One of our latest innovations in fraud and cybersecurity addresses a fundamental issue affecting predictive analytics: Data doesn’t sit still. What do I mean by that? As any data scientist can tell you, a model development project begins with the lengthy process of collecting, identifying, cleansing and normalizing data. This is often the longest part of the process of building a model. There are many checks of data integrity to ensure proper data quality. The paradox is that the data we’re spending so much time working with isn’t necessarily the data we really care about. The data we really care about is the data that the model will analyse in the future — which will be different than the data we’re studying to build the model. Analytics methodologies include data governance to ensure that a model developed on today’s data will make good predictions based on future data. For instance,... [Read More]
To some degree, we’re living in a world where we are cursed with our own success. Financial institutions have seen tremendous benefits from analytics, and as a result, they are using predictive models on an increasingly broader scale, to measure capital reserve requirements and manage complex customer decisions. But as my rap doppelganger would say: “More Models, More Problems.” The greater complexity and number of predictive models in use makes it even more difficult to track and manage model performance, not to mention comply with regulatory requirements. Since the financial crisis, banking regulators have increased their scrutiny of how institutions use predictive analytics. These days, regulators are not only concerned with the safety and soundness of the analytics themselves, in terms of how the models are built and whether they are still validating. Regulators are also focused on the impact of the decisions—that is, who gets a particular decision and... [Read More]
For several years, I have been actively “fighting the good fight” in the area of cyber security. Beyond my anti-fraud work here at FICO, I also participate in various industry efforts focused on preventing cyber crime, most recently joining the board of directors for the Cyber Center for Excellence. Cyber security touches our lives daily, whether it's protecting our national infrastructure, securing payment systems, or installing virus protection for personal computers and devices.
The recent data breach at Target caused many to rally around the adoption of EMV payment cards. But while it’s a step in the right direction, it would not have prevented the loss of data estimated to affect more than 70 million customers. The Target breach, however, does point to the need to monitor the computer networks for malware designed to steal PII (personally identifiable information). These are costly problems for both consumers, who need to stay vigilant of any PII misuse, and to financial institutions, where an estimated $200 million will be spent just to replace...
Back in school, cramming for an exam may have been acceptable (although stressful). But it’s never the best option for lenders preparing for their next regulatory “exam” on model risk management. Fortunately, there’s no need to cram for an audit if you adopt good model management practices from the get-go.
Your preparation must begin well ahead of an audit, a point made clear by the 2011 Supervisory Guidance on Model Risk Management issued by the US Office of the Comptroller of Currency (OCC) and similar guidance in the Federal Reserve’s SR 11-7. These guidelines require sound and robust processes for model development, validation, implementation, use and governance. This has always been good policy, but in the regulatory climate following the post-financial meltdown, it becomes even more critical.
How can you make sure you’re prepared? Here are a few tips:
Understand the scoring models you use, both those developed in-house as well as those from third-party vendors. The guidelines state that bankers must demonstrate a clear understanding of the...
As frequent readers know, I'm a strong advocate for using adaptive analytics in credit card fraud detection. Some may wonder: Could the same analytic technology also boost fraud protection for debit cards? The answer: Absolutely!
First, a quick primer on adaptive analytics. These models continually "adapt" traditional neural network fraud models in response to real-time fraud tactics, some of which may have not been present during model training. This ability to learn from in-production fraud trends helps financial institutions recognize newer schemes and consequently catch more fraud.
We recently tested the value of using adaptive analytics on a FICO debit consortium model, specifically our recently released FICO® Falcon® Fraud Manager UK-Ireland Debit 12.0 model. Adaptive analytics detects 6% more fraud accounts at a 10:1 account false positive ratio (AFPR). By keeping the detection rate fixed, adaptive analytics reduces the model’s false positive ratio by more than 10%, up to a 40% account detection rate (ADR), as shown in the...
My analytics team has been working on some exciting new technologies that improve fraud detection while reducing declines of legitimate transactions. One such research innovation boosts performance of traditional fraud models, which use a single customer’s past purchases to determine whether a current transaction is consistent with that customer’s historical behavior patterns. The new analytic technology goes further by examining patterns of transactional behavior across millions of customers to continuously update “behavior archetypes” for each customer in real time. As such, the analytics provide a more complete and textured view of customer behavior.
Our new analytic technology leverages a large-scale consortium-wide data set of historical behavior and, with some modern Bayesian methods, computes data-driven archetypes based on patterns across many customers. Think of these archetypes as the “atoms” in the periodic table of customer behaviors. Real customers are “compounds,” composed of a flexible mixture of various archetypes. At a...
Individual cardholders are creatures of habit. Cardholders have "favorites"—or recurrences—over a wide variety of entities in their transaction streams. These entities might include favorite ATMs close to work or home, favorite gas stations along a daily commute, preferred grocery stores, and preferred online stores for shopping.
To improve fraud management, we’ve been developing analytics that identify these cardholder favorites. This new analytic technology helps distinguish between “in-pattern” normal customer spending and “out-of-pattern” suspicious transaction activity. This enables faster fraud detection at much lower false positive rates (declines on legitimate transactions).
How does it work? An advanced analytic algorithm maintains favorite lists within the card transaction profiles. These "Behavior Sorted Lists" are updated with each transaction so that the patterns of favorites evolve over time. The more frequent entries appear with greater recurrence and are ranked at the top of the list, while less frequent entries...
Financial
institutions have had a difficult time adapting to the latest regulatory guidance regarding
model validation and management. But making the right improvements can also translate into better analytic
performance and risk management.
To
both comply and compete, it's critical to build an organizational policy for
comprehensive model and credit policy management. This framework should include
the following tried-and-true practices:
Have
clearly stated credit policies; review these regularly. We recommend reviewing these every six months since they have a
direct impact on your bottom line. In the US, the Fed and OCC require a review
of policies at least annually.
Prepare a
suitable data sample. Regulators
require you demonstrate your model validation sampling techniques are complete,
responsible and relevant, since incorrect or inaccurate sampling can impact
model performance.
Ensure
model segmentation transparency. In
general, you’ll need to clearly document how you segmented subpopulations and
how this supports business...
Those of you who follow this blog know that I regularly discuss analytic innovations that boost fraud detection. One such innovation is adaptive analytics, so-called because these models continually "adapt" traditional neural network fraud models in response to real-time fraud tactics that were not present at the time of model training. This helps financial institutions combat newer fraud schemes that arise between fraud model developments.
FICO research regularly demonstrates that adaptive models provide performance improvements and extend the useful lifetime of static neural network fraud models. Results from our latest research study show that these results hold true worldwide.
More specifically, the study showed that banking institutions worldwide can significantly improve fraud detection rates and reduce transaction false positives. We analyzed consortium data supplied by FICO® Falcon® International Credit clients, comparing the addition of adaptive models to using only neural network models.
Figure 1 below shows how adding adaptive models...
How are companies using Big Data analytics to understand and collaborate with today’s connected consumer? Join us at FICO World 2013 to discover answers from experts and network with your banking peers. Registration is now open for the conference, which will be held April 30-May 3 in Miami. FICO World has become the leading international conference on analytic strategies. This year, we have a packed agenda with more than 80 presentations from 70 banks and retailers on fraud management, analytic innovation, risk management, collections and improving the customer experience (view a list of sessions). Attendees will also hear from keynote presenter Kenneth Cukier, data editor for The Economist and co-author of a new book on Big Data. Learn more or register for FICO World 2013.
