When Marvel released its first Iron Man movie in 2008, the studio replaced Edwin Jarvis, Tony Stark’s loyal butler in the comic books, with J.A.R.V.I.S., an artificial intelligence (AI) assistant that both runs Stark’s home and acts as a sounding board. For most people, J.A.R.V.I.S. remained an only-in-fiction concept until 2011, when Apple introduced its natural-language assistant, Siri. Suddenly, the floodgates to the future of AI were open.
Today, virtual assistants help users find information, remember appointments and errands, and even shop. In a retail-driven capacity, for example, Google Home and Amazon’s Alexa will place orders on command.
Run out of toothpaste? No biggie—all you have to do is ask, and Alexa will send more.
Retailers are also using predictive algorithms and machine learning to suggest relevant products to consumers based on past purchases and other retrievable personal data, and chatbots that help shoppers solve their specific needs. Together, these innovations are transforming the way consumers are able to shop online, an activity which has remained largely unchanged for more than a decade.
As Nikki Baird writes for Forbes, “For retailers, AI-driven personalization is an opportunity to gain access to relationships between products and customers that were previously hidden within the wash of clickstream data.”
What does AI-driven retail look like?
There are three main fields where AI offers retail growth potential: replenishment, complementary purchases, and predictive retail.
Replenishment is the most straightforward. When a customer buys a consumable product—whether it’s laundry detergent, toothpaste, shampoo, or cat litter—the retailer knows the customer will eventually exhaust the item. The surest way to keep that customer returning for additional purchases is through a replenishment program. Amazon is the leader in the field, with two customer-driven methods for replenishment: a recurring delivery option, which lets customers decide in advance how frequently they want their items to be replenished, and the Dash button or Wand, which lets customers reorder with the touch of a button or scan of a barcode. Pet supply specialist Chewy.com is another online retailer that offers auto-replenishment. If Fido goes through a bag of dog food every four weeks or so, you can easily set up your account so that your preferred brand ships to you just in time to keep him in kibble.
Advance scheduling and buttons aren’t the only means of encouraging follow-up purchases. For retailers that offer loyalty programs—like Sephora, Bloomingdale’s and Walgreen’s—email is an easy solution to anticipate customers’ needs, reminding them that it’s time to top off their supply. For example, Sephora’s Beauty Insider program logs every product a shopper purchases. Because Sephora knows how frequently shoppers buy a given product, it can send “Restock Your Stash” reminder emails when customers are likely to be running low.
Retailers are also using predictive algorithms and machine learning to suggest relevant products to consumers based on past purchases and other retrievable personal data, and chatbots that help shoppers solve their specific needs.
Complementary product suggestions offer the opportunity to upsell a customer right away by reading into how they’re browsing or buying. Lingerie brands such as Victoria’s Secret, ThirdLove, or Aerie offer metatag-enabled suggestions for coordinating panties on every bra product page, as a lightweight example of this, and Target uses a similar approach with any items that can be purchased as sets, like swimwear or even dishes. Retailers can also offer maintenance and care products with purchases, such as a delicates detergent or a laundering bag with lingerie, shoe polish or leather conditioner with a pair of shoes, or stain shield for a rug.
For a retailer like Gilt Groupe, the opportunities to upsell with complementary product suggestions can be endless. If a customer buys a hotel stay through Gilt Travel, Gilt can recommend an entire suite of items for the same trip: sunscreen, sandals, and swimsuits for a beach stay; location-specific restaurants and spa treatments through Gilt City; even a new set of discounted designer luggage.
Take these suggestions a step further as AI becomes even more advanced, and it’s not difficult to imagine how new retail technologies could analyze your Google calendar and offer targeted suggestions based on specific itineraries, personal preferences, and purchase history. Instead of having to specify a schedule for a regular shipment of toilet paper, AI will learn from past behavior and other data inputs to know when you are running low.
And with that, we come to predictive retail, which is by far the most advanced form of AI to hit the industry—but also the most prone to error. To generate suggestions, machine learning technology applies what it knows about a customer from past searches and purchases, product feedback, and other data to customize the online experience and present the items it thinks the customer is most likely to buy. Of course, there are variables that can affect those predictions. If friends or family members share an account or device, it could skew the results. Furthermore, past searches don’t necessarily signal what future purchases will be.
Instead of having to specify a schedule for a regular shipment of toilet paper, AI will learn from past behavior and other data inputs to know when you are running low.
Right now, the traditional mode of personalization is to show the customer more of the same. If a shopper looks at button-down shirts, the site will suggest more button-down shirts. Qubit, an AI-personalization platform driven by heuristics, is banking on an alternative model: suggestions using notions about how buyers learn about new products, which products they really like, and which factors most influence them to buy.
Nick Smyth, VP of Sales in North America at Qubit, explains to Retail Dive, “Just because you looked at that shirt doesn’t mean you’re interested. We shouldn’t make giant inferences from limited data sets.” Qubit’s heuristics model, by contrast, allows brands and retailers to more accurately identify and segment their customers by understanding some of the “mental shortcuts” they take to their purchasing activities, like if they buy products according to the latest trends or supply limitations.
AI can make commerce effortless
AI and machine learning can help brands anticipate and cater to customer needs in a way that saves time and increases loyalty. With Sephora’s reorder emails, customers never have to leave home or even go back and browse the web to find the products they need. With Neiman Marcus’s Snap. Find. Shop app, customers can submit photos of products they like, and find similar items without searching through a store or endless web listings. With The North Face’s IBM Watson-powered chatbot, customers receive help finding the right jacket for their specific weather conditions without the hassle of reading hundreds of style descriptions. Collectively, that information can help retailers grow its repeat customer base—offsetting the initial AI investment and cost per acquisition—and create a more compelling and profitable product assortment.
Retail AI might not be as advanced as J.A.R.V.I.S. yet, but—much like Iron Man’s trusty digital sidekick—it can help the industry prevail in the face of competitive challenges, consumer expectations, and a constantly-evolving landscape.