Our website uses cookies

Cookies enable us to provide the best experience possible and help us understand how visitors use our website. By browsing Essential Retail Magazine, you agree to our use of cookies.

Okay, I understand Learn more

Ocado Technology adds machine learning fraud detection

Ocado has added sophisticated fraud detection and prediction to its analytics offering. The new system uses an advanced machine learning algorithm, developed using TensorFlow, powered by Google Cloud. 

Part of the Ocado Smart Platform (OSP), which the retailer licenses to grocers including Morrisons and Groupe Casino, the fraud detection technology will allow OSP users to rapidly detect fraudulent patterns through an intelligent machine learning model, which will learn and adapt to trends to better spot and block incidents at a faster pace.

Speaking to Essential Retail, James Donkin, general manager for the Ocado Smart Platform, said this technology uses machine learning to better identify fraudulent transactions.

“As a general rule we look for transactions where no one paid us for goods,” he said. “One advantage with machine learning is that it can start to pick up patterns than we spot manually.”

He explained that with hundreds of thousands of orders made every week, there was a need for something more scalable for businesses than what a human could identify. For example if someone was buying large amounts of products with a resale value, such as razorblades, alcohol, perfume or even baby formula, this pattern would be detected and highlighted as a potentially risky order.

Ocado said the motivation behind using machine learning for fraud detection was twofold: speed and adaptability. As fraudsters change their tactics, machines can learn new patterns and adapt far quicker, reducing the burden on the human analyst who only have to perform a final check to confirm whether they should cancel the order or not based on the prediction made by the model.

The company said: “The model has been a great success, improving Ocado’s precision of detecting fraud by a factor of 15. However, we are keen to continue improving. We are now tackling our next challenges: investigating algorithms that could allow us to explain our predictions in more detail, assessing whether we can transfer learnings from one retailer to another, and considering what tools could help us to streamline our process.”