As cashierless shopping evolves, fintechs strip Amazon's model to its essentials

Cashierless stores have generated a lot of attention, but there’s a heavy lift to ensure the concept works outside of the relatively controlled environment of beta tests and proof-of-concept stores.

Even Amazon Go, a rapidly expanding pioneer of the cashierless concept, has had to tweak its model to allow cash and adapt to local tastes. Other developers are digging into their toolkits to manage diverse store layouts, and more importantly, sell merchants on the idea of deploying new checkout technology without much work going into the deployment.

Amazon Go glassgate turnstiles
An Amazon.com Inc. employee scans in to shop at the Amazon Go store in Seattle, Washington, U.S., on Wednesday, Jan. 17, 2018. After more than a year of testing with an employee-only focus group, Amazon Go opens to the public Monday in downtown Seattle, putting to the test the online retailer's technology that lets shoppers grab what they want and leave without paying a cashier. Photographer: Mike Kane/Bloomberg
Mike Kane/Bloomberg

“The first generation is what we saw last year,” said Krishna Motukuri, CEO and co-founder of Zippin, a San Francisco-based developer of cashierless merchant technology.

Demonstrating the store actually works, that payments register and show up on digital receipts, and people aren’t just walking out with items for free was 2018’s argument, according to Motukuri.

“Last year was a showcase or a demo store, but in a curated environment,” Motukuri said. “Since then we have worked on making the deployment process easier for stores and tweaking the technology with store locations based on the constraints that these locations have.”

Zippin is piloting cashierless technology with four retailers that it did not identify. It is also analyzing input from consumers that have used the company's concept store in San Francisco.

Amazon Go is the most recognizable brand among cashierless stores, using cameras and other methods to determine which items to charge shoppers after scanning them in through an app. Other versions of this concept rely more heavily on smartphone interaction, with retailers such as Sainsbury's in the U.K. testing the model.

In Sansbury's store, consumers are asked to scan items with their smartphones. Other chains, such as Casino in France and Bingobox in China, also use the mobile scan method to minimize retrofitting and to accommodate larger and diverse store layouts.

Retrofitting autonomous checkout into existing store layouts is a huge challenge, and it's why to date, autonomous checkout has only been deployed in greenfield stores that were essentially built around the checkout system, said Michael Suswal, co-founder and COO of Standard Cognition, which opened its first cashierless store in San Francisco in September 2018.

"Solutions that require sensors, special shelves and turnstiles exacerbate the issue,” he said.

Amazon made news recently when it was revealed the e-commerce company is using synthetic data to introduce intentional errors to train its artificial intelligence technology to make quick corrections to Amazon Go locations. Amazon has opened a little more than a dozen Go locations, but reportedly plans to open as many as 3,000 in the next few years.

As its network grows, Amazon’s challenge is to scale its Go model to accommodate different traffic flows and other variables, such as natural light, different inventory stocks and how people enter and exit stores that may be on the street, in shopping malls, office buildings, airports or other venues.

Zippin is using a similar strategy to determine how stores will perform in different environments.

“How well will the AI work?” Motokuri said. “What we do is introduce new scenarios, beyond what customers would typically do to ensure user experience.”

Part of the store design is accommodating existing technology with the addition of new hardware such as cameras. Zippin is considering the stores’ existing internet capacity. A typical supermarket has a normal internet connection, but adding extra cameras to track shopping activity could increase a store’s internet requirements.

Zippin uses edge computing, which allows internet of things data to be processed closer to where it is created. This is meant to reduce stress in data management, requiring less of an IT footprint while boosting analytic capabilities.

“We’re not using a server, but a smaller model that is on the ceiling itself,” Motokuri said. “That is important because stores don’t have the space for large services, especially smaller urban stores."

Standard Cognition's approach is also to put cameras on the ceiling in an attempt to avoid deeper tech upgrades.

“Standard designed its system to be completely camera based, and all the cameras are on the ceiling, so it's very simple to retrofit into any existing layout,” Suswal said. "The work can be done in overnight shifts so the store doesn't even have to close for installation."

The data improvements also contribute to other parts of Zippin’s strategy, such as diversifying check in and payment types. Most cashierless stores work by charging purchases directly to the consumer’s e-commerce account, such as Amazon Prime, for example. But not everyone will have an e-commerce account for a brick-and-mortar store, and not every retail chain will want to identify consumers the same way, according to Motukuri.

“One remaining friction point deciding how consumers will check into a store. Will they use a payment app or another method?” Motukuri said. "The technology has to allow for a variety of methods for both."

For reprint and licensing requests for this article, click here.
Mobile payments Digital payments Retailers
MORE FROM AMERICAN BANKER