WHY CARDS GET REFUSED
Anyone who’s had their card declined knows what happened next. The service agent verified my identity, asked me to review recent purchases, and then apologized profusely and removed the block on the card. But since I’m curious about the machinery behind such decisions, I dug a bit further.
I asked him whether I had correctly informed Visa of my trip—I had. I asked if he had access to past travel purchases showing I was in California—he did. And I asked him why the other purchases had been approved but this one hadn’t—apparently, Target stores in the East Bay are subject to credit card fraud, particularly in the purchase of personal electronics.
We’re wooed by the promise of Big Data to make our lives better. So beyond just venting about my first-world camping problems, I want to use it as an example of why, even armed with all the right data and tools, we don’t act.
Credit card fraud is a huge problem for the finance industry, costing $190B in 2011 (though much of this is online, not physical retail.) So there’s certainly motivation to tackle the issue. You’d think it would be easy to predict well—there are few industries with this much authenticated data available. In my example above, there are plenty of details that could have been used to improve the prediction.
Based solely on the data that the credit card company, and Target, had:
The system could have looked at the geography of purchases to construct a travel profile, seeing that I was headed towards Target and giving it confidence that I was a legitimate buyer.
Moreover, it could have looked at the things I’d bought—mostly junk food and boxed wine—to see that I wasn’t buying the kinds of things fraudsters often acquire.
Finally, it could have looked at the places I’d shopped before, and see that I’ve shopped at this Target in the past, and not filed a stolen card report, and approve the transaction.