Insights

Just-released AI software lets shopping center cameras capture and analyze shoppers’ movements

October 7, 2021


Matt Hudgins
ICSC

Pathr.ai’s spatial intelligence software, powered by artificial intelligence, analyzes how people move and interact in a space. It already was serving retailers, and today, Pathr.ai rolled out Sensor Layer v2.0, which can integrate with landlords’ existing security cameras and other sensors to analyze how people move in shopping centers, office buildings, factories and other commercial buildings. Commerce + Communities Today contributing editor Matt Hudgins chatted with Pathr.ai founder and CEO George Shaw to learn what the technology can do for commercial landlords.

Retailers study in-store behavior to refine store layouts and route foot traffic for increased sales, but how do you envision landlords using your technology?

Our vision always recognized that the technology we’re building has uses outside retail, and we’ve started into commercial real estate with shopping mall operators. One of the biggest uses there is optimizing lease rates based on how many people are walking by locations in the mall, recognizing the relative value of each space. It’s analogous to valuing a website by tracking how many people see a webpage and allowing landlords to set lease rates in a data-driven way. For office buildings, based on how people move through the space, we’re able to tell how long tenants are likely to wait for elevators or amenities like fitness centers. How people move through lobbies is pretty important. For example, are they visiting the coffee cart or how often are they interacting with the building staff? When landlords understand how people move through physical spaces, they can make those spaces more intelligent.

How does spatial intelligence work, and how does your software use it?

Spatial intelligence is about recognizing people or objects in images or other data and tracking their movement. What you do with that tracking is spatial intelligence. One of the first and most mature areas in spatial intelligence is in sports analytics, where they are using it to understand the plays and strategies of the game. At Pathr.ai, we’re taking ideas and cues from sports analytics and putting that into every other environment. You can think of it as which plays are being run in a retail store, which plays are happening at scale at a mall. Our software integrates with a client’s existing infrastructure to collect data and measure human movement inside a physical space anonymously. Basically, we turn the movement of people into anonymous dots with no personally identifiable information like age or gender. But you can learn a lot from how that dot moves around and that is relative to your business outcomes. You can understand how people interact, where they spend time, how they shop and so forth. Even more, you can detect suspicious behavior, enabling companies to comply with the General Data Protection Regulation and handle fraud in a nonbiased way.

What size property can the platform serve?

We are happy to work with single properties, but the system scales broadly, as well. Our initial user for the Sensor Layer v2.0 software is a major U.S. shopping mall operator. A small space like a luxury retail store just implies a smaller device. Scaling to large properties or a portfolio is just a lift-and-shift or copying the system to additional or larger servers. There’s not a lot of retraining of the system; it doesn’t have to learn each new location, and it already knows what people look like and how to track them.

How is Sensor Layer v2.0 unique among spatial intelligence systems?

It leverages existing infrastructure because the way to scale AI is to be able to use sensors that are already there. There are millions of video cameras in the world, and we’re one of the only solutions that leverages those existing cameras. Two, we’ve moved entirely to Intel chip sets. Using central processing units and video accelerators from Intel allows us to not use discrete or standalone graphics processing units, which is a massive cost savings. A lot of AI companies use GPUs, and we were one of them. But we’ve made the move to CPUs specifically so the tech can scale more readily. CPUs are less expensive and more readily available, especially with the global chip shortage, so our clients’ capital expenditure for a server is lower. Due to that, the cost to run the server is lower because it uses less electricity and it’s easier to cool it.

What is the process to use your platform?

We start out by connecting to their existing infrastructure and putting a server in place to run our software. We typically loan that server while they try us out, so it’s a very lightweight investment for a trial. During that period, typically 30 days, we are designing an analytics package specifically for that customer with the analytics capabilities most relevant to them. They ask us business questions, and we figure out the analytics that will drive those business questions. Once those 30 days are over, we then go to a subscription model paid per month and per location. They can run the software on any server or in the cloud, but we always recommend Intel processors because the software runs better on those.

You mentioned privacy above. Why is anonymity important?

We have a privacy-first stance. We feel strongly that consumers want privacy, so we’re aligning with that and we’re in a position to lead the industry to be much more privacy sensitive. We’ve also found we can still provide maximum value to our customers in a privacy-sensitive way. We are compliant with [the EU’s] GDPR and the California Consumer Privacy Act right out of the box because we’re well-anonymized. One of the only differentiations our platform makes [among individuals] is that this dot is an employee and that dot is a customer, and we make that assessment just by how the dots move.

Read on ICSCInsights