Retailers Combine Facial Recognition with Artificial Intelligence
Widespread use of facial recognition technology isn't just for phones anymore. Retail companies are launching projects to drive revenue, cut costs, eliminate theft and improve the customer experience by better understanding customer behaviors. The overall facial recognition market is expected to jump from US$2.8 billion in 2014 to an estimated US$9 billion by 2023, according to Crystal Market Research.
Source: Crystal Market Research
“If you think about the top 40 or top 80 companies you know, almost all of them are thinking about facial recognition, or they've all at least looked into it,” Peter Trepp, CEO of the facial recognition software company FaceFirst, told BuzzFeed. He noted that hundreds of retail stores—and soon to be thousands—have implemented his company's technology.
Last March, 7-Eleven rolled out facial recognition technology across 11,000 stores in Thailand. The system can suggest products, analyze in-store traffic, monitor product levels and even gauge the emotions of customers as they look at products. In October, the new National Soccer Hall of Fame opened in Frisco, Texas, USA. The facility created a system allowing visitors to opt for a tailored tour that serves up images and stats related to their favorite teams when they walk past certain displays. CaliBurger, a burger chain with more than 40 global locations, tested a pilot project in 2017 where a kiosk recognizes the face of returning customers to display previous orders and make ordering easier. In 2018, the chain rolled out a separate pilot project to test a payment system based on facial recognition.
But stakeholder reception isn't always straight-forward. According to In Moment's 2018 customer experience report, 75 percent of consumers find most forms of personalization somewhat creepy—and 22 percent say they'd react to that creepiness by shopping elsewhere.
Since stakeholder buy-in will have a huge impact on how facial recognition projects are received, organizations will need to design systems where user opt-in and privacy are paramount. Testing and iterating the tech based on feedback will be key.
7-Eleven's project is one of the largest facial recognition implementations ever. It'll also be heavily used. Approximately 10 million people shop at Thailand's 7-Eleven stores each day—roughly one-seventh of the country's population.
To implement the technology, Remark Holdings installed cameras connected to its KanKan system, which can recognize both faces and gestures. The system records data on customer traffic and foot patterns and how emotions change as customers move through the stores. That information can then make recommendations fueled by artificial intelligence (AI) around which products to stock at particular stores. It can also identify members of 7-Eleven's loyalty program, creating an opportunity for personalized promotions.
But with great amounts of data comes great responsibility. Remark says it has taken steps to avoid storing any images of faces on its servers. “No human faces or images ever leaves the KanKan system or goes on the public network,” the company said in a press release.
And although the company has kept quiet about the project's ROI since the rollout was complete, 7-Eleven has since announced plans to implement similar technology in stores across Japan and Taiwan.
Still, critics argue that some of these projects may be flawed at the most basic level—the very code they're built upon. A 2018 MIT Media Lab study found that some facial recognition technology is far less accurate for individuals with darker skin. When testing algorithms from Microsoft, Megvii and IBM, the gender of darker-skinned men was misidentified 12 percent of the time and 35 percent of the time for darker-skinned women. Lighter-skinned women were misidentified only 7 percent of the time and lighter-skinned men were misidentified only 1 percent of the time. This significant gap in accuracy raises concerns about bias and racial profiling, and it may hamper the benefits facial recognition projects are aiming to realize.
Before portfolio leaders rush to implement these projects, they should thoroughly vet the technology. “This is the right time to be addressing how these AI systems work and where they fail—to make them socially accountable,” Suresh Venkatasubramanian, a professor of computer science at the University of Utah, told The New York Times.
—Suresh Venkatasubramanian, University of Utah, Salt Lake City, Utah, USA, to The New York Times
Retail organizations might want to borrow some lessons learned from the airline industry. Delta, for instance, has launched and closely monitored several pilot projects of a boarding system that uses facial recognition software. The biometric scanners, which are optional to use, help save time when people check in for their flights by allowing their face to act as the boarding pass. Last year, the company launched a new project in Detroit, Michigan, USA that is built upon key learnings from previous pilot projects at other airports. “This new phase will allow us to get even more feedback from customers and employees,” Gil West, Delta's COO, said in a release.—Ashley Bishel