Amazon Go: The Next Level Of Shopping Powered By Analytics Technology

Amazon today unveiled its first convenience store called "Amazon Go" in Seattle, which features hassle free shopping without any checkout lines or even any physical money / card payment machine!

Here is a video by Amazon telling more about Amazon Go.



So what is the technology that powers the Amazon Go innovation?

Amazon does not share a lot of details around the technology behind Amazon Go other than references to AI, computer vision and data from multiple sensors used to allow customers to only be charged for the stuff they grab. However here are some possibilities uncovered by GeekWire based on some earlier patent filings by Amazon:

  • The Purchase Processing Technology: Amazon's patent filings describe a system that uses technology including RFID to detect when a shopper takes an item from the shelf, and then syncs the data to a handheld device. This points towards a system logging the items as the shopper goes along, rather than a checkout POS system which causes lines in traditional stores. When customers exit the store through a “transition area,” the system senses that they’re leaving and adds up the items and charges their Amazon account.
  • The Purchase Verification Analytics: What if the inventory management system cannot confidently determine the exact purchase? The patent filing notes that the system could use a shopper's past purchase history to help identify an item when it’s picked up. For example, if the user historically purchased an item from a specified location, the probability of purchase of the same item from that location may be more. As the video mentions using multiple sources of sensor data, the data from other input devices may also be used to assist in determining the identity of items. For example, if it is determined that an item is placed into an inventory location based on image analysis, the weight of the item may be determined based on data received from a scale, pressure sensor, load cell, etc., located at the inventory location. Together these data points can help narrow the list of potentially matching items, thus generating a higher confidence score for item match. 
Ofcourse, since the patents were filed more than two years ago, it's very much possible that a lot has changed. However, the information carried in them helps explain many of the features highlighted in the Amazon Go promo video. It will be further interesting to see how Amazon Go concept tackles different challenging use cases - say for example a family going to the stores for purchase where multiple people are doing the purchase at different locations. Lets continue exploring!