The session was opened by Frank Watson, Kangamurra Media, introducing Jon Myers of MediaVest, who provided a historical overview of the click fraud issue, including a "how it was" slide from 2007 (this is a fast-moving issue). He cast the issue as a David and Goliath conflict, in which Google and Yahoo! clearly occupy the heavyweight category.
Jon provided some really interesting maps, showing the locations where fraud risk was highest. While India, southern South America, Mexico, etc. were still in the red, it's interesting to note that the risks are up in the US, but somewhat down in the former Soviet Union and China.
The magnitude of fraud (UK alone) he calculated at over 26K Pounds Sterling during just the time consumed by the current session! In contrast, he noted help on the way from a growing number of auditing firms and councils. Finally, he provided some helpful hints for self-monitoring discipline . . . after all, if you are the potential victim, it's your responsibility to spot the problem first.
Andrew Goodman, Page Zero Media, continued with Preventing Click Fraud Through Proactive Account Strategy. His opening assertion was interesting, that "rogue behaviour" seems endemic to the Internet (due at least in part to the "savvy gap" and relative anonymity of the web). Given that, he did not recommend that your in-house SEM people should be spending their energy on "policing" when they have other more pressing issues to attend to. He showed a graphic from PPC Assurance, a 3rd party monitoring service, which showed how they could help you identify and address click fraud issues.
His key advice was to "build it (campaigns) right" to avoid fraud, leading with a "granular, well organized campaign." He also had thoughts regarding keywords, cautioning against the use of "head keywords" and surprisingly "tail keywords" as well. He suggested using two word phrases, and going for the "torso" of the keyword list to avoid the obvious threats. His advice on monitoring was "track, iterate and adjust" looking for odd traffic, high bounce rates, etc. (If, for a given keyword, time spent on page less than 5 seconds is 60+%, good chance it's fraud).
Finally, we heard from Shuman Ghosemajumder, Business Manager for Trust & Safety, Google. He described the ways that Google protects against click fraud, including their treatment (and non-charge) for invalid clicks. By casting the invalid clicks web wide (large "false positive'' pool), they attempt to save users from improper charges. He noted that Google does, in fact, have an economic incentive for addressing the issue, which relates directly to their need to compete (for dollars) on the basis of advertisers' ROI.
Shuman provided an explanation of Google's 3-part system for invalid click detection, Proactive (Real-Time) Filters and Offline Analysis and Reactive Investigations. According to Google, < 10% of clicks are filtered today, most proactively. Less than .02% are detected reactively. As for tools, some employ simple rules (e.g. >1 click per IP within n seconds) and some are found through statistical anomaly detection, a far more sophisticated technique.
Thanks to Frank Watson, for moderating a great session, and an excellent, if short, Q&A.