Don’t Treat Your Customers Like Criminals: Understanding Behavior with Machine Learning

Don’t Treat Your Customers Like Criminals: Understanding Behavior with Machine Learning

Don’t Treat Your Customers Like Criminals: Understanding Behavior with Machine Learning

Guest Bio

Luke Reynolds

Chief product officer at Featurespace, Luke Reynolds has been in the fraud management world for more than 25 years.

More Info

Keeping up with the fraudsters is a constant challenge for fraud and risk operations teams. For more than 20 years, I managed fraud prevention within the banking industry and saw first-hand that these criminals are constantly evolving their attack methods. The severity of the attacks proves how essential it is to use the latest technology systems to tackle fraudsters and prevent against blocking genuine customers' activity.

Banks, card issuers and merchants that meet these challenges by outsmarting risk gain a significant competitive advantage.

What are the fraud challenges facing organizations today?

During my years in fraud prevention management, I was always impressed by the hard work and dedication of fraud teams in seeking new ways to tackle the evolving methods criminals will use. I also experienced their frustrations with outdated consortium data and heavyweight fraud management systems, which lacked the flexibility to build the rules and models they wanted. With these outdated methods, too much fraud gets through and too many genuine customers get blocked.

Transactional card fraud continues to be a target for these legacy systems, hitting banks, card issuers, merchants and their customers where it hurts most: revenue, reputation and resources. The payments industry is also seeing a rise in card-not-present (CNP) fraud — especially since the introduction of chip-enabled cards in the U.S. — as fraudsters attack vulnerable new channels where legacy systems struggle to spot changing fraud types. recently reported that CNP fraud rose 12 percent during the holiday season alone at the end of 2016.

Faced with these challenges, how can businesses reduce fraud risk while allowing genuine customers to transact?

The answer: real-time, adaptive machine learning

A significant change I have seen across the industry is financial organizations adopting the latest machine learning technologies that use adaptive behavioral analytics to gain unique, informed foresight and control over real-time, individual customer risk.

Legacy systems that rely on outdated, static data sets to understand customers are exposing banks to unnecessary fraud risk. Both customer behavior and fraudsters’ attack methods are constantly changing and to truly understand behavior and spot anomalies, organizations need an adaptive approach. This means understanding each individual customer’s behavior in real-time — aggregating their behavioral profile across both monetary and non-monetary events.

Understanding individual behavior in real time

People, like businesses, change over time — and their income, lifestyle and spending habits evolve.

By detecting deviations across all customer activity, a machine learning behavioral analytics approach constantly adapts to changes throughout each customer’s life cycle to spot new fraud attacks as they occur. At the same time, this deeper customer understanding means that genuine customers are easily recognized by the fraud system and they can transact without being blocked.

Adaptive behavioral analytics works by building a normal pattern of behavior in real-time for each individual customer and their peer group — and then spotting the exact moment when this behavior changes. All this takes place within a self-learning fraud prevention system, which automatically updates itself as new behavior types are identified.

What makes this unique is the ability to distinguish between natural human volatility and true anomalies, which means new types of fraud are detected automatically while simultaneously and dramatically reducing the false decline rate for genuine customers.

The good news is that this is done using data already captured by financial institutions. Practically, this means organizations get a sophisticated understanding of each individual customer within an automated risk-management system. When an individual’s behavior changes, the system can alert the fraud team, and a clear reason code will help fraud teams accurately and quickly interpret the fraud attack, so that financial institutions can take further action.

The benefits of real time

Card issuers, banks and merchants are already seeing significant benefits from adaptive behavioral analytics, including:

  • Quick, accurate fraud decisioning and acceptance to protect customers and increase revenue
  • Average of 70-percent reductions in volume of genuine transactions declined — and even greater reductions in newer fraud target areas, such as CNP and contactless — so fraud teams can focus on the most urgent fraud alerts
  • Clear, real-time understanding of the customer base for fraud and customer experience management

Why now?

Machine learning and adaptive behavioral analytics are enabling businesses to turn advanced customer understanding into smart business risk decisions. After all, there's no need to risk your customer protection, your revenue or your reputation.

It's possible to have all three protected — and the power of adaptive machine learning is making it possible.

The statements and opinions of the writer do not necessarily reflect those of TSYS.

Other Articles by Luke

Luke Reynolds

Chief product officer at Featurespace, Luke Reynolds has been in the fraud management world for more than 25 years. Prior to joining Featurespace, he worked directly in fraud and security operations in the banking and payments industry for more than 20 years, including roles as Callcredit's commercial director of fraud and ID and Lloyds Banking Group (head of retail audit, head of fraud and head of group security and investigation). Luke also worked in fraud management at the UK Card Association and NatWest. He is passionate about helping organizations implement machine learning technology to improve their fraud and customer management.

Share this story via email or social networks