Updated: Feb 27
In-Store Analytics comprises software and hardware, both have seen significant improvements in recent years making in-store analytics not only a useful tool to quantify your store’s performance, but also a must-have in order to survive against tough competition, in particular from online retailers.
The hardware is responsible for recognising humans, and if possible, also identifying them. There are various technical methods - some are simple, some smart, and some highly technical. Here are the 7 leading technologies in 2020 for counting people and “more”.
Active Infrared or “Break-Beam” (AR)
Active Infrared or Break-Beam is the most common type installed today. Every time the invisible light beam is disrupted by people walking in or out, it counts 1 strike. The AR system works based on the number of times the infrared light is broken or passed through. It comprises of sensors that have an infrared emitter and receiver. These are placed at different, opposite ends.
This method seemed the best option when it came out but for obvious reasons is outdated, inaccurate and insufficient for today's requirements. The measurements are inaccurate, it can't distinguish between entering and exiting traffic, and at the end it only gives you 1 metric: number of store visits.
The system is kind of an analog since by the end of the day, the user needs to divide the number of times that ray was passed through by 2 to get a suitable count. In addition to that, intelligent AIRs are bidirectional in their movement evaluations.
Mostly used in shopping mall entrances and chain stores.
$50 to $150 per sensor set.
The break beam sensors provide inaccurate data, especially when two people enter or leave the room at the same time
The assessment of human movement is complex. This is because the sensors rely on signal processing which gets distorted when people bring bags with them.
The system uses depth data combined with machine learning, and computer vision to count people. The system is efficient enough to count people anonymously.
To count how people use a certain space without identifying a specific person, i.e. fully anonymously. It is used typically in corporate buildings, including offices, conference rooms, and other work areas.
This technology subject to the quality and the placement of the sensor is fairly accurate (70-80% correct counting) and 100% anonymous as the depth image are unable to see faces or any detail of a person.
The system doesn’t track identity and each sensor covers a relatively small area (100sqf) which means you need many sensors if you need detailed info.
$800 per sensor + software/dashboard fees per user (depends on user volume)
These battery operated motion detectors can be installed using tape. They operate based on movements. These can easily be installed in almost under every desk since they are low power consumption, and only activate when there's a movement. Upon catching one, the device transmits the movement signal to a nearby device that has a 4g connection.
It gives seat specific data and is battery operated.
The battery operated scale needs to be checked to keep them working. In addition, with every movement, it will transmit a signal, which means the more movements, the less battery life.
Can only be used for very specific requirements, e.g. counting how often a person uses a specific desk or any common office facilities.
$50 to $200
These devices use body heat and computer vision to determine objects and movements.
Security critical environments with low people density
Can see in the dark. No privacy concerns.
Inaccuracy - because the system relies on human movement, the device fails to give accurate results when the person is still. Moreover, the results also vary when the person is carrying a laptop that emits heat.
Multiple heat objects in close or touching proximity may not be counted.
$1,500 per sensor
The system has an emitter and a receiver at both ends. These sensors bounce inaudible sounds from the people passing through them, transmitting movement.
Mostly used in the robotics industry
Low power consumption. No privacy concerns.
The accuracy of the systems is still questioned. The ultrasonic sensors need a solid platform to bounce off the sounds correctly. In case a person is wearing a fuzzy coat, the device will get confused because the waves will bounce inconsistently.
However, highly sophisticated sensors are offering better accuracy.
$50-$100 per sensor
Cameras or “Optical Sensors”
Cameras with face recognition is a very modern method. It recognises and counts faces in the store. The output data therefore can include number of people in which store area, but because of advanced face recognition software, it can also tell you the gender, age group and race of the visitors.
Most of the cameras use flat color imaging. However, smart cameras use pixel colors that change second by second. These allow us to identify and determine people or objects in the scene.
Cameras are typically used in retail stores or shopping malls, and corporate offices.
Cameras with high-end specifications carry impressive on board computer vision. These can generate enough information for people tracking, facial recognition, object detection, etc. Since the use of the machine and deep learning has aided the development of the latest tech, cameras are becoming a feature-rich method for people counting and behaviour analysis.
The main drawbacks are the "handoff problem," privacy culture and other security reasons.
Handoff problem occurs when there are multiple cameras in a place, requiring their placement to be made in a way that they don’t overlap. If not dealt with, the cameras need intelligent processing to identify people disappearing in FOVs.
Lack of privacy is perhaps the biggest of the camera issues. Some cameras can “anonymize data locally,” making the person or object obscure or downsampling the image resolution. But to the customer the cameras remain a privacy breach.
The installation has substantial impact on the quality of the data results and will require fine-tuning or even re-installation with changes in the store interior (new display may block a view).
The cost varies depending upon the intelligence and the analytics system for the device used. Companies that install cameras and charge hardware fees alongside the dashboard fee. The fee varies from $510 to above $1,000 per camera, and $150-$200 per month for the software intelligence and dashboard. Installation requires professionals and is usually as or even more costly than the hardware itself.
WiFi Tracking Or “Probe Request Sensors”
Mobile devices with build-in WiFi and/or Bluetooth are continuously sending out “probe requests” as the devices are looking for known networks to connect with, like for example your home or office Wifi. These probe requests are sent from a globally unique MAC address which means each probe request sent out comes from a unique address.
WiFi tracking routers can receive and count those probe requests. For this to work, the mobile device’s WiFi must be turned on (but NOT connected!).
Multiple routers inside a room can also triangulate the mobile device by the relative strength of the mobile device’s signal.
Mostly used in retail and airports
Because in today’s world where we call carry a WiFi-enabled mobile device, this method has become accurate (90%). Also, the prices have come down substantially.
But besides the accuracy and price, the main benefit is the ability to identify people without infringing on any privacy concerns. The unique MAC address identifies the person (actually the mobile device) without giving any info about the person or his phone - the MAC address acts like an anonymous ID number. This enables the system to know if the store visitor has been in the store before, and also to track the movement within the store.
WiFi-based People Counters are today the new leader in-store analytics hardware.
Installation is simple and can be done by store owners themselves (plug-n-play). ASAP iRetail has combined the tracking router with a temperature sensor and a smoke detector which provides additional functionality and makes the device appear less intrusive.
No demographics can be identified, such as gender, age, race, etc.
The location tracking is medium-accurate; it's accurate enough to say in which department the person is but not detailed enough to say which shelf the person is looking at.
Normal prices are average $1,000-$2,000 per tracking device. But ASAP iRetail is offering a set of 3 tracking devices for $235 via its online shop.