Revealed: how Google manages click fraud

July 25th, 2006

Automated click fraud deviceIn February 2005, Google, Yahoo, and Time Warner. were sued by Lane’s Gifts & Collectibles in a class-action lawsuit over click fraud. The company alleged that Google and the other companies had been improperly billing for pay-per-click ads that were not viewed by legitimate potential customers. The case was settled earlier this year and as part of the settlement Google agreed to have an independent expert examine their click fraud detection methods, policies, and procedures and make a determination of whether or not they were reasonable measures to protect advertisers. The expert was Alexander Tuzhilin, a Professor of Information Systems at NYU. A summary of his report has been posted in Google’s official blog:

“The bottom-line conclusion of the report is that Google’s efforts against click fraud are in fact reasonable. At several points in his report, he calls out the quality of our inspection systems and notes their constant improvement. It is an independent report, so not surprisingly there are other aspects of it with which we don’t fully agree. But overall it is a validation of what we have said for some time about our work against invalid clicks.”

The full report contains lots of interesting information on Google’s approach to dealing with click fraud. Here’s the high-level description of the approach

“Google has built the following four ‘lines of defense’ for detecting invalid clicks: pre-filtering, online filtering, automated offline detection and manual offline detection, in that order. Google deploys different detection methods in each of these stages: the rule-based and anomaly-based approaches in the pre-filtering and the filtering stages, the combination of all the three approaches in the automated offline detection stage, and the anomaly-based approach in the offline manual inspection stage. This deployment of different methods in different stages gives Google an opportunity to detect invalid clicks using alternative techniques and thus increases their chances of detecting more invalid clicks in one of these stages, preferably proactively in the early stages.”

The report also has a good overview of the pay-per-click model and Google’s AdSense program. Professor Tuzhilin concluded that the basic approach is sound and the thresholds set entirely by the engineering team with no input from financial officers. Here’s his bottom line:

“In summary, I have been asked to evaluate Google’s invalid click detection efforts and to conclude whether these efforts are reasonable or not. Based on my evaluation, I conclude that Google’s efforts to combat click fraud are reasonable.”

If you don’t have time to read the full 47 page report, the Search Engine Watch blog has summarized some of the most interesting findings.

While this report is very interesting to the technical community, not all of the plaintiffs are happy with the $90M settlement and some are fighting it. In a related development, CNET reports that Google has announced it will show advertisers the number of invalid clicks on their ads starting Tuesday.

Full disclosure department: the UMBC ebiquity site carries AdSense ads and uses the resulting income to support Mr. Capresso.