Retailers in the physical world are at a real disadvantage when it comes to knowing the customers in their stores, and what they do once they get there. Customers online leave behind all kinds of data in their wake — who they are, how often they visit or purchase items, what they’ve looked at and purchased, when they’ve exited, what they keep coming back to see, how much they spend. The list goes on, but suffice it to say, the jokes about digital retailers knowing more about customers than they know about themselves didn’t appear in the culture out of nowhere.
The real world, on the other hand, can be a much trickier place to gather data. Brick-and-mortar players often deal with some minimal data sets — how many customers came into a location, what sales and leading items turned out to be. This data certainly isn’t nothing, but it pales in comparison to the data troves that digital retailers have.
It’s a problem that Aura Vision CEO Daniel Martinho-Corbishley told Karen Webster he first encountered while he was finishing his PhD and working at a different retail startup. He and his co-founders realized two other things as well.
“We learned that most of retail [had] these [closed-circuit television (CCTV)] cameras that were terrifically underutilized, and they all suffered from a lack of insight on the floor,” he said.
Martinho-Corbishley’s PhD was in computer science, specifically computer visioning. When he looked at those languishing CCTV cameras, he realized that retail already (mostly) had the main piece of hardware it needed to start pulling better insights from the floor. It just needed software that could take in all that visual data and turn it into useful, actionable demographic data for retailers. Thus, the idea for Aura Vision was born.
How It Works
Simply explained, Aura Vison software plugs into a store’s existing camera system — the whole integration takes about 15 minutes, according to Martinho-Corbishley. The camera continues to do what it has always done: take in and stream video. The software takes a copy of that stream, processes the visuals with machine learning and spits back anonymized, actionable data. To make sure that data is anonymized, it is always aggregated into 15-minute blocks.
“From there, we can give [retailers] non-identifying data — things like 80 people walked in, 30 were female, 35 were between the ages of 16 and 24, 30 percent turned [and] left immediately upon entering the store to look at displays, [and] how many people converted. They get a profile of what their customer looks like and how they act, without seeing any identifying data,” he said.
The anonymization, he noted, was the first major decision the brand made when it came to designing its product — which, ultimately, turned out to be critical. There are all kinds of technological solutions for tracking consumers and their behaviors when they are in-store — beacons and mobile Wi-Fi tracking are two of the most popular — that run afoul of the European Union’s GDPR regulation. Those solutions require systems that store identifying details from consumers’ phones, and there is no easy or obvious way to get consumer consent for that kind of data storage, which GDPR requires.
Aura Vision, he noted, side steps that issue because it isn’t storing information (the demographic data is inferred from the video feed), and has the advantage of being GDPR-compliant. It is also a more simple and scalable solution than those of many competitors, which require the use of costly stereo sonic sensors throughout the retail locations they serve. Those solutions are effective, but they are technologically involved, expensive and hard to scale.
“If a [retailer] wants to track their full store, they will need a lot of sensors — a 500-square-foot store needs 10 to 20. It is costly to set that up, and it can take days to calibrate, which is days the store isn’t open,” he said.
Aura Vision’s first design challenge wasn’t computer visioning — that is where its academic and technical background began. The first real challenge, he explained, was creating something that could easily integrate with the vast majority of CCTV camera products on the market, and be applicable to a large range of retailers out of the box. More simply, it needed something powerful enough to work with any camera system, and generalized enough that the company could go anywhere and offer it.
The Big Insights
The biggest insight perhaps, according to Martinho-Corbishley, is that there isn’t one big insight that works for every retailer. Each retailer is unique, and what works brilliantly in one place will utterly fail in another, even if it looks similar on the surface.
What Aura Vision does best, he noted, is unearth those individual differentiators so retailers can better act on them. Sometimes, that means their AI will see something the store does not.
A good example was a retailer in London on the High Street that had all its exterior-facing displays pointed to the right. A fairly innocuous choice, except for the fact that 80 percent of the foot traffic in front of the store was walking left to right and not seeing the display. Once the data was in, the displays were flipped, and foot traffic in the store almost doubled.
Sometimes, the retailers see demographic information that would otherwise be hard to catch: Men in furniture stores are more likely to buy than their female counterparts, particularly if they end up in certain parts of the stores. That kind of data, he noted, forces those retailers to ask how to bring in more men, then how to direct them to the right part of the stores.
“The answer is different for every retailer — what sector are they in, how is the store laid out,” he said. “The power of what we can do with that is run A/B tests at scale. We can track results at every store location pretty easily. We can help retailers understand what physical features are drawing consumers, or if events and campaigns at the store [can] actually measure the impact.”
It is why Aura Vision sees the demand for its services most acutely among specialty retailers, he noted — particularly those with stores that have a “showroomy” feel to them.
“A retailer can be selling shoes, but selling them in a more specialized, experiential way,” said Martinho-Corbishley.
Among those specialized players, the company is particularly popular with large, well-established brands that are ready, willing and able to take in all that demographic data and apply it. Aura Vision also sees a lot of up-and-coming, direct-to-consumer eCommerce brands making the transition into the physical world of retail — and looking to have some of the same visibility to their consumers as they do in their digital shops.
Not every shop will necessarily be a great fit for this technology, he explained. Grocery stores, for example, are notoriously difficult places in which to operate — and one can see the biggest brands in the game working double time, trying to overcome the innovative hurdles.
There is still more Aura Vision would like to develop its product to do. Right now, the company is working to use its tools to help its retail partners optimize areas like store staffing. However, he noted, given the trend in retail over the last few years to shift away from the transaction itself and more toward the experience of buying, it is more important than ever that physical retailers get some insight into the journeys their customers are taking.
“That is where we are really useful for the merchant — in helping them understand how they delivered the experience, how the consumer engaged with it and how they can improve it going forward,” he said.