June 9th, 2006

classic cylon

(no subject)

You may recall sometime back I mentioned someone was going to do a click fraud experiment in which they'd purchase some ads, click on them, and see at what point, if any, Google's fraud detectors are tripped. According to this entry, the experiment has run. The resultant thread is rather long, with some very interesting perspectives given regarding the degree to which it is possible to detect click fraud, various methods of doing so, alternate business models (mine is fixed fees), and how market forces may render the entire debate moot.

FWIW, the individual running the experiment claimed to use up a fair amount of the ad spend without raising any of Google's alarms. There apparently will be a more formal report forthcoming, pending the review of an NDA (nondisclosure agreement) the experimenter has with a former employer.

The results don't really surprise me; I described means by which this could be done while I worked at AV. I'm not sure this will have much of an impact, and it could very well be dismissed by Google. (After all, Google is not (yet) under any obligation to release the actual click streams for the experiment.) Another possibility is that Google might pursue legal action against the experimenter. Although the test involved only the experimenter's ads, Google's terms of service state they may seek civil or criminal penalties against anyone who attempts to defraud their service. There is no exception made for anyone running their own tests. However, if the experiment was actually run, and Google chooses not to contest it, it could be a point in favor of the plaintiffs in the click fraud cases filed against Google.