This is the 1st part of the Nethone Guide to Fraudproof Payments in which I delve deeper into what it means to use Machine Learning to increase the revenue by fighting fraudsters and recognizing good customers.
Nowadays, fraud prevention is all about Machine Learning (ML). It is a corollary of the rapid growth of computing power and popularization of ML techniques in many fields of science and business. Cars are driven by the “taught machines” and our faces are recognised by models that learned from millions of pictures. It goes without saying that exchanging a set of static rules with an adaptive model being an equivalent of thousands of static rules, brings the sharpshooter’s eye to the fraud tool.
Saying all that, one can still often hear that “ML is not the Silver Bullet”. Well indeed, it is not, but by no means because of any shortcomings of the technique itself. ML gives the power to crunch insurmountable amount of data in real-time, a feature no human would be able to achieve. It is a great tool but an agnostic one, it depends on what data you feed it with, how much data you have and what context it is assigned.
Furthermore, ML is not a single object nor a single technique. It is not one algorithm nor a set of rules that you apply to the question at hand to get an answer. ML is an art in itself, one that bares ripest fruits to those who study it and are ready to tinker with it. This is why Nethone brings the best specialist in the field who really know when to apply a random forest algorithm or a neural network to the problem. Both techniques are proven to work miracles with data but first you need to know to what kind of data each responds better.
Also for this reason, Nethone builds custom models for each partner… even more, it is not that a model is trained for each partner but each model is created for each partner’s case – chargeback prevention, fighting friendly fraud, marketing optimization and any other business relevant needs. There is no limit to Nethone capabilities when it comes to squeezing an insight out of available data. This approach truly utilizes the possibilities brought by ML, allowing for high accuracy, giving clear cut results based on hard data.
We, at Nethone, also appreciate the fact that building a model that performs great is not enough. The world is changing and the web world is changing even faster, so it is in constant flux. The models need to be retrained online, so they can evolve with the partner’s traffic to always stay on the edge and keep the revenue growth at maximum pace.
ML is a piece, an important one, of the puzzle which, as we prove each day, leads to a precise solution closer in analogy to a laser beam rather than a silver bullet. Nethone brings together Machine Learning, authorial research into data harvesting, expertise in the field of fraud prevention and merchant’s business insight. Putting all these pieces together is what gives our partners fraud/chargeback levels far below 1% and no fear of dropping a sincere customer.
Should you have any questions or comments, don’t hesitate to contact me. Those will be published and answered below, so please stay tuned.
Q: You say it’s a Laser Beam solution but still state that your partner’s experience chargebacks/fraud?
A: The level below 1% very often reflects human mistakes, performed single (not repeated) operations on freshly stolen cards and other that in no way differ from healthy operations and are in no way discernible.
At Nethone, Aleksander is responsible for business and product development, workflow management and ensuring comprehensive operational excellence at the company.