EP062 - Deep Dive Amazon-Go Store concept On Monday 12/5/16, Amazon surprised the industry by debuting a new retail concept called Amazon Go.  The 1800 square foot brick and mortar convince store’s most notable feature is that it requires no checkout.  Just put your items in your shopping bag and walk out of the store when you are done.  Payment is handled implicitly (similar to paying for an Uber). The store is for Amazon Prime members only (requires a mobile app to act as a key and let you in the store), and uses a combination of camera and sensors to track shoppers and the items they put in their bags.  The store is open to Amazon employees only for this month, and will be open to all prime members “early next year.” Amazon has a website for the store, and a demonstration video Context and History In the video and on the info site, they talk about having been working on the store for 4yrs The store likely leverages image recognition technology Amazon acquired via Snaptell 2009 The store clearly leverages elements of two patents Amazon filed in 2013 and 2014 USPTO20150012396 TRANSITIONING ITEMS FROM A MATERIALS HANDLING FACILITY USPTO20150019391 DETECTING ITEM INTERACTION AND MOVEMENT  This is one of several retail concepts Amazon is testing in the Seattle area this year. The store is located in downtown Seattle on the edge of the Amazon Biodome (their new HQ complex) at  2131 7th Ave Two other grocery pickup store concepts are under construction are in Balard and Sodo (north and south of downtown), and appear close to completion. Amazon has an open bookstore open in University Village, 4601 26th Avenue NE  (North of downtown) The store is offering a number of new (to Amazon) products including ready to make meal kits (similar to Blue Apron), made to order food, and freshly prepared food. Technology The video the explicitly call out computer vision, deep learning algorithms, sensor fusion, “just walk out technology” #JWOT Computer vision is likely used for  people tracking (patent even talks about using microphones to track peoples location and cell triangulation).  This likely means the store has to be designed to have no blind spots for the camera.  That would mean no restrooms for example, and may make the technology harder to retrofit into existing stores. Sensor fusion using a combination of LIDAR (laser based radar) and cameras for detecting and tracking products Deep learning for product recognition and training.  One insider has said that they can see 30% of a product and accurately recognize it. Amazon makes the analogy to self driving cars which use a similar combination of AI/Cameras/and Sensors for example. The store may or may not use RFID.  The patents reference RFID but the video (and unconfirmed reports from insiders) to give any indication of RFID use.  This is a big deal, as requiring every product in the store to have an RFID tag adds significant operational costs.  This store is a much bigger game changer, if RFID tags are not required. Although Amazon emphasized the huge customer experience benefit of not having to checkout.  The store would also offer other advantages such as inventory accuracy (which would assist BOPIS, Out of Stocks, etc…) The store doesn’t appear to use any form of digital price tags.  This is interesting as Amazon’s previous store concept (Bookstore) uses online prices in the store (which change frequently).  If the prices are truly printed on paper on the shelf, it’s unlikely Amazon is offering the same prices as it’s website. Open Questions: How well does the technology work (what edge cases can trip it up) How well will customers adopt the new use case and which elements will be most compelling How will the store handle families shopping together, parents with kids, etc… How will Amazon address all the privacy concerns and issues with tracking shoppers in the store Will the store carry alcohol and/or tobacco, how will it handle age verification Does the store in-fact