor plan need surveying and keeping current, and how often the building changes. A venue that refits tenancies every quarter pays for map maintenance indefinitely, regardless of which positioning technology sits underneath. Ask for the cost of year three, not year one.

What should you do next?

Walk your own building with someone who has measured others, and do it before you shortlist. Every venue breaks indoor positioning in a different place, and the entrances, the lift lobbies and the underground levels are where you find out which place is yours. Book a scoping call and we will go through your floor plans, your entrances and the routes your visitors actually get lost on.