[Note: This article was originally published March 21 on the Qualcomm Ventures blog.]
Since its Amazon Go announcement in early December, Amazon’s newest brick-and-mortar initiative has sent shock waves through the retail industry with its promised checkout-free shopping experience. Retailers across every segment have been scrambling to figure out the impact to their own businesses and how the Internet of Things will shape shopping journeys and shopping experiences in the not-so-distant future.
Qualcomm recently sat down with two of RetailNext’s IoT smart store pioneers, Arun Nair, co-founder and CTO, and Bridget Johns, head of marketing and customer experience, to gain their insights on the technology behind Amazon Go, the application to other retail segments and the next frontiers in the ongoing evolution of smart store physical retail.
How will Amazon Go work, and what parts of the technology solution will be immediately transferrable to other companies?
Arun Nair: “Amazon has been a bit cagey about the technology behind Amazon Go, and its communication to this point has been very high-level.
So, to ascertain the technology involved, I think it’s beneficial to first examine the shopping experience Amazon is touting, and in doing so I think there are four distinct elements to the Amazon Go solution:
- Amazon needs to know when a shopper enters the store,
- They need to know who each shopper is,
- They need to know which items shoppers picked up, and lastly,
- They need to know when the shopper exits the store
Broadly speaking, there are two categories of technologies available to solve these types of problems: one is video and the other is wireless, essentially Bluetooth beacons and RFID.
With video, Amazon can apply several computer vision techniques to identify both people and objects, and track their movement in real-time.
In theory, using just video, Amazon can solve all four elements above. Amazon can know a shopper entered the store, using face recognition they can know who the shopper is, using object recognition they can identify items as they’re being selected, and they can tell when a shopper exits the store.
However, although computer vision is very accurate, it’s not 100 percent accurate, and Amazon Go is a use case that requires 100 percent accuracy. After all, no shopper wants to see items on the receipt she didn’t buy, and I’m sure Amazon doesn’t want to give away items for free.
Amazon will likely try to solve for accuracy by using additional sensor inputs, and while each of these will have its own unique drawbacks, like video, when combined, these various sensor inputs will push overall accuracy to 100 percent.
Bluetooth beacons can be used to know when a shopper enters or exits a store, identify that shopper, as well as track movement through the store, and item-level RFID tagging allows for very reliable scanning of products in bulk. .
My guess is Amazon will tag each item in the store with an inexpensive passive RFID chip, and will scan items in bulk as a shopper exits the store. The item recognition data from RFID will likely being supplanted by real-time object recognition data from video, so that Amazon will have two independent sources of data.
As for the individual shopper, it seems that, at least for the moment, shopper identification will be done in a relatively low tech way, by scanning a QR code generated by the smartphone app as you enter the store. Once a shopper ‘checks in,’, she will be tracked using video so any items selected can get associated with her, either as she shops, via video, or as she exits the store, via RFID. And, of course, Amazon will be able to complement this data as well using beacons.”
What are the limitations of the Amazon Go concept and how might that impact its effectiveness and its transferability to other retail segments?
Arun: Outfitting a store with full video coverage is expensive, which makes deploying this technology across an entire retail chain with thousands of stores a very costly proposition.
For each product, there is a training process to teach the video analytics system what the product looks like from various angles. This makes the Go concept difficult to scale for, say, a typical apparel retailer with a large number of constantly-changing SKUs.
Also, for product identification to work well, each product needs to be packaged in a specific way that makes it clearly visible to cameras and uniquely distinguishable from other products. This is difficult to pull off if you’re not the manufacturer for all the products you stock in your store, or if the products cannot be packaged into boxes – again, a very common scenario in apparel retail.
Amazon is clearly targeting the friction point of checkout queues when shopping for groceries. What other store friction points are ripe for solutions?
Bridget Johns: “Amazon chose to tackle grocery and eliminate the checkouts queues, but this is probably not the path that all of retail will take. Other categories like apparel will need to evolve their stores in other ways. Smart fitting rooms are already becoming a ‘must-have’ for many apparel retailers, and I believe there is still a tremendous amount of opportunity to further evolve and better that experience. And, as I think about different categories of retail, each has their own unique challenges and opportunities to evolve. While the successful go-forward retail strategies won’t all look like Amazon Go, let’s all hope they look different in five years than they do today.”
How will the Amazon Go concepts affect other retail segments?
Bridget: “The ability to measure, in parallel, product movement – with RFID, for example – and human movement – through video analytics – opens a whole host of opportunities to service shoppers in new ways.
With video, retailers measure very precise movements of shoppers through large environments, and RetailNext’s Aurora all-in-one sensor further moves the needle on this technology. With stereo cameras, retailers measure very precise tracks with high degrees of accuracy and an ability to segment tracks by age, gender, employee vs customer, etc. provides a tremendous amount of information about what is happening in-store.
The advancements in RFID reader technology, particularly with new fixed readers being brought to market by companies such as Intel and Impinj, gives a very good view of where all tagged product is at all times. Intel goes a step further and embeds video into its Responsive Retail Platform to provide, for the first time, the ability to bring shopper and product movement together in one place at the same time. Software like RetailNext’s SaaS platform then takes all of those tracks and gives insights into what is happening within a store, and from that data retailers like Amazon can get really creative about what that means from a customer-facing application perspective.”
What other parts of Amazon’s e-commerce shopping experience might be coming to brick-and-mortar retail next?
Bridget: “Amazon Go’s checkout-less shopping experience takes even the web site’s one-click checkout to a new level, and that’s exactly the point of the digital opportunity that’s available at brick-and-mortar stores.
There’s plenty of opportunity for physical retail to better enable shoppers to both provide and read/watch product reviews, or compare like products to one another easily and efficiently. And, just think about the opportunities to share accessory ideas in apparel along the lines of ‘purchasers of this item also purchased these accessories.’”
Arun: “Online commerce has simple and yet very powerful search capabilities, and much of physical retail is begging for just such a solution, particularly grocery and big box stores that have hundreds of thousands of SKUs. Search would, of course, be applied differently at each retailers, as some retailers like Costco and TJ Maxx very much nurture their ‘treasure hunt’ ethos. But, certainly, finding products and having insight into product inventory levels are two friction points that greatly detract from the shopping experience for many customers.”
Amazon Go is currently in beta testing with Amazon employees only. It’s scheduled to open to the public early in 2017, at its initial 1,800 square foot location at 2131 7th Avenue in Seattle, at the corner of 7th Avenue and Blanchard Street.