Over the past two decades, commerce’s expansion to digital channels has created an entirely new ecosystem of payments information and with each passing moment, the amount of that information increases:
- In the U.S., there were 21.5 billion ACH transactions in 2017, an increase of 5.7 percent from the previous year and in that same time, same-day ACH transactions increased 478 percent;
- The Federal Reserve’s latest Payments Study shows that the total number of card payments increased at an annual rate of 7.4 percent year-over-year in 2016; and
- Economists at the World Bank estimate that in 2016, $574 billion was sent from the U.S. to other countries.
This data possesses virtually endless business potential, but the challenge is accessing, incorporating and extracting value in this transactional information in real time, which is where machine learning comes into play. A form of artificial intelligence (characterized by a computer’s ability to make intelligent decisions), machine learning is the process of the computer learning from the data it receives to improve its performance on a given task. The more data and historical context that is provided, the greater the computer’s accuracy. In business, this has demonstrable applications across all areas: marketing efforts become stronger through more precise and timely offers, while also revealing customers’ varying propensity to respond positively; operationally, training, inventory management and information sharing becomes increasingly automated; and in risk mitigation, by enhancing the detecting and reporting of suspicious behaviors when processing transactions.
When it comes to payments, risk mitigation is particularly challenging; whereas honest consumers operate within the framework of your business, fraudsters seek to exploit your system anyway they can. And with more targets for skilled cybercriminals to attack, combating this threat requires the ability to both analyze consumer behaviors at the individual level and adapt to fraud in real time – something that machine learning facilitates. And the technical incompatibilities that once prohibited business with mainframes from taking advantage of machine learning are now a thing of the past.
Serving as a gateway for mainframes, GT Software’s technology allows organizations to interact with new API’s, enabling machine learning technology to analyze internal data and use it to prevent illegitimate transactions, while allowing honest ones through. Leveraging a powerful and versatile API, GT Software modernizes mainframes in a fraction of the time and cost typically required, providing a gateway through which clients can onboard the latest technologies – and in the fight against payments fraud, there is no better technology than Featurespace’s ARIC™ platform.
The Featurespace ARIC™ platform captures all relevant information for each customer’s transactions and leverages Featurespace’s adaptive behavioral analytics to build individual profiles, which are constructed using the unique behaviors of each consumer. As it collects data, the platform monitors for anomalies in transaction patterns, which are weighted and used to help the system make accurate predictions in real-time as to whether the transaction should occur or be prevented. The customer profiles that are incorporated into the decisioning model become increasingly robust with each transaction, allowing the platform to constantly evolve its ability to detect any irregularities. This is important because whereas traditional fraud prevention tools can detect known fraud, Featurespace’s software learns from new fraud attacks and adjusts accordingly to ensure organizations are keeping pace with cybercriminals.
At the crux of any effective machine learning model is its accuracy. An inability to properly discern whether activity is truly illegitimate results in a good transaction being halted. Anyone who has experienced this can tell you how frustratingly inconvenient it is and for both consumer and businesses, friction at the point-of-sale is one of primary reasons a sale and customer is lost. Alternatively, if the model doesn’t detect a fraudulent transaction, it’s needlessly increasing the organization’s risk exposure and jeopardizing customer relationships. Simply put: either scenario is bad for business.
Cybercriminals have adjusted their attacks in the digital age and it’s time for companies to respond accordingly. Data science has never been this advanced and its accessibility through streamlined modernization removes all barriers that previously deterred organizations with legacy systems. Machine learning systems that capture and leverage large volumes of data in real time have leveled the playing field, allowing businesses to increase revenue and outsmart risk.
– Dave Excell, Founder and CTO of Featurespace