Original article can be found here (source): Artificial Intelligence on Medium
Pragmatic AI for Retail and Beyond
The growth of online retailing has led many pundits to predict the demise of the retail store and the shopping mall. In a few years, many believe that Americans won’t ever go to the store, they will simply click a few buttons to dispatch a delivery truck.
I see a different future, one where brick-and-mortar retailers transform their businesses using predictive AI and machine learning. It’s a future where retail shops become dynamic, responding to the needs and behaviors of their customers, where shopping malls again become destinations — where the “wow” factor comes back to shopping.
The key lies in data that’s readily available today: The visual information gathered by the cameras already installed in shops, malls and other places for security or loss prevention. Capture those images and convert them to anonymized digital “objects” and you can analyze and understand the movements that occur in a physical space. The result is a vast database from which AI and machine learning can extract crucial insights about everything from what to put in a store’s window to the best location in a mall for a specific store.
This is what Deep North has developed: Technology that offers our customers decision-making insights necessary to be successful in the Age of AI. We are looking at nothing less than a transformation of the customer experience that a business is able to offer its customers within a physical space.
Our goal is to give our customers a detailed understanding of consumer behavior as it relates to their specific physical environment. We use computer vision and AI to provide real-time insights that support profitable decisions by retailers (and those in other industries, from commercial real estate to transportation hubs). We provide the ability to create a kind of personalization in the brick-and-mortar world by providing insights that previously could only be captured online.
Our technology is designed to do nothing less than enable retailers to make shopping a far different experience for their customers — and in the process improve their bottom line.
The Contrast Today, The Difference Tomorrow
If you want to understand how this will change retailing, think about buying something online today versus in a mall.
Online, you enter a search for “men’s blue shirt with oxford collar.” You get back a list of options. Click on one and put it into your basket, and now you’ll be asked if you want to add some jeans, shoes or a belt to your purchase. You’ll get that shirt; the merchant gets a nice upsell — and gets data it can use to target other offers to you.
Look for that shirt in a mall, and first you have to decide which stores to visit. You look in the windows of one or two and don’t see what you want. You go into third, find some shirts, but not quite what you want and not in your size. Nobody comes to help you. You rummage around, find a different shirt in your size, walk to the register, see the long line, and leave without buying anything.
We believe that should be a far more satisfying experience if the retailer could visualize this physical space as a kind of web page that changes in real time with customer traffic. This is what our technology does. We can (anonymously) understand the movement of customers in the store and what movements and actions lead to conversion, so the retailer can optimize the placement of merchandise.
The system learns what time the store (or specific areas of the store) has peak traffic so the retailer can assign staff to focus on operational efficiencies. The page can alert the retailer if a customer has been lingering by a certain display for more than a minute and send a staff member to work with the customer to find them the right items and upsell. It can even let the staff member know there’s a long line at the register and to offer the customer mobile checkout.
Just knowing how many people are in a store at any given time is an important data point for a manager. It can be the difference between deciding to open another checkout counter and line abandonment. It can help identify times when high traffic doesn’t translate to high conversion, and along with analysis of gestures and behavior, help identify the reason. Right now, most stores don’t have any data until a customer swipes a payment card.
That’s a far different and more customer-centric experience than we see today — and for the store, a much better road to sales. And it’s only the beginning of how harnessing existing visual assets with AI and machine learning will transform retailing.
Beyond the Store to the Mall
This technology creates a mega-opportunity for mall operators and developers to understand everything from location to marketing to get people to cross the magic 50-yard-line.
For example, the insights from shopper behaviors, traffic flows and heat maps make it easy for mall management and prospective tenants to identify the ideal location for a new store. It will provide insights on how shopper behavior differs early in the week compared to weekends, so that individual stores and the mall itself can do more effective on-site marketing and presentation.
These insights go beyond shopping to other areas of the customer experience. Traffic patterns can inform the design of maintenance schedules, so that rest rooms, for example, are serviced not on a fixed rotation but based on current usage. Analyzing views of parking lots along with local traffic reports and even the weather can help make decisions on which lots to open and where to direct traffic. It can alert management to the approach of emergency vehicles, environmental issues, even people where they should not be.
This will let malls fulfill their role as a destination center providing a great experience to shoppers, increasing the time and money people spend there.
Visualizing with Existing Technology
The idea of using data to provide insight to retailers, shopping centers and other high-traffic locations isn’t unique to Deep North. There are companies who try to do that with sensors, RFID beacons, Wi-Fi and cell phone tracking. Two problems arise quickly. One is the lack of accuracy — the data collected is only 30–60% accurate and then extrapolated. The other is the enormous cost of the technology along with the software — which requires a team of programmers to get anything beyond stock analysis — plus being locked into a single vendor.
Visual systems make a lot more sense, but as cameras are already in place, you can’t ask customers to install special cameras for the purpose. The customers we see have no interest in deploying and paying for additional infrastructure, when the average store has several cameras already, and a typical big box store has dozens. Making our vision work means quantifying what a camera can see — and seeing through the cameras already in use.
In fact, we don’t need access to all those cameras. With the information from 60–70% of them, we can let you visualize the performance of every store in a mall at any time and see what’s happening on a tablet, from anywhere. Not only is vision-based analysis more accurate and less expensive, it can scale at light speed. The average chain apparel store could be online in about two hours; a major urban department store or large mall could be up and running in a weekend.
What Drives Our Vision
We’re so focused on video for a simple reason. If a picture is worth a thousand words, then at 30 frames a second, a minute of video is worth just a bit under two million words. Here, it’s two million words in context — in a map of a physical space that lets us uncover amazing amounts of data, the needle in the haystack that generates great ROI. Technology that can run on premise, on our cloud, on the customer’s cloud or the public cloud, and where customers don’t have to write a single line of code.
Yet as excited as we are about our technology, what matters more is that our approach is very pragmatic. We’re not building technology for its own sake or to create a solution in search of a problem. We are building products that customers want, not ones we think they want. Pragmatic AI really should be our tagline.
That approach is what, over the last three years, has earned us some marquee customers, who have trusted us so much they have invested in us and are on our board. Our most recent funding round comes not from the usual venture capital investors in technology companies, but from the very retailers who see the immense value and paradigm shift that springs from our technology. We can’t be more thankful, because they are our validation, the sort of validation that a company like ours rarely gets. It tells us that our customers — and their customers — are ready for the visual-driven future that is our vision.