How to Choose a People-Counting System (Buyer's Guide)

Knowing how to choose a people-counting system is harder than it should be. Every vendor claims high accuracy, every brochure looks the same, and the pricing is often hidden behind a quote form. It is easy to end up comparing apples to oranges.

This buyer's guide fixes that. It gives you five criteria to score any vendor on equal terms: hardware cost, accuracy validation, privacy, integration, and pricing model. We will walk through each one in plain English, with practical questions you can put to a sales rep.

By the end you will have a reusable scorecard. Use it to shortlist two or three vendors, run a small pilot, and validate the numbers in your own store before you sign anything.

Key takeaways

Score every people-counting vendor on the same five criteria: hardware cost, accuracy validation, privacy, integration, and pricing model.
Validate accuracy yourself with a manual clicker count over a couple of busy hours, never from a brochure figure.
Software that runs on your existing IP cameras avoids buying a sensor per door; dedicated sensors add cost at every entrance.
Favour transparent per-camera pricing over per-store quotes so you can forecast cost across many sites.
If privacy and GDPR matter, pick a system that keeps video on-site and outputs anonymous counts.

Start with the decision, not the technology

Before you look at a single sensor, write down the questions you need answered. People-counting is a means to an end, and the end is usually one of these: How many people walk in? What share of them buy something? How busy are we right now? How long are people waiting in line? Are we staffed for the rush?

Each question points to a different metric. Footfall and conversion drive merchandising and marketing decisions. Real-time occupancy matters for safety and capacity. Queue and wait-time data helps you staff the till. Get clear on which questions are worth money to you, because that determines which features you actually need.

Only then does technology enter the picture. There are four common ways to count people, and it helps to know them at a high level.

Infrared beam counters sit across a doorway and tick up when something breaks the beam. Cheap and simple, but they struggle with groups and cannot tell a person from a shopping cart.
Thermal and time-of-flight (ToF) sensors mount overhead and detect body heat or distance. Good in low light, but they are dedicated devices you buy per entrance.
3D stereo sensors use two lenses to judge depth and handle groups well. Accurate, but among the most expensive per door.
Software on existing cameras uses computer vision on the IP or CCTV cameras you already have, with no new sensor at the door.

The right choice depends on your situation. A single store with one entrance has different needs than a 40-site chain. A venue that already runs IP cameras can avoid buying hardware entirely. For a side-by-side breakdown of each approach, see People Counter Types Compared: Cameras, Thermal Sensors, Beam Counters and Wi-Fi.

The rest of this guide gives you a scorecard to compare any vendor, regardless of which technology they sell.

Hardware cost and what you actually have to buy

The biggest cost difference between systems is hardware. Dedicated sensors are bought per entrance. If you have 15 stores with two doors each, that is 30 sensors to purchase, mount, and eventually replace. The sensor price is only the start.

Add up the full picture over a typical three-to-five-year life:

The sensor unit itself, per door.
Mounting and ceiling work, sometimes with a contractor.
PoE switches and cabling to power and connect each device.
Replacement when units fail or a vendor sunsets a model.

These line items are easy to underestimate at quote time and they recur across every site you roll out.

Software-based counting changes the math. CountPort is pure software. It runs on the RTSP or ONVIF IP cameras you already own, processing the feed on a back-office PC, a mini-PC, a Mac Mini, or supported cameras. There is no sensor to buy per door. If your cameras already cover the entrance at a sensible angle, your hardware cost can be close to zero.

That is not a free lunch in every case. Existing cameras have to be positioned reasonably well, and very wide entrances may need a camera with the right field of view. But for most stores that already run CCTV, the saving over a sensor-per-door system is real and it compounds across sites.

For the full breakdown of where the money goes, read Software vs Hardware People Counters: Which Is Right for Your Stores?. You can also see how the camera-based approach works on our technology page.

How to validate accuracy before you sign

A brochure accuracy number tells you almost nothing. It was measured in a vendor's test conditions, not your doorway, and the headline figure rarely matches the messy reality of a busy Saturday with groups, prams, and people carrying boxes.

So validate in your own store. Here is a simple method that costs nothing but a couple of hours.

Pick one entrance and two genuinely busy hours.
Stand where you can see everyone coming in and use a manual clicker (or a phone counter app) to count each person.
Pull the system's count for the exact same window.
Compare the two and calculate the difference as a percentage.

Do this on more than one camera if you can, and ideally on a busy and a quiet period. If a vendor will not let you run this test during a trial, treat that as a warning sign.

It also helps to understand what affects any camera-based count, so you can rule out fixable issues before judging the software:

Camera angle. A near-overhead view counts more cleanly than a shallow side angle.
Mounting height. Too low and people occlude each other; too high and detail is lost.
Lighting. Strong backlight from a glass storefront, or very dim entrances, both make detection harder.
Wide doorways. A single camera may not cover the full width.
Groups and items. Clusters of people and large carried objects are the classic edge cases.

CountPort uses body-only detection. It identifies the shape of a person rather than any facial features, and it never uses facial recognition. That keeps it privacy-friendly, and it means accuracy depends on the factors above rather than on capturing anyone's identity.

We deliberately avoid quoting a single guaranteed accuracy percentage, because an honest number depends on your camera placement and conditions. The right way to think about accuracy is to validate it where it will actually run. For more on reading vendor claims critically, see How Accurate Are People Counters? Reading Accuracy Claims Honestly.

Privacy and data handling: the questions that matter

Not all people-counting systems are equal on privacy, and the difference matters both legally and reputationally. The key distinction is what the system stores.

Some systems capture and retain images, or extract biometric features to recognise individuals. Others process the video and output only anonymous counts, keeping no personal data at all. For a retailer in a GDPR jurisdiction, that distinction shapes your obligations.

There are three practical questions behind it: Where is the video processed? Does anything leave the building? How long is data retained? A system that processes on-site and stores only aggregate numbers is far simpler to defend in a privacy review than one streaming footage to an external cloud.

CountPort is built around this. It uses body-only detection and never facial recognition. Video is processed on-site (on-prem), so footage does not leave your building. It stores no biometric data and produces anonymous counts. You can read how that works on the counting feature page.

Whatever you are evaluating, put the same short checklist to every vendor:

Data location. Is video processed on-site or sent to a cloud?
Access. Who can view the footage, and is any footage even retained?
What is stored. Counts only, images, or biometric identifiers?
Deletion. How long is data kept and how is it deleted?

Score the answers side by side. A vendor that cannot answer these clearly is telling you something.

Integration: making the data useful

A footfall number on its own is interesting. Combined with your other data, it becomes actionable. The whole point of counting is to connect it to decisions, and that means integration.

The most valuable connection is sales. Footfall multiplied against POS transactions gives you conversion rate, the share of visitors who buy. Counts also matter next to staff schedules (are you covered for the rush?) and inside your existing reporting or BI tools (so the data lives where your team already looks).

When you evaluate a vendor, ask integration questions directly:

Is there an API to pull counts into your own systems?
Can it export to BI or spreadsheet tools your team already uses?
Does it correlate footfall with POS to calculate conversion automatically?

CountPort produces a set of metrics you can put to work right away: footfall and visitor counts, real-time occupancy, queue and wait-time, dwell, zone and heatmap views, conversion (footfall combined with POS), peak-hour patterns, and capture rate. Each one maps to a decision, whether that is staffing the till, redesigning a slow zone, or judging a window display by how many passers-by it pulls in.

One feature sits right next to integration and is easy to overlook: staff exclusion. Employees walk in and out all day. If they are counted as visitors, your conversion rate looks worse than it is, because the denominator is inflated. Staff exclusion filters employees out so the conversion number you report is honest. Ask whether a vendor offers it; without it, your figures drift.

A quick note on a related industry term. Some vendors talk about re-identifying the same person across multiple cameras to avoid double-counting. That is a real concept in the field, but it is not something CountPort does. CountPort counts per camera and excludes staff; it does not de-duplicate the same visitor across separate cameras. Be clear with any vendor about exactly what their counting claim covers.

Compare pricing models, not just price tags

A low headline price can hide an expensive model, and a higher per-unit price can be cheaper overall once you forecast it across sites. Compare the model, not just the sticker. Buyers typically meet three.

Pricing model	How it works	The catch
Per-sensor hardware plus cloud fee	Buy a device per door, then pay an ongoing cloud subscription	Hardware cost scales with every entrance; two fees to track
Opaque per-store quote	A custom price per location, given on request	Hard to forecast across sites; little price transparency
Transparent per-camera subscription	A published, flat price per camera per month	Predictable; scales cleanly as you add cameras or sites

Per-store quotes are the trickiest for multi-site budgeting. Every location can carry a different number, so building a chain-wide forecast means chasing a separate quote for each store. A published per-camera price, by contrast, is simple arithmetic: cameras times the monthly rate.

CountPort uses flat per-camera pricing. Lite is $29 per camera per month and Pro is $39 per camera per month, published openly so you can model your own rollout in a spreadsheet before talking to anyone. The numbers are on the pricing page.

For a deeper comparison of every pricing model and the hidden costs in each, read How Much Does a People-Counting System Cost? Pricing Models Explained.

Build the business case

A people-counting system has to pay for itself, and the return usually comes from three places. Higher conversion, because you can see which displays, staffing levels, and layouts turn visitors into buyers. Smarter staffing, because you schedule people against real footfall patterns instead of guesswork. Shorter queues, because wait-time data tells you when to open another till before customers walk out.

Quantifying that is worth doing before you buy. For example, a store that lifts conversion by even one or two points on a meaningful sales base can cover a modest per-camera subscription many times over. We lay out a worked example you can adapt to your own numbers in What's the ROI of People Counting? A Framework for Building the Business Case.

You do not have to commit a whole chain on day one. Run a pilot in one or two stores. Validate accuracy with the clicker method above, connect the data to your POS, and watch a real conversion number for a few weeks. A pilot turns a sales claim into evidence you can take to a budget holder.

If you already work with a CCTV or security partner, ask whether they can deliver the system for you. CountPort works through integrators and offers a white-label channel, so a buyer with an existing partner can fold counting into a relationship they already trust. See the partners page for how that works.

A buyer's scorecard you can reuse

Pull the guide together into one checklist. Score each shortlisted vendor against the same rows, and the right choice tends to make itself obvious.

Criterion	What to check	What good looks like
Hardware cost	Sensors per door vs. use of existing cameras, plus mounting and cabling	Little or no new hardware; runs on cameras you own
Accuracy validation	Can you run a manual clicker test during a trial?	Vendor welcomes an in-store comparison; no fixed-percentage hand-waving
Privacy	Where video is processed, what is stored, who has access	On-site processing, anonymous counts, no biometrics
Integration	API, BI export, POS correlation, staff exclusion	Counts flow into your tools and tie to sales for true conversion
Pricing model	Per-sensor, per-store quote, or per-camera	Published, predictable, easy to forecast across sites
Support	Onboarding, response times, pilot help	Clear contacts and a real trial path

Weight the rows by what matters to you. A single privacy-sensitive venue should lean hard on the privacy row. A tight-budget independent might weight hardware and pricing highest. A 40-site chain will care most about predictable per-camera pricing and clean integration. There is no universal weighting, only the one that fits your business.

The discipline that makes the scorecard work is asking every vendor the same questions and recording the answers side by side. It is easy to be charmed by a strong demo and forget to ask the awkward pricing question. A shared list keeps the comparison fair.

From here, the next steps are simple. Shortlist two or three vendors. Run a pilot in one or two stores. Validate accuracy in your own environment, on your own cameras, with your own clicker count. Then decide with evidence rather than a brochure.

Frequently asked questions

What should I look for when choosing a people-counting system?

Compare vendors on five things: how much hardware you have to buy, how you can validate accuracy in your own store, how the system handles privacy and video, what it integrates with (especially POS), and how it is priced. Scoring every vendor on the same criteria makes the choice clear.

How accurate are people-counting systems?

Accuracy depends on the technology, camera angle, mounting height, and lighting, so treat any single percentage with caution. The reliable approach is to run a manual clicker count over two busy hours and compare it against the system's output in your own environment before you commit.

Do I need to buy special hardware to count people?

Not always. Some systems require a dedicated sensor at each entrance, while software like CountPort runs on the RTSP or ONVIF IP cameras you already have, installed on a back-office PC, mini-PC, or supported camera, so there is no sensor to buy per door.

They can be, if they avoid storing personal data. CountPort uses body-only detection, never facial recognition, keeps video on-site, and stores no biometric data, so it produces anonymous counts. Always ask a vendor where video is processed, what is stored, and who can access footage.

See it on your own cameras

The fastest way to test any of this is on your own store. We will walk you through running CountPort on the cameras you already have, and you can validate the count yourself with the clicker method in this guide. Request a demo and we will set it up with you.

How to Choose a People-Counting System: A Buyer's Guide