The challenge of combating fraud continues to evolve, as fraudsters become more sophisticated and markets digitalise.
Panteha Pedram, Director of Risk Operations and Products at Worldline, writes for Payment Expert on what she feels is the best approach to take when tackling fraud.
The pandemic has exacerbated payments fraud as e-commerce continues to surge and fraudsters are becoming more and more sophisticated. According to Statista, payments fraud will remain a long term, significant, and expanding issue as global fraudulent transactions are set to amount to $38.5 billion a year by 2027.
There are several challenges for merchants to overcome when looking to implement successful fraud prevention tools. With the increasing sophistication of fraudsters, these challenges can vary enormously from one geography or sector to another, and it is incumbent upon merchants to engage an optimal solution provider, or fraud prevention partner, to help them understand these market idiosyncrasies and have the best possible opportunity to address and overcome these variables. It is also crucial that fraud management practices and solutions evolve in line with the methods and tactics used by fraudsters.
Historically, static rule-based solutions (bespoke fraud tooling) were developed before machine-learning (ML) made its entrance onto the market some years ago. While each methodology has its advantages, they have their own individual shortcomings. For example, static rules are seen as being less dynamic, while machine learning is perceived as being less user-friendly due to the potential complexity around its configuration.
To meet the challenges presented by today’s increasingly sophisticated fraudsters, the optimal solution is more and more likely to be one that combines both the traditional static rules with next generation machine learning. By using this so-called ‘hybrid’ approach, businesses can get the best of both worlds, as they are able to retain a certain level of control provided by the more traditional fraud protection solutions, while also being able to harness the power of data through the scientific advances of machine learning models.
A hybrid solution needs to overcome the most common shortcomings of any legacy solution. This includes those presented by the inability of one system to operate seamlessly and successfully in every geography – or having to be expensively recalibrated to address multiple regional complications. Likewise, a successful system would ideally be capable of being tailored simply, in order to fulfill the unique requirements of any market vertical in which the merchant operates now and in the future. It should also be able to minimise higher operational costs as manual reviews become less necessary due to the more advanced, inbuilt machine learning component.
Additionally, the ability to maintain the right balance between approving more orders from legitimate customers and the early identification and weeding out of fraudulent transactions from the outset is both important and challenging. A hybrid solution should help provide this delicate balance of detecting fraud patterns and recognising good customers in order to optimise fraud detection.
An ideal hybrid solution would require a significant level of sophistication to be able to combine the best of the static rules and machine learning model in a single entity. It would encompass the combination of data, knowledge and science, as well as the technical sophistication required to make the rules work. Once up and running, the data science element will become ever more critical to ensure the longer-term success and longevity of the solution. Likewise, a solution should offer multiple models that can be applied and customised ubiquitously to each merchant’s very distinct requirements, i.e. it is not a “one size fits all” solution but rather a “bespoke to the merchant” system.
Critically, any viable solution needs to be future-proof to ensure it stays up to date with technological innovations and is capable of meeting the challenges posed by increasingly sophisticated fraud methodologies.
A hybrid solution is a much more effective deterrent as it benefits from being able to learn and adapt to the ever-changing fraud patterns, as well as being tailored to an individual organisation’s specific requirements.
Any global merchant who wants to increase the revenue impact of their current fraud management systems can benefit from implementing a next generation of hybrid fraud management solution that combines rules and machine learning on top of the most comprehensively available dataset- to increase accuracy and efficiency beyond what rules or advanced machine learning alone could achieve.
If your payment solution provider or fraud prevention partner can also offer dedicated expertise to make such a system tailored to your needs, congratulations, you have found the ideal partner.
