Walmart Taps AI for Grocery Substitutions

Walmart lifts the veil on how it moved from manual to AI substitutions when online grocery orders come up against an out of stock.
Jamie Grill-Goodman
Editor in Chief
Jamie Grill-Goodman profile picture

Anyone who has shopped an online grocery store before knows the drill. You carefully select your items, fill your virtual cart, arrive at the store and find … not everything you received is what you asked for.

At Walmart, when an online grocery order would arrive, personal shoppers would go through a manual process to determine the best way to manage a substitution when an out of stock would occur. But with nearly 100 different factors that can go into that decision, they could not realistically manage all the variables. And during the COVID-19 pandemic, the surge in demand presented a challenge.

To improve the substitution process, Walmart turned to artificial intelligence and said customer acceptance of substitutions has increased to over 95% since the deployment of the new proprietary technology.

“The decision on how to substitute is complex and highly personal to each customer,” Srini Venkatesan, EVP, Walmart Global Tech, said in a blog. “If the wrong choice is made, it can negatively impact customer satisfaction and increase costs.”

To help ensure a substitution that would result in a happy customer, Walmart got to work on developing a deep learning AI technology solution to help identify the next best item for customers if an item they selected is out of stock. The solution considers hundreds of variables, such as size, type, brand, price, aggregate shopper data, individual customer preference, and current inventory, in real-time to determine the best next available item. It then preemptively asks the customer to approve the substituted item or let the retailer know they don’t want it, an important signal that’s fed back into the learning algorithms to improve the accuracy of future recommendations, said Venkatesan.

Now, when a Walmart personal shopper is preparing an order and hits and out of stock, the system suggests the alternative product and the tech shows where the item is located in the store.

“We continue to iterate and enhance this technology, and our customers are responding positively,” Srini Venkatesan, EVP, Walmart Global Tech said in the blog. “The best part is, throughout the entire process the system is learning and getting smarter based on customer input and actions. Our goal is never to be out of stock and never to have substitutions. But, when it happens, the technology we’ve built helps ensure customers get the next best thing.”