‘Hush hush’ plan that could stop self-serve checkout theft
RIGHT now in a clutch of Australian supermarkets, their locations under wraps, a "hush, hush" plan is underway that could eliminate the temptation to indulge in a sly bit of self-serve checkout theft forever.
The days of surreptitiously swiping ginger while pretending it's a cheap potato may soon be gone. And that completely legitimate mistake you keep making of swiping an organic avocado as a non-organic one - well that won't be happening for much longer either.
That's if a new checkout technology, dreamt up by three Melburnians who met at a coding competition, catches the eye of the big retailers.
In trials, it's already managed to work out what's in shoppers' baskets before they even reach the checkout.
Keaton Okkonen, co-founder of robotics firm black.ai told news.com.au the system of cameras and sensors could virtually eliminate what he terms self-serve "stock bleed".
And he has reason to be confident. His company has just attracted $1.2 million of seed funding to supercharge their self-serve system.
Mr Okkonen has also taken a swipe at Amazon's register free supermarkets branding its technology as "stupid" and far more expensive and cumbersome than black.ai's.
The supermarkets are loathe to reveal how much they lose through self-serve registers. But in June, research firm Canstar Blue said that 7 per cent of people surveyed admitted to stealing an item without scanning it, while 9 per cent said they had scanned an item as a cheaper alternative.
The items most likely to be stolen were fruits and vegetables. Overall, theft costs Australian retailers $4.5bn a year.
HOW IT WORKS
The technology is being trialled in a clutch of supermarkets in New South Wales and Hong Kong although Mr Okkonen said the locations were "all very hush, hush" as were the identities of its grocery partners.
Unlike other measures to halt self-serve loss, which are focused on the new register, the new system actually kicks in the very second a customer takes a product off the shelf thanks to an array of sensors and cameras embedded in ceilings and fittings.
It's designed for a brave new world where it will be able to track robots as they roam the aisles or out on the streets. But for now, it's pretty good at tracking shoppers and the things they buy.
The sensors and cameras build a picture of the customer and the products they select. Having tracked the customer around the store once the checkout is reached the shopper and their shopping is recognised.
As they checkout the register can actually prompt the customer as to what the item is without them having to search through hundreds of images of fruit and veg. Eventually, the system could be so savvy, checkout won't be needed.
Mr Okkonen insisted the self-serve sensors weren't about catching casual crims.
"We're not trying to catch people who are actively malicious - most of the time when people press the wrong product, it's not intentional.
"So (with the new system) when you pick up an organic carrot, the checkout suggests it's an organic carrot which makes it easier for you to do the right thing".
And it helps the retailer's margin.
Last year, researchers at Queensland University of Technology looked into whether "moral triggers" could nudge us to being more honest shoppers. This could include the register suggesting the correct product for us.
"Everyone has a deviance threshold, everyone can be bad up to a point but all of our grey areas are a bit different," research fellow Paula Dootson told news.com.au.
"This is about changing the behaviour of people that steal just a little bit because that's actually worse for the supermarkets than the few people that steal a lot."
Up to 8 per cent of shoppers swiped the wrong item but Mr Okkonen said his system was correctly identifying produce "in the high 90s" per cent wise.
The technology was so snazzy it's claimed it can tell the difference between different varieties of the same product, say a kipfler from a Dutch cream, through sight alone.
Black.ai's system can also be used to automatically judge stock levels on shelves and reorder more cornflakes or corn if stock is running low.
Despite the data the sensors build-up on each customer, Mr Okkonen dismissed privacy concerns. He said the computer wouldn't know your identity and likened it to how cookies on computers can track individual users without knowing particularly personal details.
"We can look at your height, rough shape, skeletal structure and clothes and come up with a profile so we won't confuse you.
"But we're not trying to know if you're Jane or David; we're just re-identifying you on your supermarket journey."
The $1 million funding is a big boost for a company that's only a little more than two years old and has just nine staff. The money will allow balck.ai to hire more machine learning researchers and engineers.
"We're just a bunch of mates who love robotics and solving difficult tech problems who all met at coding competitions.
"We were all competing against each other and who better to start a company than with people who might be a little better than you at certain things?" he said.
The trio, which aside from Mr Okkonen also includes Sebastian Collier and Karthik Rajgopal, have attracted international attention and have already been whisked away by Microsoft to meet its researchers at the firm's Seattle HQ.
But he has little time for another big tech giant playing in the supermarket space - Amazon. That company has opened a handful of Amazon Go branded supermarkets with no registers at all. Like black.ai, it features sensors that detect what a customer has picked up and then deducts the amount from their back account.
But Mr Okkonen said Amazon's solution to digital grocery stores was solved through "brute force" while his system was easier to build and cost less.
"Amazon Go has thrown a huge, frankly stupid, number of sensors over every square foot. We can deploy less sensors but still map out what you're doing and what you've picked up."
Far from being a pie-in-the-sky concept for far into the future, he said within 18 months it could be rolled out in your local supermarket.
"People are already charged a mark-up based on the supermarket's losses due to the 3 to 8 per cent of produce mis-identifications, so if we can reduce that stock bleed then in an ideal world that will mean cheaper process for consumers."