Sam’s Club Launches Machine Vision Technology to Expedite Shopping

Sam’s Club is deploying vision and AI to automatically analyze members’ purchases as they exit the store. The system ensures they’ve paid for everything in their shopping cart.

Members simply walk through an arch loaded with commercially available cameras that collect images of products in the shopping cart. Algorithms, including computer vision models using various machine learning methods, analyze contents and compare them to an electronic receipt.

The new process is designed to alleviate a “pain point” in which customers line up at a club’s exit, so an employee can review their receipt and compare it to the products in their cart. “Members said they want a faster and more convenient shopping experience and consistently rated the potentially lengthy wait times at exit—especially during busy periods—as a primary pain point in their shopping experience,” Sam’s Club (Bentonville, AK, USA) says in a news release.

Since January 2024, the retailer has deployed the new exit technology at 120 stores and plans to add the machine vision technology to all 600 stores by the end of 2024.

The move to add the exit system follows the implementation in 2016 of Scan & Go, a phone app that allows members to scan their purchases as they load them into their shopping carts and pay for them with a credit card on file. By using the app, they don’t have to wait in line at the checkout registers. As of January 2024, the app had been downloaded 9.6 million times; more than one in three shoppers regularly use the app, according to a blog post.¹

Now, members who use Scan & Go and then walk through the automated exit arch, skip both lines.

Figure 2: AI-enabled computer vision algorithms process product images in shopping carts.

The Anatomy of the Machine Vision System

Arches at the exit are divided into two lanes. Each lane includes three cameras—one overhead and two others—to capture the sides and bottom of a shopping cart. Sam’s Club uses a combination of 2K and 4K cameras. The system captures images using only the ambient light inside the club.

After a shopper passes through the arch, algorithms complete these steps²:

Computer vision algorithms detect items within the cart.
More computer vision algorithms extract visual features, such as shapes and colors, from the images and then cross-reference those features with products in Sam’s Club’s inventory system.
Other algorithms match the products in the cart to an electronic store receipt, noting any discrepancies.
The system sends those results to a mobile device in the hands of a store employee stationed at the exit.

The entire system was designed in-house, and all image processing and analysis is done in the store and available in seconds, Sam’s Club says.

Although the algorithms were trained on images of all products sold at Sam’s Club before the system was pilot tested at 10 locations, training is a never-ending process. “To reach the highest levels of accuracy, we continuously train our computer vision models using millions of images of Sam’s Club items in various positions and lighting. This enables our systems to recognize items regardless of how they are positioned in carts,” Sam’s Club says in a blog post.²

Once a new product is added to the stores, it is automatically flagged, so images of it can be labeled and used for training.

“We are constantly looking at ways for Sam’s Club to be the most convenient membership club and will continue to prioritize using technology to provide a truly differentiated and delightful experience for our members,” says Chris Nicholas, CEO of Sam’s Club, in a news release.