Banking Analytics Blog

Infographic: The Analytics Big Bang

2013-06-19T06:00:00-07:00

A new FICO infographic “The Analytics Big Bang” traces the evolution of predictive analytics since the dawn of the computer age. It cites compelling evidence that the analytics industry is at an inflection point, and offers a glimpse of what comes next.

Note: click on the infographic above to enlarge it, or view a larger version here.

So, where are we headed next? Put state-of-the-art analytics and the means to drive real-time decisions from it into a cloud computing infrastructure, and the power of analytics becomes available on-demand to organizations of all shapes and sizes. Suddenly, every organization can take advantage of profound efficiencies and opportunities for collaboration that transform what we can accomplish, how fast and at what cost.

Andrew Jennings, FICO’s chief analytics officer and frequent blogger, summed it up: “We've now reached a tipping point where the convergence of Big Data, cloud computing and analytic technology is leading to massive innovation and market disruption. We foresee predictive analytics being used to solve previously unsolvable problems, and bringing enormous value to businesses, governments and people.”

Fraud Turns Low-Tech as Criminals Go Back to Basics

2013-06-17T04:08:15-07:00

In Skyfall, James Bond put aside the invisible cars and exploding pens from previous films, and set about defeating the villain with old-fashioned guns and handmade bombs. This gadget-light approach also characterizes the real-world villains who are perpetrating credit card fraud.

At a recent meeting of the FICO Fraud Forum — our fraud consultants, modellers and product specialists from around the world — it was observed that criminals in the U.S. and Europe that have been thwarted by the sophistication of anti-fraud technology are going back to some very basic techniques. As my colleague Martin Warwick noted, there is more trickster and confidence fraud today, as criminals look to get good information out of unsuspecting cardholders.

We have seen a rise in phishing and vishing in both regions. Social engineering is seeing a sharp increase in Europe and the US too — in other words, getting the cardholder’s data from them directly, or even from a bank’s offshore service providers, who may not have the same level of fraud prevention measures in place.

One of the relatively prevalent schemes in recent times demonstrates this low-tech approach. You get a phone call from a person who says they are a representative of your bank, and that there has been suspicious activity on your card. The person then asks some “security questions” before divulging any specific account information. Some unsuspecting customers provide answers in good faith; others unwilling to do so are encouraged to phone their bank back on a telephone number the customer trusts. You hang up — but the other person doesn’t, they just wait for you to pick up again, they play a dial tone sound, and when after you dial you are “connected” to your bank – only it’s the same call. The person (or a substitute) takes the call, and asks for your card and personal security to identify you. They can then use this information to access the customer’s true account records, to fabricate cards for use in non-chip-and-PIN environments, and even to make large payments. In the worst, and most insidious cases, fraudsters specifically target elderly and wealthy customers and even make arrangements to send an accomplice to the home address to collect the card and the PIN for “forensic investigation” purposes whilst promising that a new card and PIN is on the way — a surprisingly simple and effective ruse to gain access to the customer’s money.

As I noted in a recent post, authentication is the new currency — if the thieves have enough information on you to pass a bank’s authentication measures, they can raid your accounts. Banks have a duty to let their customers know that they themselves are quite often the weakest link in the anti-fraud chain.

Does Big Data Help or Hinder Customer Centricity?

2013-06-13T10:00:00-07:00

The promise of Big Data looms large as banking institutions worldwide launch major customer centricity initiatives. Today we have the means to capture and analyze much bigger quantities of data than ever before, and to make meaningful connections between different types of it. We can analyze data in-stream for real-time decisions. We can distribute analytic tasks in a massively parallel manner across many processor nodes, then algorithmically assemble their outputs into a single result.

But is any of that helpful for achieving customer centricity?

It’s most helpful when we can systematically extract the most valuable analytic insights—causal relationships—from Big Data. These insights enable us to understand individual customer behavior and sensitivities, anticipate needs, and predict likely responses to offers and treatments. In some situations, we must find and act on such insights as data is streaming in. In others, we can use out-of-stream methods to dive deeply for them.

Big Data computing infrastructures are making it practical to employ automated machine learning algorithms for this purpose—but human expert oversight is essential to ensure results make business sense and are useful in operations. And, ultimately, whether any of these insights impact customer centricity depends on how quickly we can pump them into operations so that they inform every decision and every customer interaction.

These essentials for turning Big Data into an enabler for customer centricity are fundamental to what I call “next-generation analytic learning.” At its core, next-generation learning elevates test-and-learn methods to a new level of efficacy. It’s a systematic, highly efficient way of continuously advancing what we know about our customers and improving how we use those insights to interact with them.

Over the next few weeks, I’ll be discussing the imperatives of next-generation analytic learning here, as well as on the FICO Labs Blog. So stay tuned…or I check out my recent Insights white paper on the topic: "When Is Big Data the Way to Customer Centricity?" (registration required). These imperatives can be practiced by every company, irrespective of size or analytic sophistication. In fact, my paper talks about how to get the most value from what you have right now, even if your data has “holes” and other imperfections, and your customer treatment strategies have been inconsistently applied.

Know Customer Favorites to Fight Fraud

2013-06-11T12:44:00-07:00

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 fall away and are replaced with new entries.

The graphic above compares performance of FICO® Falcon® Fraud Manager 6 models with and without the Behavior Sorted List technology. Falcon 6 models with Behavior Sorted Lists outperform the legacy Falcon models without it, looking at both fraud detection and false positive reduction. I’ve highlighted the relative improvement in fraud detection (“REL”) at a couple account false positive ratios (“AFPR,” or the number of accounts the model identified incorrectly as fraudulent for each actual fraud account found). Because of this boost in performance, Behavior Sorted Lists are being incorporated into all Falcon 6 consortium models released this year.

The Young and the Cardless

2013-06-10T10:27:02-07:00

My admittedly kitschy title highlights a new trend that we’ve observed after analyzing credit behaviors across different age groups: more young consumers are forgoing the use of credit cards.

First, let’s look at a breakdown of outstanding debt for consumers in the 18-29 age group between October 2007 and October 2012. Except for student loan debt, these consumers have reduced their outstanding debt across all categories.

While the growth in student loan debt garners a lot of attention, the fairly dramatic reduction in credit card debt is also enormously noteworthy. Lower credit card debt is not only associated with consumers carrying lower balances on their cards, but more consumers not having any credit cards in general.

As the chart above illustrates, over time there’s been an increasing number of consumers within each age group that no longer have a credit card. This change is most dramatic among younger consumers.

This trend can be attributed to two main influences. First, the CARD Act has put more requirements on issuers before a card can be issued. In order to open a credit card account, for example, consumers under 21 need to show an ability to pay off the debt or have a co-signer on the card.

Secondly, the impact of the recession seems to have had a pronounced effect on young consumers’ attitudes toward credit. These results are consistent with reports indicating that younger consumers are more likely to use debit cards instead of credit cards for purchases. They may also reflect the growing use of mobile payments by this younger generation.

As more young consumers eschew credit, credit card issuers may need to rethink established lending strategies for attracting and building relationships with these consumers. As part of that, issuers may benefit from supplementing traditional credit data with new data sources to develop more refined marketing and risk management tools for this dynamic demographic group. This will allow them to grow their portfolio responsibly and stay relevant in an increasingly competitive and changing landscape.

Regulating Credit Scores: When Good Intentions Make Bad Policy

2013-06-07T11:17:27-07:00

In previous posts, my colleagues have expressed concern over well-intentioned government proposals that seek to effect social policy by tinkering with the FICO® Score. Such actions threaten to diminish the predictive power of a risk management tool used by most lenders. This presents potential problems for both banks and consumers who may ultimately be impacted through higher rates and reduced access to credit.

Consider the case of a bill proposed recently in Montana. This legislation was intended to help consumers who were shopping for a mortgage, auto, student, or equipment loan, so that their credit score would not be harmed by multiple lender inquiries. The bill would have created a state law to require FICO to change its scoring logic to bypass all mortgage, auto or student loan inquiries in the last 30 days to 45 days.

This proposed outcome is not so different from what actually happens now with the FICO® Score when a consumer is shopping for credit. For quite some time, FICO Scores have used sophisticated logic to account for rate-shopping behavior. In fact, all mortgage, auto and student loan inquiries (but not “equipment loans”) that occur 30 days prior to scoring have no effect on the FICO Score. A recent blog post discussed the score’s rate shopping logic.

So why was the Montana legislation a problem? Because any change to the FICO® Score model has serious repercussions for lenders and consumers alike. Banks rely on the FICO Score as an objective, empirically derived risk management tool to help them accurately assess credit risk. They know that the development of FICO’s credit scoring models is a data-driven process, and that the predictiveness of the tool will be compromised when policy proposal mandate changes to the model without any analysis of whether the data supports the mandated changes.

The Montana bill also posed an additional problem of creating new operational costs for lenders. If adopted, the legislation would have required banks to use a Montana-specific scoring model when Montana residents seek credit while in-state or out-of-state. By making this change to the FICO® scoring logic, some banks might be forced to re-validate the new Montana FICO scoring model against their own portfolios. And if the Montana legislation were adopted, other states might follow its lead by imposing their own tweaks to the FICO scoring model. The impact on national lenders would be significant.

Consumers would also be negatively impacted if the operational challenges facing banks result in delays and higher costs, or, worse, cause some lenders to stop offering credit products in this relatively small consumer market. And if the FICO® Score became a less predictive underwriting tool, lenders would be more hesitant to extend credit to marginal borrowers.

Attempts to effect social policy by manipulating credit scores not only hinder the ability of banks to effectively assess credit risk. They can also result in many creditworthy borrowers being denied access to credit, or borrowers being extended more credit than they can safely handle.

The Montana problem was ultimately averted as FICO, banks, credit unions, retailers and credit bureaus convinced legislators in the Montana House of Representatives to table the bill. The message that the entire industry must continue to communicate in cases like these is that there are other, better ways to effect public policy changes.

Credit Card and Personal Loan Delinquencies Pull Down Russian Credit Health Index

2013-06-05T06:12:38-07:00

Q2 data from FICO and the National Bureau of Credit Histories (NBKI), Russia’s leading credit bureau, shows that the rise in credit delinquencies is still continuing — but only on some products.

Overall, Russian borrowers’ delinquent credit repayments rose slightly in April 2013, continuing a steady climb that started more than a year ago. The country’s FICO® Credit Health Index — which measures the percentage of loans and cards that are delinquent by 60 days or more — dropped one point from last quarter, to 108 points.

All regions but one have shown a drop in credit health since January of last year, but the same can’t be said of credit products. While delinquencies rose on unsecured loans and credit cards in the last quarter, late payments on mortgages and auto loans are falling.

As my colleague Evgeni Shtemanetyan points out, we’re only seeing small increases in delinquency, but the trend isn’t good. Personal loans and credit cards are at the bottom of Russian borrowers’ payment hierarchy, which means they need extra scrutiny from risk managers.

Analyzing Credit Trends by Age: Part Two

2013-06-03T09:53:35-07:00

In my last post, I shared research showing how older consumers have been increasing their debt loads, while younger consumers were doing the opposite. This post will dig deeper, looking specifically at credit trends by both age and loan type.

The graphic above summarizes the average outstanding debt for each age group by debt type. Of note:

Student loan debt is consistently higher for all age groups. This is no surprise given the rising cost of education over the last few years. Understandably, younger consumers have much higher levels of student loan debt. Some of the growing student loan debt for the older age groups may be explained by these consumers going back to school or parents co-signing loans for their children.
The magnitude of deleveraging is inversely related to age. That is, the younger the age group, the greater the relative change in outstanding debt level across all types of debt, except student loans.
All age groups reduced their outstanding credit card debt. The younger the age group, the greater the reduction in debt. One reason why older consumers may not have reduced their card debt as much is that some may have been forced into early retirement, and thus did not have a nest egg large enough to handle the transition.
Older consumers increased their outstanding auto and mortgage debt. It’s possible these consumers took advantage of low interest rates and low home prices to purchase investment and vacation properties.

Of course, this is only a high level analysis of what’s changing, and within each age group, there will be exceptions to the rule. Still, these findings underscore important changes in debt usage and credit behavior trends. They are a reminder that the credit landscape can be dynamic, and as a result, lenders must continuously track such changes and evolve their product offerings accordingly to remain relevant.

Stay tuned for my next credit trends post, where I'll take a closer look at how younger consumers are rapidly changing their attitudes towards credit.

Analyzing Credit Trends by Age

2013-05-30T09:42:51-07:00

One of the things I enjoy most about being a data scientist is when I get to slice and dice data in various ways to identify important new trends. Recently, my team completed an analysis of how consumers of different ages changed their credit behavior before, during and after the Great Recession. Our study reveals some interesting trends.

The chart above shows average debt of consumers by age group at four points in time: well before the recession (October 2005), immediately before the recession (October 2007), immediately after the official end of the recession (October 2009) and the most recent time period (October 2012).

Not surprisingly, outstanding debt peaked in the period immediately before the recession for most age groups. After October 2007, there was a steady decline in debt for all age groups—except for consumers 60 years and older.

In fact, the most notable observation is that consumers aged 60 and over were the only age group who increased their debt levels. If we only look at the October 2005 and October 2012 periods, consumers 40 and over have more debt today than they did in 2005. By contrast, younger generations owe less in October 2012 than they did in October 2005.

Next, my team looked at how these changes impacted consumer FICO® Scores, given the importance of indebtedness on credit risk. The chart below shows the percentage of consumers by age group with FICO Scores greater than or equal to 760. For example, 11.2% of consumers in the 18-29 age group score 760 or greater in October 2012.

We observe two contrasting movements at the two ends of the age spectrum. A greater proportion of young consumers score higher in recent time periods, while a smaller proportion of older consumers have high FICO® Scores. Keep in mind that carrying lower debt has a positive influence on the FICO Score.

In an upcoming post, I’ll peel back another layer of the onion to better understand what is driving these changes in outstanding debt.

UK Cardholders Are Careful to Avoid Overlimit Penalties

2013-05-28T06:02:41-07:00

There were three important messages about cardholder behavior coded in the latest FICO data on UK card performance.

First, despite a struggling economy, UK cardholders are doing a good job staying within their credit limits. Overlimit accounts among classic cards were at their lowest level — just over 2 percent of cards — since FICO began benchmarking performance in 2002. I’ve spoken before about credit cards as “survival tools,” and it certainly seems that cardholders in the UK are protecting theirs.

The second message is that newer cards look riskier. The percentage of balance paid reached a three-year high last quarter for “veteran” accounts open for more than five years (26 percent in January) and “established” accounts open for one to five years (30 percent in February). However, the same measure reached a two-year low in March for new accounts (just under 14 percent). New cards are also closer to their limits, with utilization at its highest level since the third quarter of 2008 — this compares with declining utilization for veteran and established cards.

Finally, the first quarter after Christmas showed an expected rise in delinquent accounts but then an instant fall in February or March. Usually, those holiday expenditures linger longer, meaning that UK cardholders are focused on paying down their debt more quickly.

All in all, it appears that UK cardholders are managing their debt carefully. The ones for portfolio managers to watch are those newer cards.

If you’re interested in getting your hands on the full complement of data every quarter from the FICO® Benchmark Reporting Service, benchmarking your portfolio against industry performance, contact Stacey West at staceywest@fico.com.