People Counting Software Visitor Counting, CCTV People Counting Camera Analytics and Video Based People Counting
People counting software is computer vision that turns an ordinary IP camera into a visitor counter. Instead of mounting a dedicated sensor over every door, you point the software at the cameras already watching your entrances over ONVIF or RTSP, draw a line across the doorway, and it counts every crossing in each direction. You get entries, exits, live occupancy and dwell time as data you can chart, export or send to another system. The trade is honest: a purpose-built overhead 3D sensor still counts a busy doorway more accurately than software on a wide-angle security camera, and the rest of this page is about when that difference matters and when it does not.
What People Counting Software Does That a Sensor Does Not
A dedicated people counting sensor does one job. It sits above a doorway, counts crossings, and sends a number to a dashboard. It is very good at that job and it is blind to everything else. People counting software works the other way around: it reads video from cameras that are already installed for security, so the same feed that gives you a count also gives you the footage behind the count.
That matters more than it sounds. When a store manager disputes a Saturday number, a sensor gives you a number and nothing else. Video based people counting lets you open the clip for 14:05 on Saturday and watch what the counter saw. Every disagreement about footfall data we have seen ends the same way, with someone asking to see the door. Software that counts from video can show them.
The second difference is coverage economics. Sensors are priced per door, so a building with six entrances needs six sensors before it has one occupancy number. If those six doors already have cameras on them, CCTV people counting software counts all six for the cost of the software. That is the whole reason this category exists.
Where the honesty belongs: software counting from a security camera mounted at an angle for face capture will not match a purpose-built overhead 3D sensor at a crowded doorway. If your business decision turns on single digit percentage differences in footfall, buy the sensor. If you want direction, trend, occupancy and staffing signal from doors you already watch, software is the cheaper and faster answer.
What Visitor Counting on Existing Cameras Gives You
Directional entry and exit counts
A counting line across the doorway with a direction, so an entry and an exit are separate events rather than one anonymous crossing. Direction is what makes occupancy possible at all.
Live occupancy across every counted door
Entries minus exits, summed across doors, updated continuously. Useful for capacity policy, queue staffing and knowing whether the building is actually empty at close.
Staff exclusion
Employees walking in and out all day inflate footfall and wreck conversion rate. Excluding staff doors, or a staff entrance window, is usually the single largest accuracy gain available.
Dwell and zone time
How long people stay in a defined area, not just how many entered it. This is where camera software pulls ahead of a doorway sensor, which by design sees only the threshold.
The clip behind every number
Any count on the chart links back to the video that produced it. Auditing a suspicious hour takes a click instead of a support ticket.
Export and API
Counts leave the platform as CSV or over an API so they can sit next to transaction data in whatever tool your finance or operations team already uses.
Who in People Counting Actually Publishes a Price
Buyers of people counting software spend weeks in quote cycles because most of this market does not publish anything you can budget from. We checked the public sites of the best known vendors in August 2026 and recorded two things: whether a price is published, and whether an accuracy figure is published. The result is a pattern worth knowing before you start a procurement, because the vendors that publish a price are not the vendors that publish an accuracy guarantee.
| Vendor | Price published on its own site? | Accuracy figure published? | What that means for a buyer |
|---|---|---|---|
| V-Count | Yes. Nano sensor family 299 to 799 dollars per sensor one time; BoostBI software 9 to 49 dollars per month | Yes, states up to 99 percent bi-directional accuracy | The most budgetable hardware vendor in the group. You can model a rollout before talking to anyone. |
| Density | Yes. Waffle 229 dollars per unit; software 8 dollars per unit per month for rooms on annual billing, 2.50 for desks, 15 on monthly billing | Not on the pricing page | Priced for workplace occupancy, rooms and desks. This is not a retail doorway counter, so do not shortlist it as one. |
| SenSource | No. Request a quote | Yes. Guarantees 98 percent accuracy and notes many customers see above 99 percent | The clearest accuracy commitment in the group, but you cannot budget without a sales conversation. |
| Xovis | No | No | 3D sensor specialist with a strong airport and transit reputation. Everything is quoted. |
| FootfallCam | No on the main site. Pricing runs through local distributors who quote in local currency | Not on the homepage | Distributor pricing means the same product can land at different numbers in different regions. Ask for the list before the discount. |
| RetailNext | No | No | Enterprise retail analytics sold as a program rather than a product. Expect a long cycle. |
| Surveillant | Yes. 199 dollars a month for 2 cameras and 649 for 6, which is 100 to 108 dollars per camera per month for the whole platform, plus a free plan covering 2 cameras | No single number, because accuracy on your cameras depends on your mounting angle. We would rather you measure it than take a figure from us. | Counting is one function of a video platform you are paying for anyway, not a separate line item per door. |
Two of the six specialist vendors publish a price. Two publish an accuracy figure. Only one publishes both, and it is not the one with the strongest accuracy language. If you are building a shortlist, that asymmetry is the fastest way to sort it: vendors that publish prices are selling a product, vendors that publish neither are selling a project.
How Video Based People Counting Gets Running
Connect the door cameras
Any ONVIF or RTSP camera works. You are picking the cameras that see thresholds, which is usually a small subset of the estate.
Draw the counting line
Place the line where people cross cleanly and set which side is in. Put it inside the door swing, not on the sidewalk where people loiter.
Validate against a manual count
Stand at the door with a clicker through one busy hour and one quiet hour and compare. This step is skipped constantly and it is the only thing that makes the data trustworthy.
Wire the counts into a decision
A count nobody acts on is a vanity metric. Point it at staff scheduling, opening hours or conversion before you widen the rollout.
Who Should Count From Cameras, and Who Should Buy Sensors
Count from your cameras when
- You already have cameras on the entrances and you are paying for video anyway.
- You run many sites and need comparable trend data across them more than you need laboratory precision at any one door.
- The questions you are answering are staffing, opening hours, layout and marketing effect, all of which move on percentages, not units.
- You want the footage attached to the number, for disputes and for incident review.
- You need dwell and zone data inside the space, not only crossings at the threshold.
Buy dedicated sensors when
- Money changes hands on the count. Landlord footfall clauses, rent based on traffic and paid audience reporting all deserve purpose-built hardware.
- Your doorways are genuinely crowded. Dense two-way flow through a single wide entrance is the hardest case for any camera mounted for security rather than straight down.
- You need occupancy accurate to a person for a safety or licensing limit rather than accurate to a trend.
- The entrance has no camera and never will, for example a fire exit or a loading door that still leaks people.
- Your camera at the door is a wide-angle overview shot at head height. Angle beats algorithm, every time.
There is no shame in a mixed estate. Plenty of retailers put a 3D sensor over the main entrance where the landlord cares, and run people counting software compared against the specialist systems on the side doors and the back-of-house where a rough directional count is entirely sufficient. Buying one technology for the whole building is a habit, not a requirement.
What People Counting Software Costs Against Sensor Hardware
The comparison people get wrong is software subscription against hardware purchase price, as though the sensor is a one-time cost. It rarely is. A sensor deployment carries the device, the install, the cloud dashboard subscription and an annual support percentage, and only the first of those is a single payment.
Current US ranges look like this. Break-beam and infrared counters run roughly 200 to 500 dollars a device at 85 to 95 percent accuracy. Thermal counters run 500 to 1,500 dollars at around 85 percent. Overhead 3D stereo, the accuracy leader, runs 800 to 2,000 dollars a unit at above 98 percent. AI camera counters run 1,000 to 3,000 dollars and up. Installation adds roughly 100 to 500 dollars per device, cloud software adds 20 to 100 dollars per sensor per month, and support contracts usually sit at 8 to 12 percent of hardware value a year.
Software on existing cameras skips the device and the install entirely and typically lands in the 70 to 90 percent accuracy band on a security camera that was never mounted for counting, or considerably better on a camera looking down at a clean doorway. Our own platform is 199 dollars a month for 2 cameras and 649 for 6, which works out at 100 to 108 dollars per camera per month for everything on the platform, and there is a free plan covering 2 cameras if you want to test a doorway before committing. The detailed hardware breakdown lives in our guide to how much people counting sensors cost.
The number that decides it: count your doors. Under three doors, a sensor deployment is cheap and precise and probably the right answer. Above six doors across multiple sites, hardware cost scales linearly while software cost scales with cameras you have already bought, and the arithmetic flips hard.
People Counting Software Questions, Answered
Can I do people counting with my existing CCTV cameras?
Yes. CCTV people counting software connects to standard IP cameras over ONVIF or RTSP and counts crossings on a line you draw over the doorway, with no new hardware at the door. The limit is the camera angle rather than the software: a camera looking down at a threshold counts well, a wide-angle overview shot at head height counts poorly in a crowd.
How accurate is video based people counting?
Software running on a security camera that was mounted for general coverage typically lands in the 70 to 90 percent band, while a camera aimed down at a clean doorway does considerably better. Purpose-built overhead 3D sensors remain the accuracy leader at above 98 percent. Accuracy also falls as crowd density rises, whatever the technology. We cover this in depth in our guide on accurate visitor counting systems and people counting accuracy.
What is the difference between a footfall counter and an occupancy counter?
Footfall is a running total of entries over a period, so errors show up as a slightly wrong total. Occupancy is entries minus exits at this moment, so errors accumulate rather than average out and any uncounted door leaks error continuously. Occupancy is the harder measurement, and it is the one that needs every entrance covered.
How much does people counting software cost?
Specialist vendors mostly quote rather than publish, but two do publish: V-Count lists sensors at 299 to 799 dollars one time with software at 9 to 49 dollars a month, and Density lists its unit at 229 dollars with software from 2.50 to 15 dollars per unit per month. Our platform is 199 dollars a month for 2 cameras and 649 for 6, roughly 100 to 108 dollars per camera per month, with counting included rather than sold per door.
Does people counting software identify people?
Counting does not require identity. The software detects a person-shaped object crossing a line and increments a number, and the output is a count with a timestamp. Facial recognition is a separate capability with separate legal exposure. Illinois and Texas in particular regulate biometric identifiers, so if a vendor bundles face-based demographics into a counting product, treat that as a legal question rather than a feature. This is not legal advice.
Do I need a sensor at every door for occupancy?
For live occupancy, yes, you need every route in and out counted, including side doors, fire exits people actually use and any staff entrance. Uncounted doors do not cause a small error, they cause a drift that grows all day. For footfall trend on the main entrance alone, one counted door is usually enough.
Which is better, a people counting camera or a people counting sensor?
A dedicated overhead sensor is more accurate at a single doorway and a camera-based system is cheaper across many doorways and gives you video evidence behind each number. The right question is not which is better but what the count is for: money and compliance justify sensors, operations and trend do not.
Keep Reading
If you are comparing specialist systems, start with our roundup of the best people counting software compared on accuracy and price. If accuracy is the sticking point, read what an accurate visitor counting system really delivers. For the hardware side of the budget, see people counting sensor costs by technology, and for turning counts into store decisions, people counting analytics for retail operations. Counting is one analytic among many on the same feeds, so it is worth seeing how it sits alongside the rest of our CCTV analytics platform.
Count Your Doors With the Cameras You Already Have
Connect an entrance camera, draw a line, and compare a day of counts against a manual tally before you decide anything. The free plan covers two cameras, which is enough to test a doorway properly.