August 9, 2006 at 17:49Google: Independent Companies Count Click Fraud Incorrectly
Google experts sorted out and published their own detailed report with a critical assessment of the work of independent analytical companies. As you know, recently several different companies at once , including AdWatcher , ClickFacts and Click Forensics , have published their own estimates of the level of click fraud in the search engines Google, Yahoo and others. According to their estimates, the number of "fake" clicks on advertising links ranges from 10% to 40%, depending on the advertising campaign and the keyword.
Google experts performed a “technical analysis” of the work of independent consultants and found several fundamental errors in their methodology, which made the estimate of false clicks too high. A detailed analysis and explanation is provided in a 17-page document (PDF file).
The main fundamental mistake of independent analysts is that they are not able to calculate the actual number of clicks on links and instead evaluate other factors, for example, the number of visits to the advertised page. In this case, if the page is reloaded or the user visits the same page again, analysts regard this as a double click on Google contextual advertising and record it as a “fake” click. This can happen if the user examines the advertiser's site, and then presses the back button and returns to the first page. In fact, there was no click fraud, but analysts recorded it.
This is not the only example of how independent analysts “come up” with fictitious clicks, which actually did not exist. Another fundamental mistake is the incorrect implementation of cookies, with which analysts and advisors monitor user behavior. One of the most well-known consulting firms, for example, has implemented cookies in such a way that clicks on Yahoo ads can count as clicks on Google ads and vice versa.
These are the two most serious fundamental errors in the work of independent experts. Because of them, the results of counting advertising conversions are seriously distorted. In some cases, distortion exceeds 50%. For example, in one of the cases, an independent consultant recorded 1278 “fake” clicks in his report, while only 850 clicks were recorded in Google’s logs.
A Google report cited specific cases of “blunders” in analyst reports. For example, in a Click Forensics report dated May 2, there were 6 clicks on Google sponsored links from one IP address in 9 minutes, and only 2 clicks in seven minutes were counted in Google logs. The February 4 ClickFacts report recorded 54 clicks from a single user over a 26-second interval, but Google actually billed that advertiser in just one click.
Further evidence of the defective methodology of AdWatcher, ClickFacts and Click Forensics is directly in their reports. There they indicate the level of conversion, including for "fake" clicks. It turns out that it is unrealistically high and in one case reaches even 5.1%, while normal transitions show a conversion of 5.8%. This is further evidence that firms cannot properly identify click fraud.
Google experts once again emphasize that serious gaps in the methodology are noted by all consultants without exception who are trying to evaluate click fraud on search sites.
Google experts performed a “technical analysis” of the work of independent consultants and found several fundamental errors in their methodology, which made the estimate of false clicks too high. A detailed analysis and explanation is provided in a 17-page document (PDF file).
The main fundamental mistake of independent analysts is that they are not able to calculate the actual number of clicks on links and instead evaluate other factors, for example, the number of visits to the advertised page. In this case, if the page is reloaded or the user visits the same page again, analysts regard this as a double click on Google contextual advertising and record it as a “fake” click. This can happen if the user examines the advertiser's site, and then presses the back button and returns to the first page. In fact, there was no click fraud, but analysts recorded it.
This is not the only example of how independent analysts “come up” with fictitious clicks, which actually did not exist. Another fundamental mistake is the incorrect implementation of cookies, with which analysts and advisors monitor user behavior. One of the most well-known consulting firms, for example, has implemented cookies in such a way that clicks on Yahoo ads can count as clicks on Google ads and vice versa.
These are the two most serious fundamental errors in the work of independent experts. Because of them, the results of counting advertising conversions are seriously distorted. In some cases, distortion exceeds 50%. For example, in one of the cases, an independent consultant recorded 1278 “fake” clicks in his report, while only 850 clicks were recorded in Google’s logs.
A Google report cited specific cases of “blunders” in analyst reports. For example, in a Click Forensics report dated May 2, there were 6 clicks on Google sponsored links from one IP address in 9 minutes, and only 2 clicks in seven minutes were counted in Google logs. The February 4 ClickFacts report recorded 54 clicks from a single user over a 26-second interval, but Google actually billed that advertiser in just one click.
Further evidence of the defective methodology of AdWatcher, ClickFacts and Click Forensics is directly in their reports. There they indicate the level of conversion, including for "fake" clicks. It turns out that it is unrealistically high and in one case reaches even 5.1%, while normal transitions show a conversion of 5.8%. This is further evidence that firms cannot properly identify click fraud.
Google experts once again emphasize that serious gaps in the methodology are noted by all consultants without exception who are trying to evaluate click fraud on search sites.