People counting is the foundation of traffic analytics in retail, malls, airports, and any space open to the public. But before you can benefit from the data, you need to install the system - and the investment required varies enormously depending on the technology you choose.
Important
This article is aimed at companies that want to make data-driven decisions about their spaces. If your organization is not yet ready to act on traffic data, the most economical approach is simply not to install any system.
How People Counting Technology Has Evolved
Evolution of people counting technologies: from infrared sensors to AI on security cameras.
Option 1: Traditional Counting Sensors
Specialized people counting sensors have been the standard solution for 30+ years. Their installation follows a well-defined architecture:
Hardware per point
Each counting point requires a dedicated sensor (infrared, stereo, or ToF/LiDAR), typically installed above the entrance.
Local connectivity
Each sensor connects to a local concentrator or directly to the store network to send data.
Cloud aggregation
Data is sent to a central server or cloud platform where it is consolidated and made available for analysis.
Sensor architecture: each counting point requires dedicated hardware connected to the network.
Option 2: AI on Security Cameras
AI-based people counting uses the security cameras already installed in your spaces - or new standard cameras - to count people through computer vision algorithms:
Existing cameras
The system connects to the cameras you already have for security, without requiring new specialized hardware.
Edge or cloud processing
A server (on-premise or cloud) processes the video streams and applies the counting algorithm in real time.
Centralized analytics
Counts and behavioral data are consolidated on a single platform accessible from any location in the chain.
AI camera architecture: existing cameras feed a central processing server, then the cloud.
Cost Comparison: Sensors vs AI on Cameras
| Traditional Sensors | AI (existing cameras) | AI (new cameras) | |
|---|---|---|---|
| Hardware per point | $1,200 – $2,000 | $0 | $150 – $400 |
| Installation per point | $300 – $600 | $50 – $100 | $100 – $200 |
| Total per point | $1,500 – $2,600 | $50 – $100 | $250 – $600 |
Note: these costs do not include software licensing, cloud infrastructure, or ongoing maintenance. Those items are similar across technologies and do not significantly change the comparison.
Example: 25 Counting Points
Consider a retail chain that wants to install counting at 25 entrances across its stores:
| Counting points | Traditional sensors | AI (existing cameras) | AI (new cameras) |
|---|---|---|---|
| 25 | ~$50,000 | ~$1,875 | ~$10,625 |
AI on existing cameras is approximately 25x cheaper than sensors. Even installing new cameras for the AI system is ~5x cheaper than dedicated sensors.
Conclusion
If your spaces already have security cameras - as most retail, mall, and airport environments do - AI-based people counting is the most cost-effective option by a large margin. The only scenario where sensors might be preferred is in environments without any cameras, where privacy constraints prevent camera installation, or where extreme precision in very specific conditions is required (such as outdoor counting in direct sunlight).
Want to see how AI on your existing cameras would work in practice?
Request my demo
