IHOP customers are ordering more food when they order online these days, thanks in part to a nudge from artificial intelligence.
The pancake chain last year launched an AI recommendation engine on its website that suggests menu items based on what a customer has already put in their cart.
The goal was to grow off-premise sales back to their pandemic level of 40% of overall revenue, part of an initiative by parent company Dine Brands called Project 40, said Dine CIO Justin Skelton during the National Restaurant Association Show on Monday.
The brainstorming process within Glendale, California-based Dine quickly turned to artificial intelligence, and then to the idea of an AI-powered tool that could promote different items to customers.
It’s a common tactic on sites like Amazon and Spotify. But it was a new concept for full-service dining, Skelton said.
The recommendation interface pops up just before checkout, asking the guest if they’re “Hungry for more” and displaying six items that the AI has deemed would pair well with the existing order.
Start with a Spicy Poblano Eggs Benedict in your cart, for instance, and the AI suggests tacking on a side of bacon, a Big Steak omelette or one of IHOP’s Splashers fruit drinks.
The AI was trained on historical order data using machine learning. And it has proven quite effective at getting customers to bite. Skelton said that 73% of customers who see a recommendation add one to their cart, and IHOP’s digital check average has risen 10% to 15% in the year since the feature launched, including menu price inflation.
But it has yet to move the needle significantly on off-premise as a share of sales. In the first quarter, they made up 21% of IHOP’s revenue pie, down slightly from 21.7% a year ago.
Almost three-fourths of guests add a recommended item. | Screenshot from IHOP.com
The company debated whether to build the tool itself or buy it from the marketplace, and ultimately decided to go with a vendor. Time was a factor, as was cost, Skelton said. Today, it costs IHOP just $1 for every $60 the platform generates.
“From a business case perspective, it’s pretty compelling for us,” he said.
Then there was the matter of making sure it worked, which involved getting it in front of real, live customers. IHOP started by putting the tool in 15 stores, and then expanded it by increments of 100 until it was at all of the brand’s roughly 1,700 U.S. locations.
The guest experience was a top priority during this process. IHOP kept a close eye on cart abandonments to make sure the recommendation engine wasn’t deterring guests, for instance.
“If 15 went and we didn’t work, then we needed to fail out of it, step back and course correct,” Skelton said. Fortunately for the brand, that didn’t happen. “After we got through the first 15, we were able to roll pretty quickly through the next 100 and so on,” he said.
The project did hit a snag at the final hour, when Dine Brands’ legal team began asking questions about the data IHOP would be collecting via the AI. Who would own it, and where would it be stored? In hindsight, those conversations should have happened earlier, Skelton said.
“When you think about AI and generative AI, we’re going through those discussions right now, and the same thing applies,” he said.
Now that the AI engine is up and running at IHOP, the next step for Dine is to bring it to its other full-service franchise, Applebee’s, and then to growth chain Fuzzy’s Taco Shop. Scaling the software opens up intriguing new possibilities, such as the ability to make recommendations and pairings that take into account guests’ behavior across the Dine portfolio.
“We haven’t really gone to that level,” Skelton said. “It’s likely gonna be a logical step though at some point, once we get this rolled out across all three brands.”
Members help make our journalism possible. Become a Restaurant Business member today and unlock exclusive benefits, including unlimited access to all of our content. Sign up here.