People counting and occupancy monitoring, using the cameras you already have
Count visitors, measure how full a space really is and see how people move through it, without installing a new sensor in every room. Our Sentinel Video Tracker works on existing CCTV feeds. It counts people without identifying them: no names, no profiles, no facial recognition. We are a technology provider with our own products and a consultant on indoor geolocation projects, and since 2011 that has meant helping teams choose the technology that actually fits the question they are asking, even when the honest answer is a cheaper one.
Not sure whether you need a counter, an occupancy number or full flow analytics? That is a normal place to start.
Start from the question you cannot answer today
Most people counting projects do not start with a technology. They start with a decision that keeps getting made on instinct: how many staff to put on the floor, whether to renew a lease, when to clean, where to put the new desks. These are the problems we hear most often.
You are guessing how busy the space is
Managers estimate occupancy from memory and from how the room felt last Tuesday. Two people in the same meeting describe the same space as crowded and as half empty, and both are arguing from impressions rather than from a number.
Your footfall data stops at the entrance
A door counter tells you how many people came in. It does not tell you where they went, which areas they skipped, how long they stayed or whether the same person walked in and out several times. Entries are a start, not an answer.
Staffing and cleaning run on habit, not on demand
Shifts, opening hours and cleaning rounds were set years ago and never tested against real traffic. You may be staffing a quiet hour and understaffing the peak on the same day.
You pay for space that nobody uses
Rent, heating and cleaning are paid per square meter, not per square meter used. Without occupancy data, a desk, a floor or an entire wing can sit empty for months before anyone can prove it.
Every system gives a different number
The turnstile, the Wi-Fi controller and the till each report a different count, so nobody trusts any of them. Before you buy another source of numbers, it is worth deciding which number you will actually act on.
Three questions to answer before you buy anything
These three answers decide the technology, the cost and how hard the privacy conversation will be. They are the first things we ask on a discovery call, and they are worth answering even if you never speak to us.
01. Do you need a number, or a map?
A single count per hour is a different product from an occupancy figure per zone, which is different again from a movement path across a floor. Counting at a door is the cheapest of the three. Flow analysis is the most expensive. Many teams ask for the third and only ever use the first.
02. How wrong can the number be before your decision changes?
If you are deciding whether to renew a lease, a few percent either way changes nothing. If you are enforcing a legal occupancy limit, it changes everything. A hypothetical example: a 300-seat canteen deciding whether to open a second lunch sitting needs peak occupancy per 15 minutes and can live with a count that is a few people out, while the same building enforcing a hard capacity limit in an event space cannot. Set the tolerance first, then buy to it. Precision you never use is money spent on reassurance.
03. Do you need to count people, or to recognize them?
For counting, occupancy and flow, the answer is almost always counting. You need to know that a body is there, not who it belongs to. Recognizing individuals raises the legal bar sharply in the EU and rarely improves the number you were after.
How we count people with your existing cameras
Sentinel Video Tracker analyzes the video from CCTV cameras you already run. It detects people in the frame and follows their position while they are in view, then turns that into counts, occupancy per zone and movement between zones. The output is numbers and zones, not clips of individuals.
It runs on the cameras you have
Where your existing CCTV covers the area well enough, no new hardware goes on the ceiling. Some cameras will not be usable: a camera aimed at a doorway to capture faces is the wrong view for counting a shop floor. We tell you which views work before anything is installed.
98% counting accuracy, on crossings and on zones
Sentinel counts people with 98% counting accuracy, and inside the covered area it resolves a person's position to 20 cm. That is more than enough to assign someone to a zone, a queue or an aisle, and to measure dwell time in a defined area. We still measure the counting accuracy against a manual count in your building during the pilot, because a number from someone else's building is not a number you can plan with. Which conditions that figure holds under in your scene, from camera angle to lighting and crowd density, is one of the things the camera assessment defines.
What comes out of it
Counts per period, occupancy per zone, movement between zones and dwell time. The form those numbers take, and the route they travel to reach your systems, is settled in the camera assessment and the first call. That order is deliberate: the binding constraints are usually your IT policy and the tools your team already reads, not the product.
Where the processing happens
This matters for privacy, for bandwidth and for your IT policy, so it belongs in the first conversation, not in the contract. Bring us the rule your organization has to follow on where video and derived data may be processed. The camera assessment is where that gets settled, with your IT team in the room, before anything is installed.
When video is not the right sensor
If you only need to know whether one meeting room is occupied, a simple occupancy sensor on the ceiling costs less and needs no camera. If you only need entries at one door, a door counter is cheaper. We will say so.
Counting people without identifying them
In the EU this is the first question your DPO, your works council and often your customers will ask, so we put it before the price. A camera pointed at people is not automatically a privacy problem, and it is not automatically fine either. What matters is what the system extracts from the image and what it keeps.
Anonymous by design, not anonymous by promise
For counting, occupancy and flow, the system needs to know that a person is at a position. It does not need to know who they are. No face is matched against a database, no identity is assigned, no profile follows someone between visits. The useful output is a count and a zone.
No facial recognition when the job does not need it
Some of our older material describes video identification based on facial features. That is not what we propose for people counting and occupancy, and we do not recommend it for these use cases. Biometric identification of individuals is a different category of processing under the GDPR, with a much higher bar to clear, and it does not make your footfall number any better.
It makes the project easier to approve
An anonymous counting system is a shorter conversation with your DPO, a shorter conversation with staff representatives and a shorter conversation with anyone who asks why there is a camera. Teams that start with identification often spend more time in approvals than in deployment. This is a commercial argument, not only a legal one.
What you still have to do
Anonymous processing does not remove your obligations. You still need a lawful basis, visible signage where people are counted, a defined retention period and, for systematic monitoring of a public or work area, usually a data protection impact assessment. We help you document what the system does and does not capture. Your legal counsel or DPO signs it off, not us.
What people counting does not solve
Four situations where we would tell you not to buy this, or not to buy it yet. If one of these is your case, a discovery call is still useful: it may save you a procurement cycle. And if what you keep losing is equipment rather than visibility of people, tagging the equipment is the direct route, which we cover on our indoor asset tracking page.
It does not tell you who someone is
If your actual requirement is to identify named individuals, for example to prove that a specific employee was in a specific room, this is the wrong page and probably the wrong technology. Anonymous counting is designed not to answer that question.
It does not fix cameras pointed the wrong way
Most CCTV was installed for security, aimed at doors, tills and corridors, often at a steep angle. Those views can be excellent for evidence and useless for counting. If your coverage is wrong, the honest answer is new camera positions, not new software, and that is a cabling job with its own budget.
It cannot count what the camera cannot see
People hidden behind shelving, queues that form outside the frame, areas with no coverage and scenes where dozens of people overlap in a small frame all degrade the count. Blind spots do not produce a warning, they produce a number that looks fine and is wrong.
It does not tell you why
Occupancy data shows you that a zone is empty. It will not tell you whether the layout is confusing, the lighting is bad or the product is wrong. Pair the numbers with someone walking the floor and asking people, otherwise you will optimize a dashboard rather than the space.
What it costs
Our minimum commitment is USD 5,000. We do not publish a price list, because the real number depends on decisions taken during the assessment, and most of those decisions are yours. These are the factors that move it.
How many cameras and how many zones
Cost follows the number of camera views to be processed and the number of zones you want reported separately. A hypothetical example: a museum counting arrivals at two entrances is a small project, while the same museum reporting occupancy in fourteen galleries is not, even though both are bought under the same heading.
Whether your existing cameras are usable
If current coverage works, hardware cost is close to zero. If cameras need to be repositioned or added, the installation and cabling become the largest line in the project, and they are quoted by your installer, not by us.
Where the data is processed and kept
Where the video is processed, how much is kept and for how long changes both the cost profile and the approval path inside your organization. It is one of the first things we ask about, because it is set by your policy before it is set by any product, and finding it out late is what turns a short project into a long one.
Integration and reporting
A live dashboard is one level of effort. Pushing counts into your BI tool, your building management system or your staffing model is another. Tell us early where the number has to end up.
How we work
Four steps, in this order. We stop at any of them if the answer turns out to be that you do not need this.
Step 01. Free 30-minute discovery call
We go through the three questions above, find out what decision the data is meant to support, and say early if a door counter or a booking system would answer it more cheaply.
Step 02. Camera assessment
You send a floor plan and your camera layout, ideally with a sample view from each camera. We come back with which views are usable for counting, which are not, and what would have to change.
Step 03. Pilot on a limited set of views
We run the system on a small number of real cameras in your building and compare its output against a manual count for the same period, so you can see the error for yourself before committing. How many views, which areas and over what period are agreed with you after the assessment, once we both know which cameras are usable.
Step 04. Rollout and integration
Remaining cameras and zones, reporting set up the way your team will actually read it, integration into your systems, and a written description of what the system captures for your privacy documentation.
Frequently asked questions
How accurate is people counting with existing cameras?
Sentinel Video Tracker counts with 98% counting accuracy, and resolves a person's position to 20 cm inside the covered area, which is well within what zone counting and dwell time require. We still measure the count against a manual count in your building during the pilot, because a number from someone else's building is not a number you can plan with. Camera angle, lighting, occlusion and crowd density all move the result, and the conditions that apply to your scene are established during the camera assessment. We would rather hand you a figure you watched us measure than ask you to plan on ours alone.
Do I have to replace my cameras?
Often not. Where your CCTV already covers the area with a usable view, we work with it. Cameras installed to capture faces at a door, or mounted at a very steep angle over a small area, are frequently unusable for counting. We tell you which is which during the assessment, before you spend anything on hardware.
Does this use facial recognition?
No. For counting, occupancy and flow, the system detects that a person is present at a position and does not identify who that person is. No face matching, no identity, no profile carried between visits. If you have read an older description of our video product that mentions identification from facial features, that is not what we propose for these use cases.
Is this GDPR compliant?
Compliance depends on your deployment, not on the product alone, so nobody can honestly sell you a pre-compliant camera system. What we can do is keep the processing anonymous, document precisely what the system extracts and retains, and give your DPO something short to review. You still need a lawful basis, signage, a retention period and usually a data protection impact assessment for systematic monitoring. Have your legal counsel confirm this for your jurisdiction.
Can I get the data into my own software?
That is usually the point: the counts are worth more inside your BI tool, your workforce planning or your building management system than in another dashboard nobody opens. Tell us on the first call which system has to receive the numbers and in what shape, because that is the requirement we design the integration around. It is a question we would rather answer for your stack than in the abstract.
What is the minimum budget?
USD 5,000 is our minimum commitment. Where the project lands above that depends on the number of camera views and zones, whether your cameras are usable as they are, how the data is processed and hosted, and what it has to integrate with.
Specifications
Product: Sentinel Video Tracker
Position accuracy: 20 cm
Counting accuracy: 98%, both on entries and exits at a line and on people counted per zone, validated against a manual count in your pilot
Hardware: Works with existing CCTV cameras
Identification: None. People are counted and located, not identified
Talk to an engineer, not a salesperson
Bring the decision you are trying to make and a floor plan. In 30 minutes you will know whether counting with your existing cameras is realistic in your building, what it would take, and whether something simpler would do the job. If the answer is a door counter, we will say so.