Picture the scene: in the south of France, a young man walks into a supermarket, picks up a PlayStation 4, and then takes it to the fruit section, where he weighs some bananas and prints out a price tag. He slaps the tag onto the games console box, takes the said box to the self-checkout line, pays €9 for his purchase and leaves the store.
This trick is being employed by thieves wherever there are self-checkouts in use. Retailers need to consider what steps they can take to minimise the impact of self-checkout fraud in their own stores.
The cost of items stolen through self-checkout machines has been estimated as high as £3bn. Losses at self-service are typically around double that at staffed registers.
Some have even suggested that the cost-savings presented by installing self-service checkouts is being offset by the prevalence of the fraud occurring via them.
In addition, waiting in line to be served remains the mother of all pain points in retail stores, while retailers are constantly seeking innovative ways to accelerate the purchasing journey for customers, removing the need to queue, while also ensuring that items are scanned correctly and securely.
AI: the answer to fraud prevention
So what steps can retailers take to solve the issue of fraud?
At Fujitsu’s AI Centre of Excellence at the University of Paris-Saclay in France, we’ve been devising a solution that will tackle the problem head-on. And it all revolves around artificial intelligence (AI).
We’ve developed a machine learning process capable of differentiating between items at the checkout. The system has learned the key characteristics of products on sale – the shape, the width, length, colour – from a bank of images taken of the products at the point of sale.
Today, cameras fitted to the checkouts take a picture of the item being presented by the customer. The AI can read this image and assess its authenticity – if it matches the characteristics the system expects of the product in question, the sale will go through.
The system is effective even in cases in which high-cost and low-cost items have the same weight, even in instances in which a barcode has been transferred – such as that of the PlayStation 4.
Enhancing all aspects of the in-store experience
But this isn’t the only application for AI in retail. In fact, AI has the potential to solve many of the biggest challenges facing retailers today – including long queues, poor stock control, and overcrowded car parks.
Long queues are a particular problem for larger retailers. It’s not good customer service to leave shoppers waiting in line for a long time, and it has an impact on the brand’s reputation. Even the sight of a long queue in a store can encourage people to leave and make their purchases somewhere else.
To illustrate how bad the issue can be: a customer that we work with today told us that if they reduce the average waiting line by one second, they can save 1 million a year – which can be multiplied across 5,000 stores.
So clearly, the ideal is to get the checkout moving faster. Well, with AI, you can do this. You can set up a system that recognizes people – through a biometric, through an account which they verify as they enter – and charges them based on the products they pick up. They can leave the store without having to go through a till.
This not only reduces the problem of queueing; it also means that stores can be kept open all day and night since the systems can run 24/7. It also means that staff in the store won’t have to sit at the checkout – instead, they will be free to do more meaningful work, planning for the next stock intake and analysing recent sales trends.
AI also has a role to play in improving store logistics. There’s nothing worse, from a customer perspective than arriving at a shop and finding that the item you want isn’t available.
But when you plug the sales figures into the right AI algorithm, it can start to predict what you will need ahead of time. It will warn you, for instance, that in twelve hours you will be out of this product so I have ordered it for you now.
In a similar way, an automated system can monitor the car park and alert customers when there is or isn’t a free space – solving another big problem for retailers.
Perhaps the most important potential for AI is the recommendation engine. In the future, a large proportion of business will come through retailers suggesting new or additional products to their customers, based on their preferences and what they have bought before. AI will be integral to this recommendation process, whether it occurs in store or online.
Ultimately, AI will enable retailers to get the most out of the store, by solving logistical problems, providing opportunities to cross-sell, and enabling employees to deliver great customer experience.
How to get there
When we speak with retailers, the majority of them are aware of how AI will help them tackle ongoing issues like fraud, and improve customer satisfaction.
But when it comes to tech infrastructure, there is room for them to catch up. Often stores have too much legacy infrastructure, which can impede the AI implementation process. Sometimes retailers are let down by really basic stuff – like the internet connectivity in store.
This means retailers do need to make some changes if they want to access the benefits of AI – and keep up with their competitors.
In the future, we will see stores with no cashier lines, and better and better recommendation engines. But in the short term, we can expect that the importance of the live store experience will grow.
We’ll see much more behavioral analytics in use, as retailers look for new ways to identify and track shoppers’ behavior inside the store – with customer consent, in accordance with GDPR.
More than just fraud protection
Ultimately, AI is hugely valuable for retailers looking to eliminate pain points within their store.
It’s not just an important tool for the future – it’s indispensable. Retail success in the short and definitely the long term will depend on the application of AI.
And while this offers up many exciting opportunities for retailers, it will also require a bit of work too.