Fraud comes in many forms, with many faces and isn’t always what it seems. Almost every day you read about the latest card skimming scams at gas stations and stolen identity horror stories.
New threats are constantly lurking in the shadows and financial institutions need to keep up. Creditors do not want to mistakenly flag a customer’s valid purchases as fraudulent, but must protect them just the same.
In light of this, influential financial institutions are always looking for more sophisticated approaches to fraud detection.
The hot topic right now is real-time fraud detection in the transaction approval stream.
The Holy Grail for the largest financial institutions in the country is being able to catch fraud before it happens and actually stop the transaction.
For the majority of lenders the emphasis is on what happens after fraud occurs - cutting the fraudster off before it happens again - not stopping it in real-time. The issue with real-time fraud detection at this level is that financial institutions have a limited amount of time to look at transactions in the approval stream. They need to be fast, accurate and not interrupt the customer experience.
While there is limited real-time detection available, the ability to bring in additional intelligence to stop fraud prior to processing payments and reduce the number of false positives is critical.
What if your best customers’ credit cards were declined because their spending habits mirrored fraudulent behavior? One customer doesn’t use their card for months and then charges two luxury cars in New York and a Persian rug in Paris on the same day.
Another cardholder makes three fuel purchases within an hour at the same gas station. These behaviors on both accounts appear suspicious and are likely going to be flagged as fraud based on a general set of rules.
While declining the rug purchase or the third gas purchase may seem like good customer service and solid protection for the issuer, it actually sets up a negative customer experience that may be hard to overcome.
Based on those examples, it is clear that it doesn’t work to apply a standard set of rules to the transaction approval process when trying to determine whether a specific transaction is fraudulent.
Creditors can add value to the customer experience by using behavioral modeling and segmentation, instead of generalized logic. Unique individual behavior tied to a payment type can save a lot of difficult conversations and perhaps a customer.
If it is known that John fills up three vehicles with gas every Friday on his way to the lake those transactions won’t be flagged.
Detecting fraud in the approval stream requires pulling in data and running realtime rules against that data. While this process is no different than detecting fraud on the back end, the key is maintaining the speed of the transaction.
There are a number of things that impact the ability to do this quickly, but the use of customer segmentation and hosted data helps lessen the extra time it takes.
There are a lot of creative ways to apply data that minimize the amount of time you need to gather information and facilitate segmentation intelligently. This process relies on innovation and the sophistication of the tools the bank uses.
As part of payment approval process, financial institutions need the tools to create their own scores and the ability to integrate their own data with additional sources to determine whether a transaction is fraudulent.
Many banks currently decline specific transactions based solely on their own data. Looking at external data sources can help banks see that someone has just declared bankruptcy or reported identity theft.
Also, segmenting people based on a cardholder’s behavior helps identify what is likely a stolen identity or fraud. If a cardholder has a $15K credit limit but never puts more than a few hundred dollars on their card then suddenly a $5K purchase shows up, the bank will know that could be a problem.
Having a tool to define their own logic and apply that logic to the payments stream is a competitive advantage for banks. It gives them more control and more secure transactions. Fraud situations are always adapting and changing.
For financial institutions to stay ahead, they need to have tools that allow them to adapt as quickly (if not more quickly) than the perpetrator of the fraud.
The implication is not that real-time fraud detection in the transaction approval stream is easy, but there are ways to begin implementing and perfecting this idea.
Years ago instant decisioning was considered something impossible to achieve and prescreen-of-one did not exist. Today those are extremely viable strategies that financial institutions rely on to be successful.
Each took effort, but that effort has paid off. They have both evolved and become stronger through a better understanding of what works and what doesn’t.
Technology advancements happen all the time and this is one more example of where the industry is headed.
In the future we will be talking about a different challenge and recalling how we thought catching fraud in the payment transaction approval stream was never going to be possible.
Tom Johnson is vice president of product development at Zoot Enterprises, in Bozeman, Mont. His e-mail is tom.johnson@zootweb.com.
