Andy Beal sat down with Google business product manager for trust and safety Shuman Ghosemajumder and got to see some graphs that previously had not been seen by anyone not working for Google. He’s recreated the graphs at his blog, MarketingPilgrim, and explained the whole deal. It boils down to this:
Google filters AdWords/AdSense clicks four times. The first two are completely automated, removing first detected invalid clicks in real time, then again using a flagging process to automatically remove some more that may have passed the first level. In the third level, Google employees manually review and remove suspicious clicks. Finally, if a click makes it past those three, it is considered effective click fraud, and gets invalidated later by advertisers and click fraud companies. These clicks are still removed in this fourth review step, and refunded to advertiser’s accounts. All in all, these make up less than 2% of all Google ad clicks.
Click fraud detection firms love quoting higher and higher percentages of click fraud that goes undetected, because the more click fraud, the more you need them to swoop in and find those clicks. I’ve heard anywhere from ten to twenty percent (and more, if the detection firm is really desperate) of clicks are fraudulent, and yet here is Google, probably the company that deals with these clicks the most, saying it comes out to less than two percent.
Here’s an idea, click fraud companies: prove it. Google is willing to quote a number, so now lets see you prove things are actually worse. Just find the clicks. Google will refund invalid clicks, so all you have to do is show the refunds for your clients coming significantly higher than 2%. Good luck.