Fill form to unlock content
Error - something went wrong!
Your content is just a step away. Please submit below.
The future of retail technology is driven by some unlikely sources, like babies learning to speak and basketball players guarding a net. On the surface, these two actions don’t seem to have much in common. But the technologies that gave researchers insights into how each group succeeds are combined to help retailers better understand their customers.
The journey started in 2009 for George Shaw, founder and CEO of Pathr.ai, a provider of retail spatial intelligence solutions. As a graduate student at MIT, Shaw worked with Media Lab Professor Deb Roy on the Human Speechome Project—a study of how infants learn languages. For the first three years of his son’s life, Roy collected data from video cameras and microphones, examining where and when the child started learning and using words.
“Even though from what we know it’s mathematically impossible, just about every child learns to speak. Clearly there was a gap in our knowledge, and the goal of the Human Speechome Project was to begin to fill that gap,” says Shaw.
The team realized that the approach to study language acquisition could also be used to understand consumer behaviors. The Media Lab technology was installed at Bank of America and Best Buy, tracking traffic to help understand what prompts people to open a loan or buy a cell phone.
Later, Shaw worked at a sports analytics startup, Second Spectrum, that tracks players on a basketball court, applying machine learning to the data set to collect insights coaches could use to help their teams win more games.
Spatial intelligence is a cognitive layer that sits on top of #AI. “It provides higher-level reasoning and is the business #intelligence layer that says, ‘Here’s what this tracking actually means.’” – George Shaw, @Pathr_ai via @insightdottech
Connecting the Dots with Retail Technology
By studying a baby’s babble and a basketball player’s dribble, Shaw discovered that regularities and interactions in an environment could create valuable retail industry analytics. The Pathr.ai Spatial Intelligence solution uses machine learning models to track the movement of people inside stores. Spatial intelligence is a cognitive layer that sits on top of AI. “It provides higher-level reasoning and is the business intelligence layer that says, ‘Here’s what this tracking actually means,’” says Shaw.
Pathr.ai’s solution leverages existing video cameras, with devices fed into a local server. The camera’s learning model is designed to detect people anonymously, producing dots moving around a map, and those dots move into the Pathr.ai Behavior Engine.
“It’s where the playbook lives,” says Shaw. “We extract business intelligence from the movement of those dots in real time to make decisions.”
“All these operations—sending real-time analytics to the store to efficiently work or instantly identifying a customer who is in need so you can send a store assistant to help—all of this is possible with AI. We have lots of business problems to be solved and there are huge opportunities where industries can leverage AI models to improve customer satisfaction,” adds Anisha Udayakumar, AI Evangelist at Intel.
Since artificial intelligence is run on the local server, Pathr.ai’s solution requires the most compute horsepower. “We’re able to run in various environments, but the most efficient and cost-effective is with systems built on Intel processors and OpenVINO™ for our computer vision,” says Shaw. “With Intel, we have the best technical solution.”
Retail Analytics Address Today’s Challenges
This is because tracking dots allows Pathr.ai to address retailers’ biggest pain points. Lower foot traffic in stores due to an increase in eCommerce makes customers who enter a location more valuable, but staffing shortages can make it challenging to properly serve them.
“We’re able to optimize each customer’s experience,” says Shaw. “If they have a more enjoyable shopping experience, they may buy more things. And we can make more efficient use of each staff member’s time and ultimately require fewer staff hours, which have become scarcer and more expensive.”
For example, a jewelry counter inside a store may see only 10 customers a day. Instead of dedicating an employee to service a small percentage of customers, you can task the person with other work. When a customer needs help at the jewelry counter, Pathr.ai detects them and sends a notification to the employee.
“It’s zone defense instead of a person-to-person coverage,” says Shaw. “It’s a more efficient use of the people you have available through dynamic staff allocation.” (Video 1)
A major grocery store chain in the United States uses the solution’s real-time data to measure queue lengths and adjust the number of open checkouts. “In a grocery store, the checkout experience is a huge part of how grocers differentiate themselves,” says Shaw. “Many of them have similar products, store layouts, and promotions. So the experience you have at the checkout matters a lot.”
Pathr.ai measures queue lengths and understands wait times. The system can make a prediction on how long a customer with a full cart of groceries will wait. If the expected wait time goes above a threshold set by the grocer, it will notify a staff member to open another checkout. If all the checkouts are open, the grocer will open more self-checkouts.
“By leveraging OpenVINO, we provided real-time queue and wait time insights in a cost-effective manner using Intel CPU-based edge servers. This eliminated the need for expensive, power-hungry GPUs and helped our client improve in-store operations and labor resource allocation,” says Shaw.
When expanding its smaller deployments into much larger deployments, like a shopping mall, Pathr.ai had to deal with the unique processing, deployment, and infrastructure challenges these layouts presented.
The company leveraged the Deep Learning Workbench in OpenVINO to explore different configurations: Open Model Zoo to accelerate deployment; Intel® DevCloud for Edge to prototype and experiment with AI inference workloads; and the OpenVINO Model Server to deploy new model versions quickly.
According to Shaw, these technologies provided by Intel helped the company achieve huge improvement in edge analytics performance, system costs, and operating costs for its large-scale deployments.
The Future of Retail Technology Is Driven by Consumers
Going even further, Shaw sees spatial intelligence eventually being used to help retailers understand other pressing issues, such as shoplifting. “Discerning that behavior in real time in an anonymous and unbiased way can help end shoplifting,” he says. “That’s a benefit not just to our enterprise customers but to all of society.”
Meanwhile, retailers need an ability to understand changing consumer expectations and act on them to stay relevant, says Shaw. “It’s up to technology providers and retailers to align with consumer desires and expectations,” he says. “We need a way to understand what consumers want when they come into a physical location and act on that. To understand behavior, we need more and better data.”
This article was edited by Christina Cardoza, Associate Editorial Director for insight.tech.
This article was originally published on November 28, 2022.