Scott Zoldi, Vice President of Analytic Science at FICO recently returned from Payments 2015 where he was invited to speak on advances in behavior analytics—specifically fraud analytics that differentiate unusual customer behavior from fraud, including collaborative profiling that anticipates new previously unseen behavior. Here are his observations and recommendations:
This analytic innovation is just one of 92 patent applications pending in our patent pipeline. Our patent portfolio is very important to FICO, and in fact we currently have 170 US and foreign patents acquired over the past 25 years.
Naturally, we’re quite proud of this accomplishment. But our goal isn’t simply to do cool research or get chest-bumping bragging rights. It’s not innovation for innovation’s sake. Rather, our goal is to create consequential products that benefit our clients, their customers and even society at large.
Let’s take fraud analytics as an example, since that’s my area of focus. The dynamic nature of fraud demands that anti-fraud technologies continuously advance in order to keep up. That’s what drives my fraud analytics team to constantly break new ground and hone our analytic approaches, and indeed FICO currently holds 89 patents in fraud, with an additional 40 patents pending.
These predictive analytic breakthroughs have had a direct and dramatic impact on fraud. Losses on US credit cards, as a percentage of credit card sales, have declined by 70% since we introduced real-time anti-fraud analytics in 1992, as shown in this infographic. While this drop is due to multiple factors—including better security features built directly into payment plastic and more vigilant consumer practices—predictive analytics played a starring role by instantaneously identifying suspicious transactions at point of sale.
Since 1992, there have been tremendous changes in how we shop, bank, and pay for goods and services, as well as in the sophistication of fraud attacks and recent cyber breaches. And fraud analytics have had to advance just as rapidly to keep pace. For instance, in 1995, we introduced modeling that extended the same real-time protection of credit cards to debit cards; in 1999, we expanded that protection to card-not-present transactions as e-commerce hit its heyday; in 2008, we filed patents on our adaptive analytics technology that allows in-production models to learn and adapt to real-time fraud events; and in 2012, we introduced the collaborative profiling analytics I mentioned at the start of this post. And we’re by no means done.
Looking beyond fraud, we face similar conditions—dynamic change and big challenges—that call for ongoing analytic innovation. One area that I am focused on with urgency is cyber analytics: streaming models for detecting cyber security attacks that may otherwise go undetected by today’s conventional technologies. As I’ve blogged previously, cyber breach and fraud detection are two sides of the same problem.
Of course, FICO is proactively looking for new and better approaches to tackle all sorts of other challenges—from finding new ways to turn Big Data into better customer experiences, to democratizing access to predictive analytics so that more of us can tap into its value without deep-pocket investment.
170 patents and counting. I’m excited to see what innovations the next 25 years will bring!
Source: FICO – The author is Scott Zoldi, Vice President of Analytic Science at FICO