3月 152018
 

fraud detectionIn the medical field, an autopsy is valuable because it helps you understand the cause of death. But, what’s more valuable is identifying the leading indicators of an illness so that you can address it before the Grim Reaper comes knocking. Best in class organizations are taking a similar approach to their fraud detection, shifting from a purely hindsight view to insights and even foresight – getting out in front of the fraud before it happens, revenue is lost, reputation damaged and regulators apply even more pressures.

Proactively detecting fraud isn’t easy though. There is the nature of the challenge itself: Fraud is a behavioral problem and one that is dynamic, complex and often sophisticated. Then, there is the data challenge – lots of it and in many different formats, including structured and unstructured. Next is the analytics. There are many techniques available, and some might be good, and others not. Finally, the technology. There is no shortage of solutions, but they can be expensive and organizations need to beware of ending up with a collection of siloed, single-point solutions that don’t tell the full story.

That said, unless you’re willing to close your business, which is the only surefire way to get to 0% fraud, you’ve got to tackle it.

How to tackle fraud?

For starters, I advise leaders to define their risk appetite and tolerance. What is the level of risk that you – and the organization – can live with? If you can live with 5%, let’s say, then that’s your true North and benchmark to measure against. Once the risk appetite is set, next comes the balancing act of strategic long-term view and tactical short-term needs plus balancing fraud prevention against the customer experience, and more. Then, make sure you have the data, technology, people, processes, governance and analytics in place to continuously measure and refine.

What we are seeing today is that analytics is a key component of moving fraud detection from hindsight to foresight. It starts with dividing risk into three classes. The first is what you know. I have fraud, it’s happening, and I can put business rules in place to detect it. It’s a repeatable pattern that usually responds well to the “if x, then y” formula. The second class is what you do not know.  This is about anomaly detections and can often be found by highlighting things that don’t happen often, but stand out when they do. The third, and most challenging class, is when you don’t even know what you’re looking for. Is it a needle in a haystack? Maybe a rusty nail? This is where AI and ML come in play.

Applying best-in-class tools allows organizations to ingest enormous sets of data, including text, voice, social, structured and unstructured data. Adding best-in-class analytics helps to sort the noise from signals, and advanced analytics including Artificial Intelligence, Machine Learning and Natural Language Processing enable organizations to move faster, by processing in real time, and benefit from iterative learning, where humans help models become smarter and smarter until they can improve themselves every single time. And, of course, the best solutions provide an end-to-end analytics lifecycle from data to analytics to insights.

There’s no question that fraud is complex and challenging, but unless you’re willing to send your business to the morgue – and close your doors forever – you’ve got to tackle it. And, thanks to advances in analytics, we can help stop fraud before it starts.

Find out more at the SAS Global User Forum 2018

Join Constantine Boyadjiev for his “Suspect Behavior Identification through Sentiment Analysis and Communication Surveillance” Breakout Session at SAS Global Forum 2018 April 10 at 3 p.m. in Mile High Ballroom Theater C.

 

 

 

 

Move fraud detection from hindsight to insight to foresight was published on SAS Users.

 Leave a Reply

You may use these HTML tags and attributes: <a href="" title=""> <abbr title=""> <acronym title=""> <b> <blockquote cite=""> <cite> <code> <del datetime=""> <em> <i> <q cite=""> <s> <strike> <strong>

(required)

(required)