Artificial IntelligenceCryptocurrencyMetaverseCybersecurityTech Forward

Retailers Are Using Cameras to Help Keep Their Shelves Stocked. Here’s How

January 23, 2019, 2:00 PM UTC

Retail store employees spend countless hours making sure that shelves are stocked with enough cereal, toilet paper and toothpaste for customers. At best, it can take workers half their shift to notice that the inventory is low. But at worst, they fail to notice out-of-stock products for days, resulting in thousands of dollars in lost sales.

Trax, a startup focused on computer vision, hopes to help store owners avoid some of that pain. It’s developing cameras and artificial intelligence that keep watch on stores shelves and automatically tracks when product inventory is low.

If the technology detects a shortage, it sends alerts to management and store employees. It can also send sales data to a store’s suppliers so that they know in real time what products have actually made it on to the shelf and what was sold at specific stores.

“We’re reducing labor costs––real major operational issues today,” says Dror Feldheim, Trax’s co-founder.

The use of computer vision combined with artificial intelligence in retail is still in its infancy, but the technology is getting a lot of attention, not just with small mounted cameras, but also with robots that patrol store aisles to track inventory on store shelves.

Variations on the idea are pitched as a possible boon to retailers in dealing with staff shortages, high employee turnover, and constant changes to their product lineups. Several big chain stores are already experimenting with the technology.

Walmart is testing robots that scan shelves at 50 stores across the country. Meanwhile, home improvement store Lowe’s has developed robots and shelf cameras that it is now testing. And finally, Kroger’s has partnered with Microsoft to create “digital shelves that can show ads and change prices on the fly along with a network of sensors that keep track of products,” according to Bloomberg.

Initially, Trax, founded in 2010, took a hybrid approach by having humans take images of store shelves and comparing those with inventory lists, visual maps of stores that show products on shelves, and sales data from a partnership with Nielsen. In the background, technology stitched the images together so it could review inventory levels and alert the client—the retailer or consumer packaged goods company that supplies the products—within a few hours about any shortages.

But Trax eventually changed its strategy to developing miniature battery-operated, wireless cameras that could be mounted on the edge of a shelf. The wireless camera, about the size of a skinny deck of playing cards, sends eight images an hour––deleting any photos with human faces.

The data first goes to local servers for on-site image processing and then to a cloud data center where Trax’s technology creates 3-D visual data. Using computer vision, the technology is supposed to provide insights about inventories that are then sent to store associates.

Depending on its size, a store may need anywhere from 500 to 600 cameras. With that kind of installation, Trax says its technology is 96% accurate.

In 2018, Trax started implementing its technology in several countries. Watson, the largest pharmacy chain in China, is running a multi-store test with Trax’s technology for its hair care and beauty products, two of its fastest-turning categories.

In the U.K., Trax worked with Tesco, the world’s third largest retailer, on product categories that including wine, beer, and spirits. For Tesco the test wasn’t about reducing the number of employees.

Rather, Feldheim says the hope was that employees would have more time “in front with clients” and replenishing stock, rather than spending a lot time checking which products were running low. Pleased with the early results, Tesco plans to ramp up its test from two stores last year to a handful of stores this year.

Trax is also working with Shufersal, the largest supermarket chain in Israel, on categories like salty snacks, cereal, and coffee. Trax debuted its technology in three of those stores and is rolling it out to fifteen more this year.

“We know that availability and out-of-stock are easy to measure and a practical issue to deal with,” says David Laron, Shufersal’s chief operating officer.

Before the cameras, some of Shufersal’s key products could be out of stock for 45 hours before being replenished. After installing Trax’s technology, the delay dropped to 30 minutes—resulting in increased sales of 3% to 5% in certain product categories.

“When you think of customer satisfaction drivers, what drives retail success is, ‘Is the product there that I want?’” says Mark Cook, vice president of retail at Trax. “Not being on the shelf equals a loss of sales.”

In addition to Trax, a number of startups are working on their own versions of camera-based inventory tracking. Simbe Robotics, a Silicon Valley startup, built Tally, a robot being tested in Target stores. Anheuser-Busch is working with Focal Systems, a Bay Area startup, to track its products inventory in convenience stores. And Standard Cognition, in San Francisco, has signed on Paltac, Japan’s largest consumer packaged goods wholesaler, as well as several unnamed retailers in the U.S.

Earlier this month, Trax announced a partnership with Google to digitize every product on retail shelves—basically turning physical objects into online visual data so that they can be more easily monitored by Trax’s technology (12 billion products have already been digitized).

Google isn’t the only cloud service Trax uses, but now it’s the preferred one. Google’s Edge computing system will help Trax process the images, stitch them together, and then convert them into data that can be analyzed so that Trax can tell its clients about their inventory.

In addition to testing Trax’s cameras, Shufersal plans to start experimenting in one of its stores with a robot that Trax is developing with a third party. Using robots with cameras to track inventory lets retailers avoid having to mount hundreds of cameras on shelves and opens the door to collecting more kinds of data.

For example, an owner may want to keep tabs on the meat counter or notice whether customers stop in front of promotional materials.

Trax hopes that the combination of robots and cameras would help stores solve problems they didn’t even know they had. It could help tell owners whether Coke should be on the right of Pepsi, for example, or whether the store should put fewer products from a particular brand on its shelves.

“For some stores a robot is good,” said Laron. “For some stores the robot isn’t relevant, maybe it’s not efficient or it bothers the customers. In a big store, robots should be more efficient.”