Who’s building the grocery store of the future?
The future of grocery stores will be a win-win for both stores and customers.
On one hand, stores want to decrease their operational expenditures that come from hiring cashiers and conducting inventory management. On the other hand, consumers want to decrease the friction of buying groceries. This friction includes both finding high-quality groceries at consumers’ personal price points and waiting in long lines for checkout. The future of grocery stores promises to alleviate, and even eliminate, these points of friction.
Amazon’s foray into grocery store technology provides a succinct introduction into the state of the industry. Amazon’s first act was its Amazon Go store, which opened in Seattle in early 2018. When customers enter an Amazon Go store, they swipe the Amazon app at the entrance, enabling Amazon to link purchases to their accounts. As they shop, a collection of ceiling cameras and shelf sensors identify the items and places them in a a virtual shopping cart. When they’re done shopping, Amazon automatically charges for the items they grabbed.
Earlier this year, Amazon opened a 10,400-square-foot Go store, about five times bigger than the largest prior location. At larger store sizes, however, tracking people and products gets more computationally complex and larger SKU counts become more difficult to manage. This is especially true if the computer vision AI-based system also must be retrofitted into buildings that come with nooks and crannies that can obstruct camera angles and affect lighting.
Perhaps Amazon’s confidence in its ability to scale its Go stores comes from vertical integration that enables it to optimize customer experiences through control over store format, product selection and placement.
While Amazon Go is vertically integrated, in Amazon’s second act, it revealed a separate, more horizontal strategy: Earlier this year, Amazon announced that it would license its cashierless Just Walk Out technology.
In Just Walk Out-enabled stores, shoppers enter the store using a credit card. They don’t need to download an app or create an Amazon account. Using cameras and sensors, the Just Walk Out technology detects which products shoppers take from or return to the shelves and keeps track of them. When done shopping, as in an Amazon Go store, customers can “just walk out” and their credit card will be charged for the items in their virtual cart.
Just Walk Out may enable Amazon to penetrate the market much more quickly, as Amazon promises that existing stores can be retrofitted in “as little as a few weeks.” Amazon can also get massive amounts of data to improve its computer vision systems and machine learning algorithms, accelerating the speed with which it can leverage those capabilities elsewhere.
In Amazon’s third and latest act, Amazon in July announced its Dash Cart, a departure from its two prior strategies. Rather than equipping stores with ceiling cameras and shelf sensors, Amazon is building smart carts that use a combination of computer vision and sensor fusion to identify items placed in the cart. Customers take barcoded items off shelves, place them in the cart, wait for a beep, and then one of two things happens: Either the shopper gets an alert telling him to try again, or the shopper receives a green signal to confirm the item was added to the cart correctly.
For items that don’t have a barcode, the shopper can add them to the cart by manually adding them on the cart screen and confirming the measured weight of the product. When a customer exits through the store’s Amazon Dash Cart lane, sensors automatically identify the cart, and payment is processed using the credit card on the customer’s Amazon account. The Dash Cart is specifically designed for small- to medium-sized grocery trips that fit two grocery bags and is currently only available in an Amazon Fresh store in California.
The pessimistic interpretation of Amazon’s foray into grocery technology is that its three strategies are mutually incompatible, reflecting a lack of conviction on the correct strategy to commit to. Indeed, the vertically integrated smart store strategy suggests Amazon is willing to incur massive fixed costs to optimize the customer experience. The modular smart store strategy suggests Amazon is willing to make the tradeoff in customer experience for faster market penetration.
The smart cart strategy suggests that smart stores are too complex to capture all customer behaviors correctly, thus requiring Amazon to restrict the freedom of user behavior. The more charitable interpretation, however, is that, well, Amazon is one of the most customer-centric companies in the world, and it has the capital to experiment with different approaches to figure out what works best.
While Amazon serves as a helpful case study to the current state of the industry, many other players exist in the space, all using different approaches to build an aspect of the grocery store of the future.
According to some estimates, people spend more than 60 hours per year standing in checkout lines. Cashierless checkout changes everything, as shoppers are immediately identified upon entry and can grab products from the shelf and leave the store without having to interact with a cashier. Different companies have taken different approaches to cashierless checkout:
Smart shelves: Like Amazon Go, some companies utilize computer vision mounted on ceilings and advanced sensors on shelves to detect when shoppers take an item from the shelf. Companies associate the correct item with the correct shopper, and the shopper is charged for all the items they grabbed when they are finished with their shopping journey. Standard Cognition, Zippin and Trigo are some of the leaders in computer vision and smart shelf technology.
Smart carts and baskets: Like Amazon’s Dash Cart, some companies are moving the AI and the sensors from the ceilings and shelves to the cart. When a shopper places an item in their cart, the cart can detect exactly which item was placed and the quantity of that item. Caper Labs, for instance, is pursuing a smart cart approach. Its cart has a credit card reader for the customer to checkout without a cashier.
Touchless checkout kiosks: Touchless checkout kiosk stations use overhead cameras that verify and charge a customer for their purchase. For instance, Mashgin built a kiosk that uses computer vision to quickly verify a customer’s items when they’re done shopping. Customers can then pay using a credit card without ever having to scan a barcode.
Self-scanning: Some companies still require customers to scan items themselves, but once items are scanned, checkout becomes quick and painless. Supersmart, for instance, built a mobile app for customers to quickly scan products as they add them to their carts. When customers are finished shopping, they scan a QR code at a Supersmart kiosk, which verifies that the items in the cart match the items scanned using the mobile app. Amazon’s Dash Cart, described above, also requires a level of human involvement in manually adding certain items to the cart.
Notably, even with the approaches detailed above, cashiers may not be going anywhere just yet because they still play important roles in the customer shopping experience. Cashiers, for instance, help to bag a customer’s items quickly and efficiently. Cashiers can also conduct random checks of customer’s bags as they leave the store and check IDs for alcohol purchases. Finally, cashiers also can untangle tricky corner cases where automated systems fail to detect or validate certain shoppers’ carts. Grabango and FutureProof are therefore building hybrid cashierless checkout systems that keep a human in the loop.