There has been a lot of noise in the fraud marketplace from new market entrants touting the power of machine learning (ML), artificial intelligence (AI) and “new data sources” to fight fraud. Regarding ML and AI, FICO Chief Analytics Officer Scott Zoldi wrote in his post, “Artificial Intelligence: Find It Right in Your Own Backyard”:
“The FICO® Falcon Fraud Management solution is the industry leader, protecting two thirds of the world’s payment card transactions against fraud. In far less than the blink of an eye (40-60 milliseconds, to be precise) Falcon scores each authorization that a merchant submits for approval.
“Here, Falcon uses adaptive analytics, a type of machine learning in which self-learning models work with the Falcon consortium models to improve the prediction of future fraudulent behavior based on fraud attacks in production. These models score transactions based on recent known fraud and non-fraud transaction data, dramatically improving the models’ sensitivity to changing fraud patterns, and keeping up with the fraudsters.”
Scott has written extensively about FICO’s use of AI and ML. My post today is about the underlying data we use to build models using machine learning analytics, the FICO Falcon Fraud Consortium. Let’s dive in.
The Predictive Power of Consortium Data
In thinking about this topic, I am reminded of one of my favorite bon mots from graduate school statistics classes, the sentiments of which have served me well over my career: “In God we trust, all others bring data.”
The point here is that data is what teaches you — not your assumptions or biases, but data. And the more data, the better. That was true in the past, and it’s even more true today, as computing power has completely reframed “the art of the possible” when dealing with large volumes of data.
In terms of detecting fraudulent card transactions, there’s only one data source that offers the most predictive power, hands down: other card transactions, and lots of them. Specifically, the transaction data from two-thirds of all the cards in the entire world, which pours into the FICO Falcon Fraud Consortium database, every moment of every day.
The breadth of this data, and its historic depth, provide a reference data set that is simply unmatched. More than 9,000 institutions and card providers around the world participate in the FICO Falcon Fraud Consortium, contributing data from all of their payment transactions, depersonalized to remove personal and sensitive information.
The FICO Falcon Fraud Consortium provides the data foundation of the FICO Falcon Platform, and has continuously improved the accuracy of fraud detection since Falcon launched 25 years ago. What’s more, the consortium provides the additional benefit of alerting all members of fraud threats and patterns seen by an issuer or institution. This “heads up” gives the entire consortium enhanced protection from threats that many member have yet to see.
Collaborative Machine Learning
The FICO Falcon Fraud Fraud Consortium is the ultimate example of collaborative machine learning. A single data lens into any given issue, including fraud detection, is likely to be myopic; collaboratively solving this problem through data sharing is an inherently stronger approach.
Think about your drive home. A million years ago, in the B.I. (Before iPhone) era, you probably took the same route, changing it up if you ran into bad traffic. Now we have technology like Waze, “the world’s largest community based traffic and navigation app” bought by Google in 2013 to enhance its mapping business.
Waze users contribute real-time traffic information to the app, giving other users up-to-the-minute insight into traffic patterns. Where I live, in San Diego, Waze can make the difference between a 20-minute cruise and a 90-minute slog during rush hour. And in using Waze, I’ve even stumbled on some neighborhoods I didn’t know existed before and remarked to myself, “So that is how I get up to those houses on the hills!”
Getting back to payments fraud, the same principle applies. Are fraudulent transactions best identified by a handful of data points, or billions? Logic and experience say that the power of global data sharing and community prevails.
Simply put, 9,000 heads are better than 1. And only FICO is bringing more online collaboration to market with the use of collaborative profiles and by leveraging our FICO Analytic Cloud. Stay tuned!
Follow me on Twitter @fraudbird.