The biggest problem with ML is that it takes time for it to react to the new fraud patterns/schemes. While rules engines have their limitations, they also have

The biggest problem with ML is that it takes time for it to react to the new fraud patterns/schemes. While rules engines have their limitations, they also have one big benefit: they allow to block recognized fraud really fast. So the actual question here is whether the ML for detecting fraud is better than a flexible rules engine and goods analysts/statisticians. In my personal experience, statistical analysis and anomalies detection effectively handles majority of the fraud. I would be interested to see a more detailed analysis of Sift Science performance with some numbers for false positives/false negatives for example, though (of course) it is probably proprietary information.