Some Adtech vendors are developing in-house solutions for handling ad-fraud. Usually they are recruiting several data-mining experts and try to solve ad fraud using data analysis. This is a great way to solve the non-sophisticated fraudsters / fraudulent traffic. However, more sophisticated fraudsters know how to create the exact behavior - CTR/CPV/CPI etc. that advertisers are expecting, making it harder to detect fraud with data analysis. Fraudsters randomize all possible variables such as browsers, operating systems, websites and more. We heard about fraudster groups that even record keyboards and mouse movements and then 'play' them randomly. These kind of methods break data-mining models. In one case, we analyzed data for a customer that had $100k fraud chargeback on the day we first met him. All of his parameters: CPM/CPC/CPI and even User LTV (life time value) seems "by the book" but all of it was pure fraud... -Z