[VengefulCynic]

Where companies like Sift interest me is moving from a place where a system determines whether or not it believes a charge is fraudulent to a place where you evaluate the likelihood that an individual transaction is fraudulent. And they're absolutely right that individual components are going to lead to tons of false positives and then eventual end-gaming of their countermeasures.

As an example, in the implosion following the dot-com boom, it was almost impossible for denizens of many Eastern European and African countries to buy things online because the incidence of fraud attributed to their country was high enough that the fraud officers at many online retailers wouldn't chance it. I have a friend who lived in Romania who had a credit card that used my address in the US and frequently required me to ship stuff for him because no companies would ship directly to him. He loves the modern era of fraud evaluation, because while his country is weighted negatively, his use of a respected international bank, an IP address in the same area as his billing address and purchase history with many online companies has greatly reduced his level of hassle due to false positives alleging that he's a fraudster.

[brandonb]

Definitely true. A lot of sites just block all transactions from, say, Africa when they start getting hit with fraud. But that's a very broad brush to paint with. It's better to use a machine learning system that looks at many signals so that people like your friend can shop online without being hassled.