Recently, Amazon unveiled the world’s first completely self-service, no checkout, grocery store — and it’s really captured the public’s imagination. Lines have stretched around the block to try the new, “no-line” experience. It’s the new topic of water cooler, coffee shop, and morning t.v. conversation. And there was even a funny skit on “Jimmy Kimmel Live!” about it (which I highly recommend).
But there is more to be excited about than the novelty factor for consumers. From a technologist’s point of view, it’s fascinating how so many new technologies (machine learning, deep learning, sensory fusion, computer vision) came together as one solution. At the same time it is even more exciting to see the re-thinking of a business model because of these technologies.
Just What Is Significant about Self-Service Shopping?
Amazon Go isn’t the first store to implement self-service shopping. China’s Bingobox (and similar stores across China and Japan) are already cashier-less. In Japan they use scanners attached to the shopping cart, almost like an Apple Store Genius Bar set up. Shoppers scan their own items as they go and then check out using payment terminals and their mobile phones.
On the other hand — Bingobox in China evolved with a clear purpose in mind that we may not think about often living in Silicon Valley — to cater specifically to their regions and the needs of the shoppers there. In North America and Europe, we are used to brick and mortar shopping. In other parts of the world, where retail stores are not as accessible, consumers benefit by having the store come to them. BingoBox convenience stores are completely mobile, so the entire inventory can be transferred to other locations to optimize for people’s preferences in shopping times and locations.
Is It Tech for Tech’s Sake? Or Is It More Than That?
As technologists, we oftentimes easily fall into the trap of, What can my newest gadget do? Or, What can my hammer hammer today?
Amazon Go is not one of these cases. They made the concerted effort to move away from a proven technology (RFID, IoT) to something new and better not because there were new emerging technologies to be utilized for their own sake, but because there were new insights they were trying to obtain beyond the scope of RFID data.
What’s Exciting about the Technology Stack for Self-Service Shopping?
To accomplish this “frictionless” shopping experience, Amazon Go took the application of technology well beyond what BingoBox and the others were doing.
We don’t know exactly what Amazon Go is using, and there is a lot of conjecture out there. Amazon is publicly saying that they’re using sensory fusion technology and computer vision and machine learning algorithms. I think it’s also obvious that they are using more (perhaps location-tracking in mobile phone, QR codes in the app that integrate with existing Amazon Prime identity, maybe radar/lidar tech to determine where you are specifically in the store).
All of this is fascinating to me, because we want to know how they are doing it, and how we can improve it and make it better and faster, so more people can more easily adopt these technologies vs incubating them for years. It also showed me that we have the opportunity to rapidly converge and simplify these different data processing techniques into either a platform or a specific type of application SDK so more retailers and shoppers can benefit.
Self-Service Shopping and the Threat to Privacy
If there are downsides to this advent in self-service shopping, then privacy concerns have to top the list. My experience in business and marketing makes me well aware of the value that this goldmine of data represents to a marketer. What it amounts to is this — shopping through Amazon Go gives Amazon access to a digital signature of the buyer’s behavior — and most importantly — intent.
They can build a buyer profile by gauging your interests and preferences, not just what you eventually buy. Before a shopper has taken an item from a shelf, they have tracked how long you linger in an area and even how long we linger over an item. This type of data gathering used to be possible only on online sites.
Many have touted the virtue of personalized/AI shopping for years, but is this what we want? These days if I click on a blue sweater on a shopping site, that history is likely to follow-me forever and I will be constantly seeing blue sweater everywhere I visit until I do something about it. Who among us still wants this? With the physical self-service shopping, there may not be this cache clearing
I think that the opportunity and the responsibility for many tech providers like Amazon is tremendous. I see the evolution of machine learning and artificial intelligence technology hitting a cross roads right now, when it comes to their impact on privacy. How will Amazon balance the benefit of automation and personalization based on history or intent, without taking away spontaneity and joy of surprise discovery?
Self-Service Shopping and Traditional Grocery Store Jobs
The rise of the self-service or friction-less checkout experience will impact not just privacy — it could impact jobs. Cashiers in grocery stores make up 32.24% of the total retail cashier workforce, and while it’s true that restocking remains necessary, even that job is a candidate for automation. The fact is, repetitive jobs are always candidates for automation.
While we may not know the specific amount of years before this catches a significant impact on the 3.5 million cashier jobs in the U.S., it does represent a wake-up call to workers and an opportunity to retailers. What retailers need to do is focus on retraining their workforce to specialize in helping shoppers with their desired outcome, not just product they are seeking. Guided, specialized customer service represents an area of growth for a retail store, and an opportunity to distinguish itself from its competitors.
Take the customer service at an Apple store’s Genius Bar, the personal shopper at Neiman Marcus, and the concierge service in a 5-star hotel — these are what makes the difference for the customer, especially in the automated hybrid digital world we live in. And automation should not hamper our curiosity, spontaneity nor creativity. Remember, there are many services that a consumer is choosing from in a walk-in/walk-out shopping experience. Retailers need to invest their workforce into these new add-on services while opportunistically embrace this new tech as appropriate.
What’s Next for Amazon and Online Retailers?
There is a lot of conjecture out there, so I might as well weigh in. I believe one of Amazon’s ultimate goals is licensing this new technology to other retailers, rather than just creating a new myriad of Amazon Go stores. Amazon has already done a tremendous job of incorporating brick and mortar stores with their online worlds to meet the goal of getting the product closer to where the customers are (for delivery or pickup), and the technology behind Amazon Go will just take this even further. We’ve already seen Whole Foods Market and Amazon Fresh complement each other creating a tier one grocery shopping experience.
This latest expansion into friction-less shopping further implies that they’re in a great market position to disrupt the retail world. Can they succeed? I don’t know… But it will be interesting to see how other online retailers and traditional, brick and mortar retailers, will respond to this.
What Are the Technical Implications Outside of Retail?
Self-service in retail is an interesting trend, opportunities beyond retail are even broader. One of my favorite examples includes self-service parking available in many large cosmopolitan cities.
Of all the tech that Amazon Go utilizes, I’m personally most interested in computer vision. I think it’s about time that we learn to better leverage video and imagery because they create very valuable insights. SAP took such an innovative approach when they helped a German soccer team prepare for the 2014 FIFA World Cup Championship with computer vision technology. With it, they could detect player movements, calculate spatial distance, understanding field dynamics. And these provide a lot of different insights to coaches and athletes without too much complexity. These use cases showcase the possibilities to the world about what computer vision can do with very basic image recognition technology.
By Ken Tsai