Ocado unveils AI-powered retail fraud detection system

Ocado develops AI solution to fight retail-sector fraud that has already increased detection rates "by a factor of 15"

Ocado Technology has deployed machine learning in a fraud detection system that, it claims, has already improved fraud detection on the Ocado Smart Platform by a factor of 15.

The system uses a machine learning algorithm created by the technology development arm of the online supermarket, combined with open-source software Tensorflow and running in the Google Cloud.

It will be offered as a feature of the Ocado Smart Platform (OSP), an end-to-end solution that enables retailers to digitise their businesses. The latter is used by the likes of Morrisons in the UK, Groupe Casino in France and Sobeys in Canada.

"Fraud can happen as a result of a genuine mistake (a customer entering the wrong personal details or using an expired card accidentally)," said the company in a blog post announcing the development.

"But, occasionally, it can also be the result of malicious intent. If left unchecked, fraud can propagate to other systems and companies and affect customer service."

The company wanted to develop a piece of technology capable of predicting and recognising potential fraud incidents. It taps into data taken from past orders.

Holly Godwin, from Ocado Technology, said machine learning could be powerful for detecting fraud because it "can learn and adapt far quicker" than other solutions.

She said: "The work of fraud agents is then made more manageable, as they no longer have to frantically analyze thousands of data points to establish fraud.

"Instead, they simply perform a final check to confirm whether they should cancel the order or not based on the prediction made by the model; it's a perfect case of humans and machines working together in harmony."

Once collecting data from previous orders - such as previous deliveries - the firm used Tensorflow software to implement a neural network into the system. It was then sent to the cloud.

"We created a list of features which included the number of past deliveries, the cost of baskets, and other information," explained Godwin.

"The more features we included in the training data, the more reliable the model could be, so we made sure that we were providing our model with as much information as possible (and we will continue to add more as time goes on).

Although the solution is still only in the early days, Ocado explained that tests have been successful so far. "The model has been a great success, improving Ocado's precision of detecting fraud by a factor of 15x," added Godwin.

"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."

James Donkin, general manager of Ocado Technology, told Computing that the system was developed in order to help tackle fraud (obviously), but also to improve customer service.

"This is particularly important because fraud doesn't affect only the retailer, it can also propagate to payment companies and other systems. Therefore, detecting and addressing fraud early is a win for the customer and for the retailer," he said.

He added that as Ocado had been early adopters of the Google Cloud Platform and its various machine learning services using TensorFlow was "a natural choice".

He continued: "In roughly six months, we had a deep neural network ready and then fed it data we collected from past orders, including cases of fraud.

"Our engineering team was very happy to see the fraud detection rate increase by a factor of 15.

"This is a perfect example of how businesses can leverage the power of machine learning and the cloud for real-world use cases."