According to a report by Tyco Retail Solutions, shoplifting accounted for $34 billion in lost sales in 2017. Although this represents only 2% of turnover, it is a sector with very thin margins.

According to Gartner Inc , retailers are expected to invest $200 billion in new technologies this year.

It is in the face of these figures that the Japanese start-up Vaak , founded by Tanaka Atsushi in 2017, has stood out by creating Vaakeye, an AI software intended to prevent shoplifting in stores.

It remains to be seen whether this technological advance does not leave the door wide open to preventive justice at the minority report.

VaakEye technology

VaakEye is a predictive artificial intelligence system that can spot shoplifters before they steal.

Indeed, this software can detect potentially suspicious body language . Thanks to a complex algorithm that takes advantage of deep learning, this AI analyzes 100,000 hours of video surveillance and 100 different characteristics such as clothing, face, joint movement, attributes, stride…

Using the information collected, VaakEye determines the likelihood of theft and alerts staff to these potential thieves via an app.

In addition to listing suspicious people and behavior, it is possible to know the decisive actions leading to shoplifting.

This technology obviously has a cost. It takes a monthly subscription of 18,000 yen including tax (about 143 euros) per camera for a suspicious behavior detection service, shoplifting behavior, display of the result by management screen and notification of the result by email; With a small bonus distribution of stickers “being analyzed by Vaakeye”. The next step is to study whether this cost is consistent with the loss caused by shoplifting.

The beta version was offered from March 2018. After testing in several dozen stores in Tokyo, Vaakeye was officially launched on March 5, 2019 . This project was made possible by funding of 50 million yen from the AI fund of SoftBank Group corp.

Demonstration tests conducted revealed a reduction of 77% or more in damage caused by shoplifting, and a reduction of 96% or more in the duration of shoplifting.

A Japanese-style Minority report?

Although this device may seem to infringe on privacy , and also fuel the debate on predictive, preventive justice, it is the use that will be made of this artificial intelligence that will be decisive.

Indeed, as it is designed, this software has a preventive role, not a punitive one. After detecting suspicious behavior, staff will be notified and they can then simply ask the target if they need help. This will be enough to prevent theft.

From a certain point of view, this software only makes it possible to carry out a more effective analysis of the results of a tool already in place : the surveillance cameras. The human, who previously only watched the customers of the store, will be able to concentrate more on the customer relationship for example.

Retailers, out of concern for security, have asked AI software providers like Vaak not to disclose their use of anti-shoplifting systems. But in the end, the biggest retailers have already adopted AI in many aspects of their business such as helping with inventory management, optimizing deliveries, live discussions with customers…

Once again, it is a question of monitoring the use made of this software, not of prohibiting it. Vaak has, for example, other more worrying prospects such as the analysis of consumer behavior in stores for marketing purposes, or automatic payment systems based on video, or even the detection of suicides or suspicious behavior in public places …

You can also download this article in PDF here .