QR code, RFID, or computer vision, 3 levels of technologies behind unmanned stores in China

Amazon Go has inspired wild passion and fire for the spreading of unmanned stores in China. All the technologies behind actually boiled down to 3 levels, at least for now.

 

First yet the lamest level, ask customers either use smartphone or self check-out device to scan the QR code to finish the transaction step. It does not require much innovative thinking and the unmanned concept simply shift the cashiers’ role onto customers themselves. However, this does not stop a string of startups popping up in this category.

 

The second level is centering on RFID, coupled with facial recognition, led by startup like Bingobox. The cost of each RFID tag in China is around RMB0.2, but can be high considering the fast-moving nature of a convenience store where every product would have to be manually tagged before stocking to the shelf. In addition, the smart tag technology is yet to be perfected, today on an average 20% of the tags do not function properly, especially reading through liquid and metal.

 

Whether in first or second level, they always ask customers to pay before leaving the store.

 

Thus comes the third level-Amazon Go without any checkout procedure – just walk out; it is so convenient to a point that so many people would stop at the end on their first trip or two in Amazon Go and ask, ‘Is it really OK if I just leave?’, told by a key executive behind Amazon Go. The holy grail here is computer vision, which takes an arbitrary scene and to understand and interpret the video (in Amazon Go’s case, this scene is understanding who took which grocery items and put them into a cart) and translate them into algorithms. By doing this, it requires a creepy setup: camera and sensor, lots of them inside the store.

 

 

It is said to avoid infringement of privacy, Amazon Go algorithms do not involve any facial recognition technology. The camera and sensor are said only tracking customers’ body movement, understanding what they do or what they remove from the shelf. Aside from Amazon Go, there are two startups in the US working on the similar solution of autonomous checkout. 

 

Standard Cognition:

 

 

Poly.ai:

 

 

In China, we identified at least two local startups which claim to achieve the equivalent sophistication of Amazon Go.

 

Jian24

The business model is working on self-built shop operation, hoping to gain solid traction nationwide.

 

Cloudpick

Its business model is simply a technology solution provider for potential clients like brands or retailers interested in launching check-out free convenience stores.

 

 

In many ways, the future of unmanned stores remains uncertain. Amazon shied away from the question of an expansionary strategy for Amazon Go at this stage and told there are no plans for Amazon Go technology applied at Whole Foods. We suspect the accuracy and efficiency of Amazon Go pending for further improvement. It is said because shoppers can enter and leave quickly, the shelves need to be constantly restocked. So Amazon said their store associates spend much of their time restocking shelves. The hype in China also seemed dampened. Bingobox, one of the media favorite picks to showcase China’s progress in the field, also stopped its pilot project with Auchan in Shanghai. Even with massive capital injection, the concept with its technology immaturity will not be that easy to scale up quickly and above all, be profitable in the short run.

 

 

Author: Cecilia Wu