People Counting Sensors for Retail, Venues and Smart Buildings: The UK Buyer’s Guide

Last updated: 13 July 2026

TL;DR: People counting sensors measure how many people enter, leave or occupy a space, and they now reach up to 99.8% accuracy. UK retailers use them to calculate conversion rates, venues use them to hold fire-safety occupancy limits, and facilities teams use them for queue management and cleaning on demand. Non-camera ToF people counting sensors avoid capturing images, which keeps you clear of most CCTV-style GDPR obligations.

Footfall is the one retail metric you cannot pull from your till. Your EPOS knows how many people bought; only people counting sensors know how many walked in, queued, gave up or wandered straight back out. This guide compares the four main technologies with real accuracy figures, explains the UK GDPR position using ICO guidance, and shows where people counting sensors pay for themselves in retail, venues and smart buildings.

What are people counting sensors and how do they work?

People counting sensors are overhead or doorway-mounted devices that detect people crossing a line or occupying a zone, then report totals to a dashboard. Depending on the model they use infrared beams, thermal imaging, AI cameras or Time-of-Flight (ToF) depth sensing. The best units process everything on the device and transmit only anonymous counts.

A counter mounted above an entrance tracks two numbers: people in and people out. Subtract one from the other and you have live occupancy. Place additional counters on internal thresholds, and you can see how visitors flow between floors, departments or zones.

The transport layer matters as much as the optics. Mains-powered AI units usually report over Ethernet or Wi-Fi, while battery-powered counters typically use LoRaWAN, a long-range low-power radio that runs for years without cabling. If that technology is new to you, start with our guide to what LoRaWAN is and how it works.

Why does footfall counting matter for UK retail?

Footfall counting turns gut feel into a denominator. Divide transactions by visitors and you get conversion rate, the single most useful number in physical retail. Without a counter on the door you cannot tell whether a quiet till means fewer visitors or worse conversion, and those two problems have completely different fixes.

The national picture shifts week by week. MRI Software data reported by Retail Times showed UK high street footfall up 4.3% year on year in late April 2026, while shopping centres slipped 1.3% and retail parks 0.7%. Coastal towns jumped 6.5% in a single week. National series exist because someone counts: the ONS publishes weekly UK retail footfall data supplied by BT Active Intelligence.

Your own store will not match the national average, and that is the point. Site-level counting lets you separate market movement from store performance: if the high street is up 4.3% and your door count is flat, the problem is your frontage, not the economy.

Footfall data also earns its keep in staffing. Match rota hours to your measured entrance peaks rather than to trading folklore, and you stop paying for empty-shop labour on Tuesday mornings while queues form unattended on Saturday afternoons.

Which people counting technology is the most accurate?

AI-based counters lead on accuracy. Retail Sensing’s comparison of nine methods puts video analytics at 97 to 99% and 3D stereo vision at 95 to 99%, with infrared beams at 85 to 95%. Milesight quotes up to 99.8% for its AI ToF people counting sensors. WiFi tracking trails badly and carries GDPR baggage.

Here is how the main options compare, using figures from Retail Sensing’s nine-method comparison and Milesight’s people counting solution page:

Each technology fails differently. Single infrared beams cannot separate two people walking abreast, which is why their accuracy collapses on wide or busy entrances. Thermal accuracy degrades when ambient temperatures climb towards body heat: Retail Sensing notes performance drops significantly above 28 to 30°C. WiFi counting broke when phone makers randomised MAC addresses, and V-Count’s technology guide now describes it as unreliable as well as privacy-problematic.

ToF is the newer benchmark. The sensor emits infrared light and times its reflection, building a depth silhouette rather than a photograph. That is how modern people counting sensors hit 99.8% while remaining anonymous by design, and why ToF units work in complete darkness.

Are people counting sensors GDPR compliant?

They can be, and technology choice decides how much paperwork you carry. Camera-based counters process images of identifiable people, so ICO video surveillance rules apply, including a Data Protection Impact Assessment in most cases. Non-camera ToF and thermal people counting sensors that process data on the device and output only totals never handle personal data in normal operation.

The ICO is blunt about camera systems. Its video surveillance guidance states that you should perform a DPIA before any processing, and that this is a legal requirement in most cases relating to video surveillance, including systematic monitoring of publicly accessible places on a large scale. Failing to carry out a required DPIA is itself an infringement of the UK GDPR that can attract enforcement action.

The trigger sits in the law, not just the guidance. UK GDPR Article 35(3) makes a DPIA mandatory for systematic monitoring of a publicly accessible area on a large scale, and the ICO lists systematic monitoring among its high-risk indicators. A shopping centre running AI cameras over every entrance sits squarely inside that description.

Non-camera people counting sensors change the analysis at the source. A ToF unit never captures a face, a gait or a number plate; it sees anonymous depth blobs, counts them on the device, and transmits an integer. Milesight markets its range as 100% anonymous detection with no personally identifiable information captured. If no individual can be identified or singled out, you are not processing personal data when counting.

You should still document that reasoning. A short screening note explaining why your counters fall outside DPIA territory is cheap insurance, and it is exactly the kind of accountability evidence the ICO expects. If you later add cameras for security, assess that system separately.

How do smart buildings use people counting sensors day to day?

Beyond retail conversion, people counting sensors drive four building operations: queue management, occupancy limits for fire safety, cleaning on demand, and space utilisation. Each replaces a schedule or a guess with a measured number, which is why counting is usually the first sensor layer a smart building deploys.

Queue management. A counter watching a queue zone knows how many people are waiting right now. Set a threshold and you can page staff to open a second till before shoppers walk out, then check afterwards how long the queue actually stayed above the line.

Fire-safety occupancy limits. Under the Regulatory Reform (Fire Safety) Order 2005, the responsible person must establish a safe occupant capacity. Kent Fire and Rescue’s occupancy guidance works this out from floor space factors, such as 0.3 m² per person for standing areas and 1.0 m² for restaurants, capped by exit capacity, whichever is lower. Live in-minus-out counting turns that static limit into something a duty manager can actually enforce on the night.

Cleaning on demand. Instead of cleaning washrooms every two hours whether used or not, a threshold-based regime cleans after every N visits. Usage-based scheduling puts effort where the traffic is, and gives you a defensible log of both usage and response.

Space utilisation. Zone-level counting shows which meeting rooms, aisles or galleries actually attract people. Estates teams use the data to consolidate underused floors, and museums and venues use it to price and programme space based on measured attention rather than anecdote.

Which Milesight people counting sensors should you choose?

Pick by doorway and data need. The Milesight VS133 AI ToF people counting sensor is the accuracy flagship at up to 99.8%, counting bidirectionally over standard entrances with no images captured. For very wide or tall openings, the VS135 Ultra ToF people counter extends the same depth-sensing approach.

Where you cannot run power, the battery-operated VS350 passage people counter counts in and out over LoRaWAN, which makes it the quick win for corridors, portable venues and heritage buildings where cabling is unwelcome.

For whole-room intelligence, the VS121 AI workplace occupancy sensor monitors multiple zones with 95% accuracy, mapping how a floor is really used, with a companion VS340 desk and seat sensor available for furniture-level detail. And for washrooms, the ToF-based VS330 occupancy sensor reports cubicle status anonymously, the natural trigger for cleaning on demand.

All of these people counting sensors are LoRaWAN or IP devices that slot into the same network as your air quality, energy and leak sensors, so counting becomes one layer of a wider building dataset. See how they combine in our industrial IoT for retail and facilities management guides.

People counting sensors FAQ

How do people counting sensors work?

Most mount above a doorway and detect people passing beneath, using infrared beams, thermal signatures, AI cameras or ToF depth sensing. The device classifies direction, increments an in or out counter, and pushes totals to a dashboard over LoRaWAN, Wi-Fi or Ethernet. Occupancy is calculated as entries minus exits.

How accurate are people counting sensors?

Published figures range from 85 to 95% for single infrared beams, 96 to 98% for thermal units and 97 to 99% for AI video analytics, according to Retail Sensing’s comparison. Milesight quotes up to 99.8% for AI ToF models. Accuracy always depends on mounting height, entrance width and calibration.

Do people counting sensors need a DPIA under UK GDPR?

Camera-based counters usually do: the ICO says a DPIA is a legal requirement in most video surveillance deployments, especially large-scale monitoring of publicly accessible places. Non-camera ToF or beam people counting sensors that only ever output anonymous totals do not process personal data, though you should record a brief screening decision.

What is the difference between people counting sensors and CCTV analytics?

CCTV analytics extracts counts from security camera footage, so you are still processing identifiable images with all the retention, signage and access obligations that follow. Dedicated people counting sensors are purpose-built to count, and non-camera versions never form an image at all, which simplifies compliance and cuts storage costs.

Can people counting sensors run on batteries with LoRaWAN?

Yes. Battery-powered units such as the Milesight VS350 send compact count messages over LoRaWAN, so they run for years without mains power or network cabling. That suits corridors, listed buildings, temporary events and multi-site estates where you want footfall counting everywhere without an electrician at every door.

Ready to measure instead of guess? Browse people counting sensors at indiott.com, from ToF flagships to battery LoRaWAN counters in our online store, or request a quote and we will spec the right counter for your entrances, zones and occupancy limits.