News, Education and Events Decoding Digital Payments & Fraud

News, Education and Events Decoding Digital Payments & Fraud

Considerations for Choosing a Machine Learning Platform for Fraud

Considerations for Choosing a Machine Learning Platform for Fraud

Sponsored Content

[Editor’s Note: November is Machine Learning Month on CardNotPresent.com. This is the fourth article in a series from our sponsor exploring how, against increasingly sophisticated attacks, artificial intelligence and machine learning are being applied to online fraud prevention. Check back here throughout the month for more content that sheds light on the newest technology in fraud prevention and answers questions you might have on how it can impact your antifraud efforts.]

Large organizations are turning to machine learning because legacy fraud solutions have reached the limit of what they can imagine by using rules alone. In a digital economy with its proliferation of channels and payment types, customers are crossing channels to make payments, demanding immediacy in transactions and calling for real-time innovations in payments.

Organizations are trying to satisfy consumers while also mitigating growing risks. To defraud businesses, today’s criminals are leveraging all the same tools that banks and large merchants have: the same market complexity, the same data abundance and the same technology.

The result of all this complexity, speed and digital innovation is explosive growth in fraud. Online retail fraud doubled last year. Juniper Research projects it to double once again by the year 2020, to $25 billion. That’s the same year when card fraud losses globally will top $30 billion, according to the Nilson Report. Meanwhile, growing trends in real-time payments are only reducing the time that bankers and merchants have to make tough decisions.

In this fragmented and fast payments space, banks and merchants have concluded they need more sophisticated tools in order to detect fraud at scale and at speed. To answer this call for better fraud detection, a growing space of machine-learning vendors are entering the scene. How can an organization begin its search for the right partner?

Here are three key considerations to take in this search:

1. Refocus on the customer with a complete view

Because banks today are product-centric rather than customer-centric, they make decisions in silos and are vulnerable to attacks on the same account across multiple channels. How can a bank know whether a customer defaulting on a credit-card bill is a risky customer for a home mortgage?

Machine learning can break down data silos by performing omnichannel aggregation and omnidata integration. The result is a 360-degree view of transactions right as they happen.

To build up a complete customer view at speed and scale, a machine learning system needs to be data agnostic, to be conceived for the purpose of extracting and loading all kinds of data, whether they are within the bank’s own system or are augmented from external sources.

2. React to fraud faster

With digital transformation comes the expectation of immediacy. Customers want decisions made at the speed of transactions, which is driving immediacy trends in payments, such as Amazon one-click ordering.

To manage the risk associated with immediate transactions, banks and merchants need a system that can accelerate the machine learning process and lead fraud analysts and data scientists to fraud drivers quicker. The benefits of speeding up the machine learning process are explored more deeply in the report, Improving Fraud Detection by Speeding Up Machine Learning.

To enable organizations to react to fraud faster, a system must be also architected for the rapid deployment of new models. Furthermore, as analysts review transactions and label them as fraud or not, the platform should integrate with those decisions and automatically learn from them to become better at recognizing future patterns.

3. Pursue explainability

As more and more banks are turning to machine learning to make good decisions, they’re realizing the need for explanations too. Transparency and interpretability in a machine learning system have two important benefits.

First, a system with interpretable reasoning lets organizations audit the machine and provide trails of explanations for compliance. Second, better explanations mean banks and merchants can drive greater engagement and deliver greater customer experience.

Today, explainability exists in the form of whitebox processing, which adds human-readability to the underlying machine logic and communicates factors behind its decisions to the human analyst. Organizations should consider platforms that can perform whitebox processing and provide transparency in decision-making.

To explore more considerations in greater depth, download the report, Choosing a Machine Learning Platform for Risk.

How to choose

The fact is this: poor fraud detection hurts.

First, there’s the financial impact of denying a legitimate customer. Banks that add friction to customer experiences will find those customers running toward a competitor that has experience “figured out.” And for merchants, a Business Insider report found that for every dollar a business prevents in fraud, it loses even more – $1.32 – to false declines.

Second, there are the cost impacts that chargebacks have on merchants and that massive fraud attacks have on banks.

Third, there’s the impact on the speed and efficiency of your operations. Large organizations need to review and process orders as quickly as possible. Poor fraud detection platforms create bottlenecks that delay order fulfillment, frustrate customers, and thwart promotions and peak periods.

The right machine learning partner can help you mitigate all these risks. Read this ebook to discover all the considerations to take in choosing the right machine learning platform.

Read The Next CNP Report Article