Homegrown Innovation: Best Practices for Building An Enterprise Fraud Management System

Hui Wang, Sr. Director, Global Risk Sciences, PayPal

Please see this post on our new PayPal Stories site.

 

Helping to protect our customers is one of our top priorities. And in rare instances, when fraud occurs, we have an industry leading team of fraud management experts in place. To help protect against fraudulent transactions, we have Purchase and Seller protections for eligible purchases.

 

Our loss rate, our expenses due to fraudulent activity compared to our revenue as a whole, is one of the lowest in the industry, at 0.32% or approximately one third of 1%. A key reason why we’ve been so successful is because of the holistic approach we take when it comes to security and fraud prevention. At the core of these efforts is our Enterprise Fraud Management System (EFM).

 

In 2009, our global risk management organization embarked on a journey to build a complete Enterprise Fraud Management System in-house. Now complete, this EFM system allows us to work with vast amounts of rich data to help keep fraud at some of the lowest levels and our customers transactions secure.

 

A lot went into building the best EFM system to do so. There were some key considerations we kept top-of-mind to keep our customers secure, and based on our experiences, we wanted to share some key learnings.

 

Keep it Simple. With isolation comes innovation. Isolate problems and make sure to solve one at a time. When building our systems, we made sure to not get caught up focusing on the perfect final product. We started with one area, and made it work end-to-end before working on other challenges. Building systems one stack at a time, allows you to easily identify problems, and nip them in the bud.

 

Test and Learn. At PayPal, we have to make our risk decisions in milliseconds. To create a system that accurate, it takes a lot of testing and learning. We are constantly layering on enormous amounts of data, to create the best system. We detect suspicious behavior using thousands of characteristics, but are still able to correlate whole events in seconds, as compared to hours for the rest of industry.

 

Strike a Balance. The best enterprise fraud management solution balances analytical and engineering knowledge. Cross-functional teams are imperative. Our teams are comprised of data scientists, engineers, project managers and business minds. When building your system, you cannot sacrifice systems or data scientists. It takes everyone to find the best solution.

 

Don’t Underestimate Intelligence. PayPal can detect suspicious behavior using thousands of characteristics – device type, dollar amount and much more. But we don’t just rely on the data – we have developed our own unique ways of using a human detective-like approach to help identify even the most hidden nuances. Combining advanced machine learning and data mining with a human oversight and approach, will be more successful than just trying to tackle the data alone.

 

Never Stay Still. It is important to keep an open mind for new generations. Always be listening, and be prepared for the next challenge. We like to call it futureproofing. PayPal has a team dedicated to developing innovative risk techniques for technologies that are three to five years from becoming mainstream – for example we developed algorithms for mobile approximately five years ago, and are now better positioned to help protect consumers from fraud attempts on mobile platforms.

 

At PayPal, we believe every transaction tells a story – which is unique from the way most companies evaluate transactions. These stories guide us on our journey to the next generation of enterprise fraud management. We have already started looking ahead, and make it a best practice to stay one step ahead – instead of constantly trying to catch up.

 

With our systems, we are able to grow and evolve, changing risk rules and models to better protect our customers now and in the future.

Original author: PayPal-Forward