A people counter answers a single question: how many people passed through a door. It can be done with an infrared beam, a 3D sensor mounted above the entrance, phone signals, a person with a clicker, or software that reads the security cameras you already have. Below: what each one measures, what it can’t do, what it actually costs — and, honestly, where the ceiling sensor is better than us.
This page compares types of devices, not brands. Every type has good manufacturers and weak ones; what follows is about the measuring principle, not the company.
They all give you a number. Not all of them give the same number, and none of them gives the number you think it does until you know how it gets it.
The same device also goes by people counter, visitor counter, footfall counter, people counting camera, people flow counter, door counter or customer counting system; the one that tells entries from exits is sold as “bidirectional”. In Romanian: contor de persoane, contor de vizitatori. In Russian: счетчик посетителей, счётчик проходимости, датчик подсчета людей, система подсчета посетителей. They are the same five principles described below, under other trade names.
A beam runs across the door; each break counts as one crossing. Two people walking in shoulder to shoulder are a single crossing. A single beam doesn’t know direction; two beams infer it from the order of the breaks. It counts crossings, not people: the courier, the employee stepping out to take a call and the customer going back for a wallet add up to four counts.
Mounted directly above the entrance, it measures the distance to the top of each head and separates people walking through together. It knows direction, so it counts exits too. For counting a door, it’s the best thing there is — it’s built for one job and does it well. It’s priced per door, needs cable and power, and sees exactly that: the door.
It listens for the signals phones send out when they search for networks. Ever since phones started changing their address on every search — and iPhone and Android do this by default — the same phone shows up as dozens of different devices. A person with Wi‑Fi turned off doesn’t exist, while a car stopped at the traffic light in front of the store window gets counted. Good for broad trends, unreliable for the daily number.
Traffic studies are still done this way today, and rightly so: for a single hour, it is the most accurate method you have at hand. The problem isn’t accuracy, it’s coverage. Nobody keeps a person with a clicker on duty twelve hours a day, seven days a week, at four locations. We use it as the benchmark: we count a few hours by hand and check the system against them.
It reads the feed from the cameras you already have. It doesn’t just look at the door: the same image shows the checkout queue, the aisle, the shelf and the employee. Nothing needs to be installed if a camera sees the entrance. Its limit is an honest one: it depends on what the camera sees. Where coverage falls short, we add one or two ordinary cameras.
In the middle column, any door counter — whatever its type and brand, since they all sit at the door and see a single line. On the right, the answers from software that reads the cameras across the whole sales floor.
If you take our word that we’re better at everything, you’ll be disappointed at the first check. So we’ll tell you up front where we’re not.
For counting a single door, a 3D sensor properly mounted above the entrance is more accurate than any software reading a camera that was put there for something else, at a different angle and a different height. If you have one location, one entrance, and all you’re missing is the visitor count, the sensor is the right answer, and we’ll be the ones to tell you so.
Conversion against receipts, the checkout queue, dead zones on the sales floor, the hours when you lose people, employees taken out of the count, and every location measured the same way, in a single dashboard. A door counter can’t give you these, however accurate it is at the door — not because it’s weak, but because it’s looking elsewhere.
We don’t publish list prices, because we don’t sell standard packages. But the cost structure can be laid out honestly, without figures, and it’s worth knowing before you ask for quotes.
Device, cable, power, labor. If you have three entrances, that’s three devices. If you have ten stores with two entrances each, that’s twenty. On top of that comes software to combine the numbers, if you want to see them all in one place. Any refit means redoing the installation.
The cameras are already there and already paid for. Where coverage falls short, one or two ordinary cameras are added. No walls to cut into, no cable to run to every door, and a refit costs a few minutes of configuration, not a technician visit.
Vizor is video analytics for customer-facing locations: stores, pharmacies, fuel stations, showrooms, warehouses. It runs on existing security cameras and passes on numbers, not images.
Customer counting on the camera that sees the entrance, with time and direction. The same method at every location, so the numbers can be compared with one another.
Matched against checkout receipts, hour by hour. That gives you the real conversion and the number of people who left without buying.
The queue at the checkout or the counter: how many were waiting and for how long. When to open the next checkout stops being a hunch.
A heat map on the floor plan and attention by zone, so you can see whether the promotion got noticed or was just set out.
Conversion by hour, bypassed zones and checkout queues, in every store in the chain.
Read →What video analytics measures, which cameras it works on, what it costs and how accurate it is.
Read →A new AI camera or AI on your existing CCTV: what each one does and when it pays off.
Read →How it’s calculated, the denominator traps, and why you read it by the hour, not by the month.
Read →It depends on the type. An infrared beam counter counts how many times a beam stretched across the door is broken. A ceiling-mounted 3D sensor measures the distance to the tops of people’s heads and separates them even when they walk through together. Wi‑Fi counting listens to the signals from phones. Video analytics reads the camera image and tracks people in the frame. In the end they all produce the same kind of number — how many crossings over a line — but they differ in what they can tell apart: direction, groups, employees, children in strollers.
Five, in practice: the infrared beam counter with one or two beams, the stereo 3D ceiling sensor, counting by phone signals (Wi‑Fi or Bluetooth), manual counting with a clicker, and video analytics on existing security cameras. The first two are mounted at the door and see only the door. The last one sees everything the camera sees: the entrance, the queue, the aisle, the shelf.
One that counts entries and exits separately, not just crossings. An infrared beam needs two beams for this and works out the direction from the order in which they are broken; a 3D ceiling sensor and video analytics see the direction directly. The difference matters when you want to know how many people are inside at a given moment — for room capacity or for how many tills to open — because that comes from subtracting exits from entries. For conversion, entries are enough.
An infrared beam is fixed to the door frame at chest height: the emitter on one side, the reflector or receiver on the other. A 3D sensor is mounted on the ceiling, directly above the entrance, at the height the manufacturer specifies, with cabling and power. Wide or double entrances need several devices. With video analytics nothing is mounted if an existing camera sees the entrance; otherwise an ordinary camera is added above the door.
A door counter is paid for once, per door, plus installation: device, cable, power, labor. Three entrances mean three devices, and every refit means redoing the installation. Video analytics on existing cameras is paid per location, monthly, and usually without hardware, because it uses the cameras already installed. We don’t sell standard packages: we first look at a location, write down what can be measured there, and then send a quote.
For one person walking in alone, yes. For two people walking in shoulder to shoulder, no: the beam is broken once, so one person is counted. A single beam doesn’t know direction either — coming in and going out look the same. With two beams, direction is inferred from the order of the breaks. It is the cheapest solution and the most sensitive to crowding, which means it gets it wrong exactly in the hours you care about most.
A device with two lenses, mounted directly above the entrance, that calculates the distance to every point from the difference between the two images. It sees people from above as little mounds and separates them even when they walk through pressed together. It knows direction, so it counts entries and exits separately. For counting a door, it is the most accurate thing on the market. Its limit isn’t accuracy but field of view: it sees the door, and only the door.
It’s possible, but the number is no longer reliable for any given day. Phones change their own address on every network search — iPhone and Android do this by default — so the same phone shows up as dozens of different devices. A person with Wi‑Fi turned off isn’t counted at all, while a passerby on the sidewalk or a driver stopped at the light in front of the store window is. It remains useful for broad trends, not for Tuesday’s conversion.
Yes, and without that the number isn’t worth much. The visitor count only becomes useful when it sits next to the number of receipts from the same hour: the ratio between them is conversion. We take hourly receipts from whatever system you have — POS, management software or 1C — and put them next to the entrance count, automatically. If they can’t be read directly, a daily file can be uploaded.
Yes, if it sees the entrance and the image is clear enough. That’s exactly what Vizor does: it runs on the security cameras you already have, whatever the brand, and counts the people in the image. No device is mounted at the door. Where coverage falls short — an entrance no camera can see — one or two ordinary cameras are added.
A door counter counts them, because it has no way to tell them apart: to it, any body that crosses the line is a number. A salesperson who steps out ten times a day is added to the statistics ten times, and in small locations this visibly skews conversion. Vizor recognizes the employees you register in the system and takes them out of the customer count. Customers are never identified.
It depends on what the camera sees, which is why we don’t give a single figure on the site. We check it at your location: we count a few hours from different days by hand and put the manual count next to the system’s number, so that both sides see the real margin of error, not a promised one. This check is part of the pilot and is repeated until the numbers settle.
It can be, but not in every form. Camera images are personal data, because people in the frame can be recognised by their appearance, so the GDPR applies to processing them; the aggregate figures that come out, from which no one can be identified, are no longer covered by the Regulation. The European Data Protection Board’s Guidelines 3/2019 distinguish simple counting algorithms, which are less intrusive, from biometric techniques. Facial recognition used by a private company for marketing or statistics will in most cases require the explicit consent of every person (Article 9), and systematic large-scale monitoring of a publicly accessible area requires a data protection impact assessment (Article 35). In the Republic of Moldova the same rules apply through Law No. 195/2024.
We look at a snapshot from your entrance camera and write back within 24 hours to say whether it can be used for counting or whether you need something else. If the answer is “you need a sensor at the door,” we’ll be the ones to tell you.