People counters explained: the 2026 guide to foot traffic counting

The same device goes by many names depending on country and industry: footfall counter (United Kingdom, Europe), foot traffic counter or store traffic counter (North America), door counter, visitor counter, customer counter, patron counter (libraries), attendance counter (worship, events), among others. The metric they all produce is traffic: the number of people who entered a defined space in a defined period.

Two things a people counter is not. It is not a security camera: it does not record who came in, and a purpose-built counter stores no images at all (more on that under privacy). And it is not a mobile-location dataset: services that estimate visits from smartphone location data describe a market; a counter measures your door. A security camera system with its NVR/DVR may also be able to count, but is typically not specialized in this, and a camera that aims to count is typically not placed correctly to also provide a clear view for security, leading to more cameras and less accuracy on both systems.

People counters started in retail — the first United Kingdom beam counters were installed at Meadowhall in 1988, covering 14 entrances (Wikipedia) — and retail is still the largest user, providing an easy and time-proven ROI. But the same sensors now count library patrons for annual reporting, museum visitors for grant applications, congregations, gym members, office occupancy and passengers.

Every automated people counter does the same three things: it detects a person crossing a count line, it decides whether that crossing is a valid count (and in which direction), and it records the count with a timestamp for software to aggregate. What differs between technologies is how the detection is done — and therefore what confuses it.

The five sensing technologies in use in 2026 are summarized below. Each has a full section, with what breaks it and where it belongs, in the technology comparison.

Because, if you serve people who do not carry a ticket or an access tag, you are optimizing against a number you have never measured. For retailers, sales figures tell you what happened, and only traffic tells you why. A store that sold less this week either had fewer people walk in, or converted fewer of the people who did, or sold them less each. Those are three different problems with three different fixes — marketing, staffing and service, merchandising — and a sales receipt cannot tell them apart. This is the Retail Equation:

Traffic × Conversion rate × Average sale = Total sales

Two of the three terms come from the point of sale. The third only exists if something counts the people who did not buy. That is the whole case for a people counter: it is the only instrument that measures the opportunities you did not convert.

The stakes have not moved online. E-commerce was 17.1% of U.S. retail sales in the second quarter of 2026 (U.S. Census Bureau), so roughly five dollars in six are still spent in a physical store. And the store visit itself is changing: ICSC and McKinsey’s April 2026 report, based on a survey of 3,004 U.S. consumers, finds shopping trips becoming less frequent and more intentional as people research before they leave home (Shopping in the Age of AI). Fewer, more deliberate visits make each one worth more — and make knowing how many you had, and how many you converted, matter more rather than less.

Your first conversion rate is a starting point, not a verdict. Almost everyone is disappointed by it. Most mall-based specialty stores sell to fewer than half the people who walk in; many sell to fewer than one in ten (our own experience since 1972). The number does not matter; the trend does. A store that moves from 10% to 11% conversion has grown sales 10% with the same rent, the same traffic and the same marketing spend. The good news: the lower your rate starts, the more room you have to improve.

What is a good retail conversion rate?

There is no single answer, and be suspicious of anyone who gives one without a vertical. In our experience, store operators who start counting usually discover a conversion rate somewhere between 12% and 20% — far below the 50% or more most of them believed (our archive). Most mall-based specialty stores sell to fewer than half their visitors and many to fewer than one in ten. Grocery converts most visits because people arrive intending to buy, so a traffic count has less value there; luxury converts least because people arrive intending to look. The independent ranges below are the ones Shopify publishes in its 2026 retail guide; no sample sizes are given, so treat them as directional.

Beyond retail: where a visitor count is the product

For a growing number of institutions the count itself is the deliverable. Public libraries in the United States report “library visits” every year through the IMLS Public Libraries Survey, which defines a visit as the total number of persons entering the library for whatever purpose — and IMLS has acknowledged that reported counts can be inaccurate because of limited counting technology and unrepresentative “typical week” sampling multiplied by 52 (our guide to the IMLS requirements). Museums use attendance to support grant applications. Places of worship track attendance trends; offices measure desk and meeting-room utilization; gyms, transit and event venues manage live occupancy against a capacity limit. The sensors are the same; the software emphasis shifts from conversion to reporting and occupancy.

Modern 3D and time-of-flight counters, installed properly, count between 95% and 99% of the people who cross the line; beam counters count 95–98% of single-file adults and less when people walk abreast. Those numbers are real, but they are not what most buyers think they are. They are the vendor’s count measured against the vendor’s own definition of what should have been counted, arrived at by having a team audit sample footage. Nobody reviews every video. Doing that would mean streaming every door to a server and paying people to watch it, which is the exact job the sensor exists to do.

1. A ceiling is not a floor. “Up to 99%” is the best result a vendor has seen over some period. “Never below 95%, verified by video, or we fix it” is the worst day you are willing to accept. Only a minority of vendors in our nine-system comparison commit to a floor at all; we back the 3D Scope II LC with a 95% floor for two years. Ask for the minimum a vendor will stand behind, and how it is verified.

2. The rules are part of the number. Most counters are tuned to count adults, which in practice means anything above a height threshold with a roughly human shape. That is already a proxy, and in retail it is a proxy for something else again: the point was never to exclude children, it was to exclude people who cannot buy. It holds until a ten-year-old is as tall as an adult. Group counting has the same shape of problem — the sensor assumes that people arriving together, spaced similarly and moving on the same path, are one party, and that breaks down quickly. This is why we publish an accuracy figure for adults and decline to publish one for children or groups: a child sits between “object to ignore” and “adult to count precisely”, and two people watching the same doorway would not agree on where a group starts. Ask any vendor what their rules are — children, staff, the customer who steps in and straight back out, the person who stands in the doorway for two minutes — and agree them before you audit anything.

3. Consistency is accuracy that holds. An accuracy rating is measured under particular conditions; you still have to factor in consistency. A sensor rated at 98% that is thrown off whenever sun floods the entrance reads 98% most of the time and far less for those hours, and once one period is measured differently from another, the comparison — the reason you count — is gone. The same goes for time not counted at all: an outage on a sensor that does not store its counts, one failing link in a chain of devices, a battery nobody replaced. Our sensors store counts through an outage, and PEARL warns you before its batteries run out. A small, steady error, by contrast, is harmless: a counter that misses the same small share of people every day still tells you the truth about your trends. How our two sensors handle each of these is set out in our accuracy guide.

4. What you are really buying is verification. A beam is simple enough to check yourself: walk the line and watch the count. A 3D sensor should let you see what it sees, live, and record a clip with the count lines drawn on it, which is why we publish the audit procedure rather than asking you to take a number on faith. Time-of-flight sits in the middle: it does produce a viewable picture — a depth silhouette with head and shoulders clearly made out, and no face, color or clothing — so you can watch it work, but you cannot tell a person from a well-shaped object. Thermal is the hardest: a low-resolution heat map where a shopping cart left in the sun looks like a warm body. Both earn their place — they count in the dark and hold the strongest privacy position of any technology — but verifying them is harder. The industry knew this: Irisys, once the largest thermal people-counting vendor in the world and now out of the market, ended up fitting optical cameras alongside its thermal sensors for no reason other than letting customers confirm the counts were people.

None of that makes the figures worthless. A television’s contrast ratio is measured on a full-black screen at the darkest setting and a full-white screen at the most vivid one, which is a real measurement of a situation nobody will ever watch. It still tells you the manufacturer was willing to be measured. An accuracy figure works the same way. What it actually says is “this will work, it will be consistent, and it will hold up if you audit it” — and the vendors worth buying from will let you do exactly that.

For one entrance, expect $300–$500 for a wireless beam sensor, $300–$1,200 for a 3D stereo or time-of-flight sensor (most fall between $650 and $1,200), up to about $2,000 for outdoor and specialized units, and $0–$150 per location per month for software. Three of the largest vendors publish no prices at all. Our nine-system comparison found hardware from $299 to $1,149.95 where it is sold outright, with quote-only enterprise units above that; consumer-grade beam counters with a display sell online for under $100, with the limitations described under tally counters.

Think in five-year totals, not sticker prices. A $300 sensor on a $150-a-month plan costs about $8,400 over five years for one door; a $500 sensor on a free plan costs about $525 including batteries. Neither is wrong — the expensive one may include cellular connectivity, hardware replacement and analytics you need — but they are not the same purchase, and the cheaper day-one number is the dearer one to own. Every published figure for nine systems, over three and five years, is in the comparison.

Return on investment is a conversion question. Take your weekly traffic, your conversion rate and your average sale. Ask what one extra sale per hundred visitors is worth per year. For almost any store with more than a few hundred visitors a week, that single point of conversion pays for the counter, the software and the cable within months — provided someone acts on the data. A counter nobody looks at has an ROI of zero.

Yes, people counters are legal everywhere we sell — and a counter built to count rather than to watch sits outside most privacy law altogether, because it never collects personal data. The legal question is not “is counting allowed” but “what does this particular sensor capture and keep”. Three categories:

One comparison is worth making before those: almost every building that adds a counter is already running security cameras that record identifiable video and keep it for weeks. A sensor that reduces what it sees to a height map and stores a number is the less invasive of the two by a wide margin — and it is usually the one that gets questioned.

United States

There is no federal privacy statute covering counting. State biometric laws — Illinois BIPA (the only one with a private right of action, and the source of most litigation), Texas CUBI, Washington RCW 19.375, and Colorado’s biometric amendments in force since July 1, 2025 — regulate biometric identifiers: face geometry, fingerprints, voiceprints (2025 litigation review). A counter that never captures a face has nothing to register. Face-based age and gender analytics, by contrast, is exactly what these laws were written for; if you want demographics in the U.S., get legal advice first.

Where it has gone wrong in US stores, it was always about identifying or profiling shoppers. The FTC banned Rite Aid from facial recognition for five years in 2023 after false matches flagged innocent customers; Macy’s was sued under BIPA over facial recognition used on Chicago shoppers; Amazon Go was sued under New York City’s law that requires a sign at the entrance of any store collecting biometric data; cameras on Walgreens and Kroger shelves and cooler doors that could guess shoppers’ age and gender made national news in 2019; and in 2015 the FTC settled with a retail Wi-Fi tracking firm that had promised shoppers an opt-out it did not provide. Portland, Oregon goes further and bans facial recognition in stores and other places open to the public. None of these involved a counter that stores no image and identifies no one.

Canada and Quebec

PIPEDA federally and Quebec’s Law 25 provincially govern personal information, with Law 25 carrying penalties of up to $10 million or 2% of worldwide revenue. Quebec’s regulator, the Commission d’accès à l’information, takes the strictest line in North America on biometrics — it refused a grocery chain’s proposed face-recognition database for shoplifting in 2025 (Torys) — and its 2024 anonymization regulation treats aggregated statistics as the reliable form of anonymous data (Fasken). Hourly totals from a sensor that stored no image are precisely that. We are a Quebec company, and this is the environment our sensors were designed in.

European Union and United Kingdom

Under GDPR, data is personal if it relates to an identifiable person. Anonymous counts do not: Recital 26 states that the principles of data protection “should not apply to anonymous information” and that the Regulation “does not concern the processing of such anonymous information, including for statistical purposes” (GDPR Recital 26). So the heaviest obligations (lawful basis, DPIA, subject-access rights) do not attach to a beam, thermal or depth-only counter. The EU AI Act’s prohibited practices have applied since February 2, 2025: they ban biometric categorization by sensitive attributes, untargeted facial scraping, emotion recognition at work and in schools, and real-time remote biometric identification for law enforcement. None describes anonymous counting, which involves no biometric data. Camera systems that do infer demographics from faces should be assessed against the Act’s biometric provisions as the high-risk rules phase in (timeline).

The right technology is decided by the doorway, not the brand. Width, traffic density, door type, ceiling height, lighting and whether power can reach the entrance settle it. This is the same logic we apply on our own demo calls, with our products named where they fit and other technologies named where they fit better.

An update of the checklist we first published in 2020. The questions have not changed much; the answers vendors give have.

We offer one sensor of each kind that this guide recommends for ordinary doors: a wireless beam for standard entrances and a 3D stereo camera for everything harder. Both ship free to the U.S. and Canada, come with a two-year warranty and a 30-day money-back guarantee, and run on software from $0 a month.

Claims about our own products come from our product pages, pricing page and help center, all linked in place. Claims about the market and the law come from the independent sources below, checked on September 9, 2026. No competitor is quoted or linked on this page; named comparisons live on the comparison page.

How this guide was made

SMS Storetraffic publishes and maintains this page. It consolidates and updates six articles we published between 2016 and 2024 — the technology series, the manual-versus-automated piece and two introductions — and adds what changed: edge AI, time-of-flight, the collapse of Wi-Fi counting, the EU AI Act, and 2026 prices. The author has worked in automated people counting since 2006. We sell two of the products mentioned and say so wherever they appear; where another technology or vendor is the better fit, the guide says that too. Corrections are welcome at our contact page and are applied at the next quarterly review.