Big Data Big Analytics

The Power of Analytics: Take Credit Card Fraud as an Example

February 17, 2011 at 7:21 pm Mary Ludloff 10 comments

By Mary Ludloff

Let me tell you a story about my credit card company, Capital One. One evening, I was sitting in my living room watching 60 Minutes and I get a phone call. Before I pick up, I hear the following: “This is John from the Capital One fraud department and we believe that someone has stolen your credit card information.”

Naturally, I pick up the phone. Poor John. He wanted to verify that I was who I said I was and before I was willing to do that, I wanted to verify that he was who he said he was (as you may have guessed from my previous posts on privacy, I don’t like to give out personal information). Once we got over that hurdle, we had the following conversation:

John:  “Within the last five minutes or so, did you or a family member make an iTunes purchase for 99 cents?” Me: “No.” John: “Did you or a family member just get back from Las Vegas and stay at fill-in-the-blank hotel which is charged to your account?” Me: “Yes.” John: “While in Las Vegas, did you eat at this fill-in-the-blank restaurant which is charged to your account?” Me: “Yes.” John:  “We believe that someone skimmed your card information and is now testing your card with a small purchase to see if you are aware that your information has been stolen. We have flagged your card for possible fraud and will now cancel it and issue you a different account number. I’ll get your new cards to you by Tuesday morning.”

Now, as you all know, I spend pretty much all my time steeped in analytics from a function (marketing) and business (big data, big analytics) perspective.  So you should all feel very sorry for John because I kept him on the phone for 15 more minutes trying to find out how Capital One’s fraud department could pinpoint that iTunes’ fraudulent purchase in under five minutes. I was impressed and you should be too because here are some facts about the credit card industry that you may not be aware of:

• Credit cards are responsible for $2.5 trillion in transactions a year at more than 24 million locations in 200 countries and territories. (Source: American Bankers Association, March 2009)
• It is estimated that there are 10,000 payment transactions every second around the world. (Source: American Bankers Association, March 2009)
• Card fraud costs the U.S. $8.6 billion annually and the bulk of that loss falls on the card issuers. (Source: Aite Group LLC)

That is a lot of data and its being generated every second! But as we’re fond of saying at PatternBuilders: “It’s what you do with the data that matters.” And in terms of predictive modeling, the credit card industry (and finance in general) is way ahead of the curve. And it should be: if you can reliably predict where and when fraud might occur, you can have a significant impact on your bottom line. Of course, this pretty much applies to any industry. For example, HP added $20 million to its bottom line by identifying fraud activity in its printer services business. This is why we included fraud detection capabilities in our upcoming release of PatternBuilders Analytics Framework.

Okay, back to poor John and Capital One. Since predictive modeling for fraud is so critical, the credit card industry, companies, and John are pretty close mouthed about what they look for. However, John did tell me that their algorithms are constantly changing based on how fraud is perpetrated. In the case of iTunes, once Capital One “discovered” that small online purchases like this were being used to see if a card was active and could be used, this information was added to their predictive model. Of course, one must also consider active iTunes purchasers versus inactive ones, etc., because you want to make sure that you flag cards with a fraudulent alert that might actually be fraudulent and not typical buying behavior for the card holder. But that’s not all. In my case, John also shared the following with me:

• I was in a small “cluster” of cardholders that had the iTunes 99 cent purchase AND happened to use their credit cards at the same restaurant in Las Vegas.
• Within this cluster, the iTunes purchase happened within 12 to 48 hours of signing off on that meal.
• Capital One was sharing this information with the Las Vegas police department because they were pretty sure that someone in that restaurant was skimming credit card information.

Here’s the corker. Once I got off the phone, my husband sent out an email to everyone that was on the Las Vegas trip and told them what happened because more than one credit card has been used at that restaurant. Interestingly enough, one of his friends discovered something similar on his card (not Capital One) and reported it to his credit card company.

Although the impact of big data is now felt by every industry and by extension, every company or organization, I like to point out that the retail and financial services industries were grappling with it long before the term was coined. Their sophisticated systems cost, in aggregate, billions of dollars but resulted in huge payoffs. Other industries are not so fortunate; while they have the same challenges, they simply don’t have the technical and financial resources available to them. This is why off-the-shelf analytics systems, like ours, are needed.

Now, a final word on credit card companies and preemptive fraud detection; it not only benefits the company but the individual card holder as well. In my case, I was not held responsible for the 99 cent purchase and saved from the hassle of reporting fraudulent activity on my card once I discovered it. This is why I like to say, “No matter the industry, analytics are a beautiful thing.”

Entry filed under: Data, General Analytics. Tags: analytic systems, analytics, big data, fraud analytics, PatternBuilders Analytic Framework, predictive modeling.