How many people came in, how many bought, where the queues are and which shelves people walk up to — from the cameras that already hang over the entrance and the tills. NeuroVision counts visitors, it does not recognise them: the report holds numbers and time, with no faces and no names. You do not need to buy separate door counters.
Sales floor analytics
Gender and age group in the report — anonymised only, as agreed during the pilot
Video analytics for retail starts with a simple question: how many people came into the store today. An ordinary door counter answers only that. A neural network connected to the store’s cameras sees more: where people stop, how many are at the tills and how many leave without buying.
We connect to the store’s IP cameras and archive. You do not need to buy and install sensors on every door to get started.
The number of visitors says little on its own. Next to the number of receipts from the till system it becomes conversion — and you can see at which hours the store is full but there are no sales.
A manager does not need to review the archive to understand why there were complaints about the queue in the evening. The system itself notes when more people stand at the tills than the set limit and sends an event to the floor administrator.
Each metric is calculated from anonymised silhouettes of people in the frame. You can connect one — for example, only entrance traffic — and add the rest as the chain gets used to the reports.
A camera above the entrance counts those who came in and out along a line drawn on the frame. Staff in uniform and children carried by parents are separate rules, set during the pilot. The report is by hour, day of the week and store of the chain.
The number of visitors per hour is matched with the number of receipts from your till system. A drop in conversion with a full floor is a signal that you lack sales staff or tills, not customers.
The system counts people in the till area and sends an event to the administrator when the queue is longer than set or waits longer than the set time. The report shows at which hours there are not enough open tills.
Where customers spend more time, which racks they walk up to and which zones stay empty. Paths through the floor are added up into a heat map for the day, not into the route of a specific person.
How many people are on the floor every 15 minutes. Occupancy is used to build the shift schedule for sales staff and cashiers, plan goods receiving for quiet hours and check how the store handles peak days.
An example is the neighbourhood store “Makovka” with one entrance, three tills and cameras that were installed for security.
The camera above the entrance and the camera above the tills are added to the console as sources. For the check, a recording from last week is taken: it shows how the system counts even before the live stream is connected.
A crossing line is drawn on the entrance frame and a queue zone on the till frame. For the queue a rule is set: “more than five people for longer than three minutes — an event to the administrator”.
The system finds people in the frame and counts line crossings and time in the zone. Faces are not compared with any database: a silhouette is enough for counting.
The store manager sees traffic and conversion for the day, and the floor administrator gets a queue event on the phone. Data from the chain’s stores is combined into one report.
A visitor report is useful only if something is changed because of it. Here are six decisions for which chains connect counting.
Sales staff and cashiers are scheduled for the hours when people are on the floor, not “as always”. Busy Friday evenings and quiet Monday mornings are visible in the report right away.
Queue events and the waiting report show where a third till is needed every day and where it is opened for nothing.
Traffic and time at the promo stand are compared before the campaign and during it. If people walk up but do not buy, the issue is the price or the display, not the advertising.
The heat map shows the “dead” zones of the floor and the shelves that people pass without stopping. The display is changed and a week later the map is checked again.
The same traffic and different conversion at two stores — a reason to look into staff, range or tills. The report combines all stores in one table.
Hourly flow data helps to discuss rent terms with the shopping centre and to decide whether to open earlier or close later.
The report holds numbers and time
How many came in, how many stood in the queue and for how long, how much time was spent at a shelf. A frame is saved only for an event — for example, a queue — and kept for as long as you set.
No face templates
Counting does not need a face: the system works with a silhouette. Face templates are not created and not stored, either in the cloud or with you.
No regular customers
The system does not recognise that this person was here yesterday and does not link visits to each other or across stores. That is why there are no “unique visitors” in the report — only entries.
No link to a loyalty card
The visitor report is not matched with loyalty cards, the receipts of specific customers or phone numbers. Conversion is calculated from the number of receipts, not from people.
Visitor counting is not biometrics as long as the system does not try to work out who is in front of the camera. That is why the report has limits that cannot be configured.
Gender and age group of visitors — anonymised, as shares in the report, without saving faces; how this works is on the page on age, gender and emotion detection by face.
One thing matters to a chain’s lawyer: at which step video analytics turns into biometric processing. Visitor counting does not reach that step. Below is a summary of the rules as of October 2026: a guide for a conversation with a lawyer, not legal advice.
The set of metrics depends on the format: for a neighbourhood store the tills matter, in a hypermarket the floor, in a shopping centre the entrances.
NeuroVision provides analytics and an API. Cameras, the video recorder and installation are yours or your integrator’s. We check the requirements on a recording before the pilot.
Best from above or at an angle over the door, so that people do not block each other. An already installed security camera often fits: you can see this from one day of recording.
For a heat map and time at the shelves, the overview cameras already hanging under the ceiling are enough. A face is not needed for this — and it is good that it is not visible.
It sees the whole line of tills and the area where people wait. One camera can cover three or four tills.
If there are too few cameras or the angle is poor, your integrator or a NeuroVision partner will add one or two cameras. We provide the API, the models and rule setup.
Pilot — 1 month free
Cloud in Russia or your own server
Your cameras and recordings
The price depends on the number of stores and cameras, the set of metrics and where processing runs. You can start with a free pilot in one store.
One store, entrance and till cameras or a recording of past days, traffic, queues and a report together with the manager.
on request
Processing on servers in Russia, a console with chain-wide reports, an API for BI and the till system.
on request
Installation on the chain’s own servers with no video leaving the perimeter, a licence for the number of stores and cameras.
Cameras, the video recorder and installation are not included in the price. How to choose between the cloud and the perimeter is explained on the video analytics page; the price matrix does not list video analytics — we calculate the price for your chain.
Words you will meet in the report, in a conversation with an integrator and at a meeting with managers.
The project «Development of a multimodal system using artificial intelligence "Neuromed" for managing the diagnostic processes of a clinic on the basis of the "Neurovision" computer vision platform» is supported by the Innovation Promotion Foundation within the federal project «Domestic Solutions»
The Neurovision computer vision platform is registered with Rospatent (certificate No. 2023661508) and included in the Russian Software Registry (entry No. 19972 of 13.11.2023)
The right to use the Platform is granted under a licence agreement concluded by Neurovision with its customers
Tell us how many stores and cameras you have, whether there is a recording of past days and which till system you use — we will suggest the metrics and a pilot in one store.
Can't put your request into words? Book a call and our team will contact you at a convenient time
Book a callThe company is included in the register of IT companies and the register of personal data operators of the Russian Ministry of Digital Development.
Address: Moscow, Presnenskaya Embankment 12 — Moscow City business centre, Federation Tower. Nearest metro stations: Delovoy Tsentr and Mezhdunarodnaya.
Materials on face recognition, algorithm accuracy and protection against deepfakes — in our blog.
A neural network watches the video from the camera above the entrance and counts the people who cross the line drawn on the frame. People in the till zone or at a shelf are counted the same way. The report holds numbers and time; faces are not compared with anything.
Most often no. The analytics connects to the IP cameras already hanging above the entrance and the tills. Whether the angle and quality are suitable can be seen from one day of recording — we review it before the pilot.
No. Biometrics begins when a person is identified by face (Art. 11(1) of 152‑FZ). Counting works with silhouettes, face templates are not created or stored, and the system does not recognise repeat visitors.
No. Organisations are prohibited from collecting and storing biometrics in their own systems for identification outside the Unified Biometric System (Art. 15(1)(1) of 572‑FZ). That is why the report has no “unique visitors” — only entries.
The number of visitors per hour is matched with the number of receipts from your till system. Receipts arrive via the API; your team or integrator sets up the link.
Only anonymously — as shares in the report, without saving faces. Read more on the page on age, gender and emotion detection by face. For selling alcohol and tobacco, age is still confirmed with a document — see the page on age verification.
Employees in uniform can be excluded from counting by a rule. We do not keep working time records or assess a specific salesperson’s work from cameras. The purposes of video surveillance are fixed in an internal act and employees are made familiar with it against signature.
In the cloud on servers in Russia, or on your servers if the analytics is installed inside the chain’s perimeter. Frames are saved only for events — for example, a queue — and kept for as long as you set.
The first month of a pilot in one store is free. After that we calculate the price by the number of stores, cameras and metrics, separately for the cloud and for installation in your perimeter. Cameras and installation are separate.
