How to Evaluate an Indoor Positioning Technology Company

Most buyers start this search with a shortlist and no way to read it. Every company in indoor positioning publishes a sub-metre accuracy figure, a logo wall, and a diagram of dots moving through a shopping centre. None of that tells you whether the system will work in your building, what it will cost over five years, or whether your legal team will sign it off.

This article is a method for reading that market. It covers what the technologies actually are, how accuracy claims are constructed, which questions separate a system that survives contact with a real building from one that does not, and what the law requires of you once location data starts flowing. It does not contain a vendor shortlist, and it is written by one of the companies in the market, so read the technical sections as the part you can verify independently and treat the rest accordingly.

What is an indoor positioning technology company?

An indoor positioning technology company builds systems that determine where a person or an asset is inside a building, where satellite signals do not reach. The output is a coordinate on a floor plan, usually fed into wayfinding, asset tracking, analytics or safety workflows.

Satellite positioning fails indoors for a measurable reason. The GPS signal arrives at the Earth’s surface at roughly -158.5 dBW for the L1 C/A civil signal, as specified in the official GPS Standard Positioning Service Performance Standard (5th edition, April 2020, published by the US Department of Defense). That is a signal weaker than the background thermal noise it is extracted from, and a concrete floor slab or a metallised glass façade is enough to end it. Indoor positioning exists because something has to replace a signal that is not there.

Companies in this market divide along three lines, and the divisions matter more than brand:

• What the system measures. Radio signal strength, radio time-of-flight, magnetic field distortion, inertial motion from the phone’s own sensors, or a fusion of several.
• What it requires you to install. Nothing, a mesh of battery beacons, powered anchors with cabling, or a reuse of Wi-Fi access points you already own.
• Who is being located. A visitor holding a phone, a staff member holding a scanner, or a wheelchair carrying a tag. These are different engineering problems with different error budgets.

A company that answers all three well for a warehouse may be the wrong answer for a 400,000 square foot airport terminal. The question is never which vendor is best. It is which vendor’s physics matches your building.

Which indoor positioning technologies are actually on the market?

Five technology families account for almost all deployed indoor positioning: Wi-Fi, Bluetooth Low Energy, ultra-wideband, magnetic and inertial sensing, and visual or LiDAR-based positioning. They differ by roughly two orders of magnitude in both accuracy and cost per square metre.

The accuracy bands above are the ranges reported across published deployments and standards documentation, and they are bands rather than numbers for a reason explained in the next section. The infrastructure column is the one that drives your budget.

Two of these have formal standards you can read yourself rather than take from a sales deck. Ultra-wideband physical layers are specified in IEEE 802.15.4 (IEEE Standards Association, 2024 edition), which defines the high-rate pulse repetition frequency modes that UWB ranging depends on. Wi-Fi fine timing measurement, the mechanism that lets an access point measure round-trip time to a client rather than guessing distance from signal strength, is specified in IEEE 802.11 (2024 edition). If a company claims time-of-flight Wi-Fi positioning, ask which 802.11 amendment their ranging uses and which client chipsets support it. The answer is short and verifiable.

Bluetooth Low Energy direction finding, which uses angle-of-arrival rather than signal strength, is defined in the Bluetooth Core Specification from version 5.1 onward. It is a meaningful capability improvement over RSSI trilateration and a meaningful increase in hardware requirements, because it needs multi-antenna arrays rather than the single-antenna beacons that most existing deployments use.

How should I read an accuracy claim?

Treat any single accuracy number as incomplete until you know three things: the percentile it represents, the environment it was measured in, and whether it describes the distance error or the probability of being on the correct floor. A vendor quoting “sub-metre” without those three is quoting a best case.

Here is why the percentile matters. Suppose a system has a mean error of 1.2 m. That is a usable number for asset search. But if the 95th percentile error is 11 m, then one in twenty position fixes puts a visitor in the wrong shop, and a wayfinding route rebuilt from that fix will tell them to turn the wrong way. For navigation, the tail of the distribution is the product. For heat-map analytics aggregated over thousands of visits, the mean is closer to what matters. Ask for the full error distribution, or at minimum the median, 90th and 95th percentiles.

Environment matters because radio positioning is a measurement of the building as much as of the device. A 40-metre open atrium, a steel-racked warehouse with ten-metre aisles, and a hospital corridor with lead-lined radiology walls produce different multipath behaviour from identical hardware. A figure measured in an empty exhibition hall does not transfer. Ask where the number was measured and what the floor plate looked like.

Floor detection is the error that embarrasses deployments and rarely appears in a datasheet. Horizontal accuracy of 2 m is irrelevant if the system places the user on level 3 when they are on level 2, because the route it computes is wrong in its entirety, not slightly wrong. Barometric pressure sensing helps and drifts with weather. Ask for floor-detection accuracy as a separate percentage.

A worked example of why percentiles matter

A 50,000 square foot retail floor with a mean positioning error of 2 m and a 95th percentile of 9 m will, across 10,000 daily position fixes, produce roughly 500 fixes with error above 9 m. If each of those triggers a wrong-direction instruction, you have 500 frustrated visitors a day on a system whose datasheet says 2 m. The arithmetic is trivial. The datasheet that lets you do it is what you are shopping for.

What does indoor positioning cost over five years?

The licence fee is usually the smallest line. Over five years, the dominant costs are survey and calibration labour, hardware replacement, and the integration work to connect positioning output to the systems that consume it.

Site survey and calibration is physical labour: a person walking the building collecting reference measurements, at a pace set by floor area and layout complexity. It recurs whenever the building changes, which in retail means every tenant refit.

Hardware purchase varies by two orders of magnitude between reusing existing access points and cabling powered anchors through a ceiling. Battery or anchor maintenance is the cost nobody models: a beacon mesh is a population of batteries with a finite life, and replacing them is a ladder and a work order, not a software update.

Map and content updates are continuous in any building with changing tenants or room assignments. Software licence is the only line most proposals make visible.

Ask every company for a five-year total, itemised, with the labour assumptions stated. Ask specifically who performs re-surveys after a refit and whether that is included or billed. A proposal that only prices year one is not a proposal, it is an opening position.

What questions expose a weak vendor?

Ask for things that are expensive to fake: a failure case in their own words, the error distribution rather than a single number, and a reference site with a building geometry close to yours. Vendors with real deployments answer quickly. Vendors without them reframe the question.

Three that test the engineering:

• “Where does your system perform worst, and why?” Every positioning technology has a hostile environment. A vendor that cannot name its own is either new or not being straight with you.
• “Show me the error distribution, not the mean.” The answer is a chart or it is not an answer.
• “What happens when the phone denies location permission?” Mobile operating systems let users grant coarse location, deny it, or allow it only while the app is in use. A system that silently stops working under a denied permission is a support burden you will own.

Three more that surface what arrives after signature:

• “What is the floor-detection accuracy, as a separate figure?” Not folded into one headline number.
• “Who owns the collected location data, and where is it processed?” A contract question with legal consequences, covered in the next section.
• “What is the re-survey trigger and who pays for it?” This is where multi-year costs hide.

One trade-off worth stating plainly: the technologies that need no infrastructure are also the ones most dependent on the device in the user’s hand and on a calibration that ages. The technologies with the tightest accuracy need the most cabling. Anyone telling you they have eliminated that trade-off entirely should be asked to explain the physics, not shown a demo.

What does the law require once you collect location data?

Precise location data is treated as sensitive in major privacy regimes, which means consent and purpose limitation are legal obligations rather than product features. The specifics differ by jurisdiction and they are enforceable.

Under the EU General Data Protection Regulation, location data is explicitly named as personal data. Article 4(1) of the GDPR (Regulation (EU) 2016/679, in force since 25 May 2018) defines personal data to include an identifier such as “location data”. Article 5 requires purpose limitation and data minimisation, and Article 35 requires a data protection impact assessment for systematic monitoring of a publicly accessible area on a large scale. An indoor positioning deployment across a shopping centre is a reasonable candidate for that assessment.

In California, the California Consumer Privacy Act as amended by the CPRA (California Privacy Protection Agency regulations) classifies precise geolocation as sensitive personal information, which carries a consumer right to limit its use. The statutory definition of precise geolocation is a radius of 1,850 feet or less, so almost any indoor positioning output falls inside it.

Separately, if your deployment serves a public entity in the United States, the Department of Justice rule on web and mobile accessibility applies. 28 CFR Part 35, as amended in April 2024, requires state and local government entities to conform to WCAG 2.1 Level AA for their web content and mobile apps. The compliance deadlines are 24 April 2026 for public entities with 50,000 or more population and 26 April 2027 for smaller entities and special district governments.

A wayfinding app run by a public hospital, transit authority or university is covered. Ask any vendor for their WCAG 2.1 Level AA conformance statement, published by the W3C as a W3C Recommendation in June 2018, and ask whether it covers the map interface itself or only the surrounding pages. Maps are where accessibility conformance usually breaks.

Three questions to put in the contract rather than the sales call: where is position data processed and stored, what is the retention period, and is any of it shared with a third party for advertising or measurement. The third question is the one that changes your legal exposure most and the one least likely to be volunteered.

How do I run a fair evaluation?

Run a paid pilot in your hardest space, define the success metric before the pilot starts, and measure it yourself against ground truth you control. A demonstration in the vendor’s environment tells you about their environment.

Three steps that produce a decision rather than an impression:

• Pick the worst building, not the flagship. The basement with the steel ceiling, the atrium with no mounting points. If a system works there it works everywhere else you own.
• Write the metric down first. For example: 90th percentile horizontal error under 4 m, floor detection correct in 98% of fixes, route recalculation within 2 seconds of a wrong turn. Agreed in writing, before anyone installs anything.
• Collect your own ground truth. Walk a marked route with known coordinates and compare logged positions against it. This is a day of work and it is the only data in the process that nobody sold you.

Two more decide what you keep once the pilot ends:

• Pilot in the real device population. Test on the oldest phone your visitors actually carry, not the newest one in the demo bag.
• Price the exit. Ask what happens to your maps, your historical data and your installed hardware if you do not renew. The answer tells you how the relationship is built.

Frequently asked questions

How accurate is indoor positioning in practice?

It depends on the technology and the building. Ultra-wideband systems using the IEEE 802.15.4 physical layer reach 10 to 30 centimetres with powered anchors. Bluetooth Low Energy beacon systems typically land between 1 and 5 metres. Wi-Fi signal-strength positioning ranges from 3 to 15 metres depending on access point density. Always ask for the 95th percentile error in a building like yours, not the mean in an ideal one.

Do I need to install beacons?

Not necessarily. Beacon-free approaches use the phone’s magnetometer and inertial sensors, or reuse existing Wi-Fi access points. They avoid battery maintenance and installation labour but depend more heavily on calibration quality and on the sensor behaviour of the devices your visitors carry. Beacons remain the right answer where you need consistent few-metre accuracy across device generations and can absorb the maintenance.

Is indoor positioning legal under GDPR?

Yes, with obligations. GDPR Article 4(1) names location data as personal data, so you need a lawful basis, purpose limitation under Article 5, and in many public-venue cases a data protection impact assessment under Article 35. The deployment is lawful when consent and retention are handled properly. It is the data sharing arrangements, not the positioning itself, that most often create exposure.

How long does a deployment take?

The software is rarely the constraint. Mapping and surveying scale with floor area and layout complexity, and integration with existing systems such as a CMMS, a patient flow platform or a retail app adds its own timeline. Ask for a schedule broken into mapping, calibration, integration and acceptance testing, with the labour days stated for each, rather than a single go-live date.

What is the difference between indoor positioning and indoor mapping?

Indoor mapping is the digital representation of the space: floor plans, points of interest, connectivity between levels. Indoor positioning determines where someone is within that representation. Wayfinding needs both, because a route is a path computed across a map from a known position. A company may be strong at one and weak at the other, so evaluate the two capabilities separately.

Can one system handle both people and assets?

Often, but the requirements differ. Locating visitors means working through their own phones, with no control over hardware and with permission prompts in the way. Locating assets means tags you choose, with a known battery and a known refresh rate. Systems built for one are frequently retrofitted for the other, and the retrofit is where accuracy claims slip. Ask for separate figures.

What is the right next step for your building?

If you are evaluating indoor positioning for a specific building, the useful next step is a conversation about that building’s geometry, device population and compliance obligations rather than a generic demo. Book a scoping call and bring a floor plan.