Comparison and selection: what to take and what will it cost

If you are only reading this article, you need three things: a summary table, a scenario similar to your object, and a pilot plan. They are here.

If you have completed the entire series, everything that was disassembled in it is collected here, in a form suitable for technical specifications and for talking with suppliers.

One note about the numbers. The following are the cost orders from open sources — foreign reports and materials from manufacturers. They are suitable for estimating the scale and cost structure, but not as an estimate: the cost of a particular project is determined by the area, design, required accuracy and the state of the infrastructure of the facility. The rule from the methodical article also applies here: a digit without conditions is not a digit.

Summary table: All technologies on one screen

Technology	What measures	Accuracy	Label	Smartphone	Recognizes a person	Water and steam	Price order	Regime in the Russian Federation	Maturity 2026
Turnstile with a bracelet	Passages, events	Point of passage	Need	I don't	Yes	Great	Low	The usual	Mature
RFID HF (wristband)	Events at reading points	Reading point	Need	I don't	Yes	Great	Low	Normal, without frequency permissions	Mature
RFID UHF (portal, grid)	Zone, presence	Zone, declared up to 0.6 m	Need	I don't	Yes	Worse for water and metal	Medium	Check the band 866-868	Mature
3D counter, flight time sensor	Flow, employment	93-99 % of the account was declared	I don't	I don't	I don't	OK, except direct sun	Low — medium	As a rule outside the PD	Mature
LiDAR	Flow, position in volume	95-99 % of the account was declared	I don't	I don't	I don't	Good	Medium	As a rule outside the PD	Mature
Radar 60 GHz	Presence, posture, fall	Presence, 0.5–2 m	I don't	I don't	I don't	OK, execution under the conditions	Low — medium	As a rule outside the PD	Average
Bluetooth RSSI	Zone, proximity	In fact, the zone	Need	Opts.	Yes	Moderately, not under water	Low	The usual	Mature
Bluetooth AoA	Coordinate	Declared 0.3–1 m	Need	I don't	Yes	Moderately	Medium — high	The usual	Mature
Bluetooth Channel Sounding	Distance, coordinate	0.3–1 m, worse in a real situation	Need	Opts.	Yes	Moderately	Medium	The usual	Early
UWB	Coordinate	10-30 cm in good conditions	Need	I don't	Yes	Good	High	Check the lanes by GKRCh	Mature
Wi-Fi, service requests	Flow (unreliable)	5-15 m, ID destroyed	I don't	Need	I don't	—	Low	Controversial and binding	Degraded
Wi-Fi RTT (802.11az)	Distance	~1 m in line of sight, ~5 m without it	I don't	Need a plus app	I don't	—	Low	The usual	Early
Wi-Fi Sensing (802.11bf)	Presence, movement	Presence	I don't	I don't	I don't	Unknown	Low	Requires study	Early
Video and AI	Flow, zones, routes	Percentage units at the entrance; routes are unreliable	I don't	I don't	Optional	Bad in steam and on the water	Medium — high	Normal; faces — biometrics	Mature
Specialized drowning detection	Critical event	The time before the alarm is tens of seconds	I don't	I don't	I don't	Main application	Medium — high	The usual	Average
Guest Application	Coordinate by agreement	1-5 m	I don't	Need	Yes	The phone is in the locker	Low on iron	Consent	Mature
GNSS and LoRaWAN	Position outside	Declared 10-20 m outside	Need	I don't	Yes	Outside	Medium	The usual	Mature

The selection tree

Three questions from the first article and the conclusion to which they lead.

1. Can I give the guest a carrier?

No → counters and sensors at the entrances and in the zones; video where allowed; application for those who install it. The routes of a particular person are not available, and this must be accepted.

Yes - > moving on to the second question.

2. Do I need a personality reference?

No → impersonal bracelet: events and routes without personal data in the analytical loop. The undervalued and most relaxed legal regime.

Yes → the bracelet associated with the contract; the accounting contour is separate from the analytical one.

3. Need a real-time reaction?

No → events on reading points plus counters. Cheap, reliable, enough for most management tasks.

Yes, by zones → add Bluetooth receivers in key zones or presence sensors.

Yes, with the exact coordinate → UWB in those zones where it is really needed, the rest is zonal.

And a security question on top of everything: will your decision change if the accuracy is ten times worse? If not— buy a cheap option.

Nine typical scenarios

1. Thermal spa complex, 3000 m2, wet areas, bar, steam rooms

Tasks: zone loading, queue at the entrance, bar conversion, locker turnover, time in hot zones, surcharges, child search.

Stack: HF passive bracelet for everyone + readers at seven to eight points (entrance, wet zone, lockers, bar, steam room, premium zones, exit, surcharge terminal); 3D counters at the entrance group and at the entrances to large areas; radar presence sensors in the locker room, showers and dressing room; video surveillance in the lobby, at the reception, in the cafe; active Bluetooth tags selectively for children and staff; analytical contour in the accounting database with the integration of cash registers and locks.

What not to take: UWB for the entire area, Wi-Fi analytics (phones in lockers), face recognition, cameras in wet areas.

Where the money is: readers and cable, meters, integration. The biggest return is from what already exists: the events of bracelets and locks.

2. Water Park

Tasks: safety on the water is a priority; capacity of the slides; capacity of the bowl; search for a child; deposit and surcharges.

Stack: the same as the thermal spa complex, plus specialized detection of drowning on the bowl; active tags for children; counters at the entrances to the slide zones; enhanced exit control.

What not to take: re-identification by video (all in swimsuits), end-to-end routes by cameras.

Where the money is: the water safety system and the video infrastructure for it.

3. Fitness club

Tasks: actual attendance of halls and classes, loading of zones at peak times, turnover of lockers, retention.

Stack: club card or bracelet + readers at the entrance and in the halls; counters at the entrances to the zones; Bluetooth receivers in the halls, if you need attendance without attachment; integration with 1C:Fitness and schedule.

What not to take: coordinate systems, facial biometrics (decided by the card).

4. Sauna or small SPA, up to 300-500 m2

Tasks: loading by the hour, visit time and surcharges, security in the steam room.

Stack: bracelet or card, readers at the entrance and in the steam area, one or two presence sensors, counter at the entrance. Everything.

What not to take: almost everything else. For an object of this size, analytics is collected from events, and hybrids are redundant.

5. Amusement park, quest, immersive show

Tasks: script synchronization, group skipping, queues, security on the territory.

Stack: UWB in scenes where the effects are tied to the position (as in the disassembled case — on the performers, not on the audience); bracelets with reading points on the locations; virtual queue on the schedule; GNSS and LoRaWAN in the open area; counters at the entrances to the pavilions.

What not to take: an attempt to link effects to each guest on the entire area is a project of a different budget.

6. Museum, exhibition, Congress center

Tasks: attendance of halls, popularity of exhibits, routes, content.

Stack: counters at the entrances to the halls (the basis); Bluetooth beacons plus a content application; QR checks as an alternative to the application; video analytics in large spaces.

What not to take: Wi-Fi analytics, facial recognition.

7. Shopping mall, food court, HoReCa

Tasks: passability, conversion of zones and tenants, peaks, the effect of events.

Stack: 3D counters at the entrances and at the anchor zones — the foundation; LiDAR or radar in the atriums; video analytics selectively; guest Wi-Fi with authorization as a channel for working with the audience (with a full legal outline).

What not to take: passive collection of MAC addresses and de-randomization.

8. Workshop, warehouse, production site

Tasks: time on sites, rapprochement of man and technology, presence in hazardous areas, search for tools and equipment.

Stack: UWB in hazardous areas and where a coordinate is needed; Bluetooth zones in the rest of the area; UHF portals for tool delivery; radars in hazardous areas; local act and familiarization of workers before launch.

What not to take: coordinate coverage of the entire area, if decisions are made by zones.

9. Clinic, hospital

Tasks: search for equipment, presence and fall of patients, time of procedures.

Stack: Bluetooth tags on equipment plus receivers in the corridors (the most payback); radars in wards and bathrooms; access events by zones.

What not to take: video surveillance in the wards.

Economics: the structure of the cost of ownership

Where does the money go

Article	Share	Note
Survey and project	Small, but critical	Without it, the arrangement of equipment will be incorrect
Equipment (sensors, readers, anchors, cameras)	Noticeable, but not the main one	It is she who is shown in the commercial offer
Cable, power supply, installation, commissioning	Basic for coordinate systems	According to open estimates, 60-85 % of the project budget
Server side and storage	Average	For video — essential: archive, storage, uninterruptible power supply
Licenses and subscriptions	Average, recurring	Count for three years, not a year
Integration and analytics	Undervalued	The part without which data does not turn into solutions
Consumables	Permanent	Bracelets lose and break; batteries of active tags
Maintenance	15-25 % of the cost of equipment per year according to open estimates	Including response to sensor failures
Training and regulations	Small	And at the same time determining whether the system will work

Open reference points by order

All figures are from external sources, indicating the nature of the assessment:

• anchor UWB — 120-350 € plus installation, according to the RTLS cost summary;
• UWB infrastructure for submeter accuracy — 15-40 € per m2, approximately 30-60 anchors per 5000 m2 (same summary);
• the anchor density is one per 150-300 m2 according to Pozyx recommendations;
• the UWB tag is $ 40-100 in one estimate, up to £ 100-300 per equipped unit in another;
• cable work and survey — from 60 to 85% of the budget according to various open estimates;
• The three—year cost of ownership for a fleet of 500 tags is $ 350-800 thousand plus 15-25% per year for service, according to an estimate of 2026.

Three rules that save the most

The area is more expensive than objects. It is cheaper to equip two hundred tags per 1000 m2 than twenty tags per 10,000 m2. Reducing the area where high accuracy is required is the main lever of economy.

Zone cheaper coordinates. The transition from "know the point" to "know the zone" reduces the number of receivers by a multiple with the same management result for most tasks.

What is already worth it is cheaper than the new one. The events of bracelets, locks, turnstiles and ticket offices at most facilities are already being collected and are not being read. The first stage of almost any project should start here.

The "constraint - > bypass" matrix

Restriction	Cheap bypass	Expensive bypass
RFID does not give a continuous track	More reading points in key locations	Active tag in the bracelet, receivers by zone
Bluetooth in terms of signal strength gives only a zone	Adopt a zonal model and build reports on it	Angular determination or phase measurement of distance
UWB is expensive due to the density of anchors	UWB only in critical areas, zonal scheme in the rest of the area	Full coverage; wait for equipment with a narrowband assistant
The label is not readable underwater	Treat the disappearance as a separate state, rely on reading points	Hybrid with sensors in the bowl area
MAC Randomization broke Wi-Fi Analytics	Replace the task: counters at the inputs	Guest Wi-Fi with authorization and full legal outline
The guest's phone is in the locker	Refuse scenarios on a smartphone in wet areas	— (not complete)
The camera goes blind in a pair	Take the camera out of the steam zone, change the angle	Radars and time-of-flight sensors in the zone
Cameras are not allowed in the locker room	Sensors of presence without an image, locks, deposit	Combination of radar and the second physical principle
Re-identification does not work in swimsuits	Collect a route by bracelet	— (video does not solve this problem)
The counter accumulates a busy error	Daily reset and reconciliation with the turnstile	Bi-directional sensors with overlap
Fall detection is immature	Time monitoring in the zone by wristband plus staff bypass	Radar plus floor vibration, simulation tests
The application has installed percentage units	QR checks without installation	A loyalty program with real value
False alarms extinguish trust	Separate channels of critical and non-critical events	Customizing the model for the object, repeated iterations
Data does not turn into solutions	One screen, five digits, one action	Analytical circuit and regular cycle of experiments

Pilot in 4-8 weeks

The only way to find out the real numbers on your object. A plan that can be inserted into the contract.

Week 0. Preparation. Identify one or two tasks that we check, and the criteria for success in numbers. Prepare a plan of the object with zones. Assign a responsible person on your side. Mark up control points.

Weeks 1-2. Deployment. Equipment — for rent or on a return basis. The minimum configuration is for one or two zones, not for an object. Basic measurement of what happened before: manual counting, unloading of existing events.

Weeks 3-5. Operation and measurements. Acceptance measurement according to the protocol from the methodical article: static by control points, routes, empty object and rush hour. For security systems — tests with simulated events. Daily log: alarms, failures, false alarms, staff comments.

Weeks 6-8. Conclusions. Comparison with criteria. Reconciliation with an independent source. Calculating the cost of a full deployment based on the actual density of the equipment, not an estimate. Solution: deploy, change the configuration, or refuse.

What must be included in the pilot agreement:

• who pays for the equipment and installation during the pilot, and what happens when the project is abandoned;
• The measurement protocol and success criteria are in numbers;
• Access to raw data, not just vendor reports;
• conditions for the transition to full implementation and fixing the price of equipment based on the results of the pilot;
• what is considered an unqualified pilot.

Separately: the pilot must include rush hour. The pilot, conducted on a weekday at a quarter load, did not check anything that is important.

Antipatterns

Bought accuracy, which is not used. The most expensive mistake in the topic. Verification is a control question about a tenfold deterioration in accuracy.

Bought data instead of solutions. The system is working, the dashboard exists, decisions are still made intuitively. It is treated by the appointment of the system owner and a cycle of experiments.

We put the equipment where it is impossible. Cameras in locker rooms and showers. It turns out at the first complaint.

They took "counting unique visitors" without reading how it works. If it is based on facial recognition, you have bought biometrics along with the entire legal regime.

They believed the accuracy figure from the presentation. Without percentile, conditions and density of equipment, the figure is unverifiable.

Missed the basic measurement. Now it is impossible to prove the implementation effect — neither to yourself nor to the owner.

Not associated with the accounting system. Four sources give four different attendance figures, and instead of working with the data, there is a dispute about which one is correct.

Did not foresee who is serving. A year later, a third of the sensors are silent, and no one knows about it, because there is no equipment status screen.

We chose equipment not for Russian bands. UHF readers in the US version; checked before purchase, not after.

Saved on the examination. Arrangement "by area" instead of arrangement by design — and then adding anchors and readers is actually more expensive.

Where is everything going in 2026-2028

Seven areas to keep an eye on if the planning horizon is more than a year.

Phase measurement of distance in Bluetooth will become widespread. The specification is being developed (Core 6.0 - > 6.2 - > 6.3 in May 2026), the chips are serial, research is underway on unconnected modes for a large number of labels. Practical consequence: an ordinary bracelet will be able to give a distance, not just a zone. What to do now: choose tags on chips with support so as not to change the fleet.

The new generation of UWB will reduce the number of anchors. The 802.15.4ab standard with a narrowband assistant, the first chips (ST64UWB, March 2026) and research receivers with a claimed fourfold increase in range address the main drawback of UWB — the cost of dense infrastructure.

UWB chips will also become radars. In the new families, the radar sensing function appears. This brings positioning and presence detection closer together: one device instead of two.

Wi-Fi Sensing will reach certification. The standard has been published, the chips have appeared, there is no Wi-Fi Alliance certification yet. When it appears, "presence without a device and without a camera" will become a regular function of network equipment — with a separate legal and ethical conversation.

Multimodal models will move closer to the object. So far, they are suitable for searching through the archive and analyzing incidents. As the models decrease and the performance of accelerators increases, some of these tasks will go to the facility.

Water safety will be standardized. Industry work on the protocol for evaluating computer vision systems for swimming pools is underway; the appearance of such a document will change the conversation with suppliers — there will be something to refer to instead of vendor figures.

Responsibility for personal data will continue to grow. The changes of 2024-2025 — administrative structures for leaks with negotiable fines and a criminal article for data trafficking — are already being applied. This leads to a practical strategy: collect the minimum, depersonalize early, restrict access. This is cheaper than any measures taken after the incident.

One-page technical task checklist

1. Tasks and solutions. What management decisions are made according to the system. Not "analytics", but a list of solutions.
2. Metrics. What we measure, in what units, with what frequency, with what we compare, which source is the main one for each metric.
3. Zones. Map item layout with zone boundaries and a unified directory of names.
4. Required level by zone. Where there is enough presence, where the zone is needed, where the coordinate is, where the real time is.
5. Accuracy. CEP50 and CEP95 by zone, permissible fraction of zone error, 95th percentile delay.
6. Capacity. Maximum number of placemarks and users, behavior during rush hour.
7. Environmental conditions. Temperature, humidity, condensate, chlorine, splashes, IP version, service life under these conditions.
8. Data. Uploading raw events via API, single session ID, time synchronization, retention periods, depersonalization, execution of requests from subjects.
9. Integration. Accounting system, cash register, locks, access system, CRM. Idempotency of exchange.
10. Legal outline. Grounds for processing, notification, informing, explicit prohibition of face recognition (if not planned), areas without surveillance.
11. The radio frequency part. Bands, capacities, EAEU compliance documents, the need for frequency assignment.
12. Operation. Monitoring the health of sensors, notification of failures, service, training, reaction regulations.
13. Acceptance. The protocol of measurement, criteria, the order of disagreements, repeated measurement in 3-6 months.
14. Stages. What is in the first turn, what is in the second, what requires a pilot.

The whole series

1. Where the guest goes and why to know it: a map of tasks and a map of solutions
2. How to read the characteristics: accuracy, delay, identity, coverage
3. Bracelet as a key: RFID from locker to Flow analytics
4. Bluetooth: from Beacons to Channel Sounding
5. UWB: centimeters and the price of centimeters
6. Wi-Fi: What's left of tracking after MAC randomization
7. Cameras and AI: how people, zones and routes are actually considered
8. Critical events: drowning, falling, overheating, congestion
9. Without a camera and without a tag: radars, time-of-flight sensors, LiDAR and thermal matrices
10. Hybrids, a guest smartphone and large open areas
11. From trajectory to solution: data, metrics, integrations
12. Law: what can and cannot be tracked in Russia
13. Comparison and selection: what to take for your task and what it will cost

Sources

Sources

• RTLS Cost Summary — price orders and cost structure (external estimate)
• Pozyx — anchor density (manufacturer's data)
• Comparison of counting and employment methods, 2026
• Overview of human counting technologies — accuracy ranges
• ST64UWB with 802.15.4ab support, narrowband IMEC receiver
• Bluetooth SIG about Channel Sounding
• IEEE 802.11bf status
• GKRCH solution No. 07-20-03-001 with applications — strips for short-range devices
• Reviews of amendments to Article 13.11 of the Administrative Code and practices under Article 272.1 of the Criminal Code (secondary sources, check with current editions)