Collusion Fraud: Time to go to the mattresses
Sonny Corleone, in The Godfather, said “I want Sollozzo. If not, it’s all-out war: we go to the mattresses.” While the origin of the “mattress” phrase is debatable, the meaning is clear: it’s time to get serious and face this threat head-on. (Yes, this is the line Tom Hanks referenced in “You’ve Got Mail.”)
Why am I writing about gangster movies? Because that’s the image that “collusion fraud” conjures up, appropriately. Collusion fraud occurs when online publishers use rotating IP addresses, or botnets, or just armies of people, to click on paid search ads on their own sites. Collusion fraud is hard to find because the fraudulent clicks are spread across dozens (or hundreds) of sites and the clicks are generated from dozens (or hundreds) of different IP addresses. So everything looks kosher. But it’s not.
Standard click fraud detection mechanisms employ various anomaly detection rules. 1,000 clicks from the same visitor in the span of a minute, or an hour, or week, is a traffic anomaly that’s easy to find. Almost all search engines and ad networks would screen this out. But the same number of clicks spread across hundreds of sites by hundreds of “visitors” could look totally normal. Anomaly detection rules won’t help find it.
Click Forensics employs proprietary technology utilizing high-dimensional cluster analysis to find publisher collusion. Ad networks rely on this technology to find fraud that they would miss with their standard in-house anomaly detection rules.
The first separate “product feature” that results from this advanced cluster analysis is the Block List that we announced earlier today. Utilizing the block list feature, ad networks can eliminate this collusion fraud from online campaigns. They won’t pay for fraudulent clicks, and they’ll deliver higher quality traffic to advertisers.
We’ll write a lot more about this technology in future posts. In the meantime, it’s time for ad networks to go to the mattresses.