In 2014, the European Central Bank estimated that the total amount of Eurozone money lost due to fraud was close to 1.33 billion euros. 60% of this money was lost through fraudulent internet transactions. Newly emerging criminal practices, ineffective fraud detection systems, and the enormous amount of money involved inspired us to develop machine learning technology to more effectively combat this growing fraud trend.
Historically, online retailers have relied on simplistic transaction decision rules to combat fraud, and they still review a substantial proportion of transactions manually. More recently, specialist anti-fraud service providers exploit artificial intelligence techniques within their products, which can be integrated with third-party payment systems. Dimebox, however, offers a complete end-to-end payment and anti-fraud solution.
A self-learning solution
Machine learning is a branch of artificial intelligence aiming to detect patterns and provide important insights by creating self-learning algorithms. The goal is to design algorithms that do not require explicit programming, learning to find patterns by observing large quantities of data. Whether it’s image recognition, text translation, or fraud prediction, programming computers to solve certain problems is no longer necessary. Computers are now able to spot fraud patterns without human intervention.
At Dimebox, machine learning is a crucial aspect of our risk management strategy. In the past, fraud detection algorithms could only detect fraud according to predefined rules that were often crude; for instance, blacklisting whole countries, or requiring shipping and billing addresses to match. This hardwired approach has proven to be insufficient, as fraudsters constantly improve their techniques, change their habits, and legitimate customers are not prepared to tolerate declined transactions.
Anti-fraud systems need to evolve beyond these outdated methods, which are cumbersome to manage, and too slow to adapt to new fraud patterns and changing customer behaviours. The focus needs to be on building fast and dynamic systems that continually adapt to new fraud patterns and customer behaviours in real-time, automatically.
Increasing conversions while blocking fraud
Dimebox have designed a flexible machine learning framework that adapts to constantly changing fraud activity, thereby controlling fraud as it evolves. Even more important, it ensures a seamless payment experience for legitimate customers, without subjecting them to arbitrary rules. The self-learning model detects normal versus abnormal behaviour by continually analysing historical transaction data, meaning that purchases won’t be blocked without a valid reason, maximising revenue and profitability.
Staying ahead of the game
By teaching machines to detect fraud by themselves, far fewer transactions need to be reviewed manually, helping our clients save precious time and resources that can be better deployed in growing their business. What is more, they benefit from an advanced and dynamic system that constantly learns, staying ahead of emerging fraud trends, and ensuring optimum customer satisfaction.