Fair Isaac Claims Pay-Per-Click Fraud Is 10% to 15% from InformationWeek says Fair Isaac is going to report today that it has found the click fraud rate of billed paid search and contextual advertising to be 10 to 15 percent, though the figures are based on what the Associated Press characterized as being a “handful” of web sites.
The InformationWeek article says this rate is for clicks that are actually billed to advertisers. The AP article makes no mention of billing but instead suggests this is simply a traffic analysis:
After reviewing a handful of Web sites since last August, Fair Isaac believes 10 to 15 percent of the advertising traffic is “pathological,” indicating a likelihood of click fraud, said Joseph Milana, the company’s chief scientist of research and development.
Less Than 5% Click Fraud Makes You “Virtually Free” and Google: Click Fraud Is 0.02% Of Clicks and Yahoo Says 12 to 15 Percent Of Clicks Are Discounted all explain in more depth the important of understanding that search engines already discount or automatically credit for many suspicious clicks.
Until we see the actual report, it’s hard to compare. More to come later, when we have more information. In addition, Fair Issacs warned against using the figure as an industry-wide representation:
But if Fair Isaac aims to have a similar impact on click fraud, it will first more complete data. Milana concedes his results are based on small number of advertisers. “There’s no way we could make a general statement about what’s happening in the marketplace,” he said.
Postscript: Google has given me a statement on this report:
As you noticed, FI is saying 10-15% of “billed” clicks. That means, if Google throws out something less that 10% and Yahoo throws out 12-15%, that their number would be another 10-15% on top of that . We’ve seen no data in our extensive research on this topic to indicate advertisers are impacted anywhere near this amount. In fact, undetected click fraud that is brought to our attention by advertisers amounts to less than .02% of clicks.
Also, the FI estimate is not similar to Click Forensics. The ~14% quoted in the Click Fraud Index does not take into account clicks filtered or discounted by the engines. Chris Sherman notes this on his coverage of their last metric announcement: At first glance, these numbers may seem alarming, but they may not take into account the discounting of questionable clicks done by most search engines. The point is that FI is coming out with an estimate that is potentially twice as high.
Finally, as I’m sure you’d agree, you can’t claim “4 out of 5 dentist prefer” and really only ask 5 dentists, which is what appears to be the case here. Shuman talked to FI and confirmed that their sample size is “less than 10″ advertisers, none of whom were using auto-tagging. This means that their log-based analysis counts page reloads and back-button activity as “clicks”.