I've updated this post with several clarifications. I've also blogged separately about how these numbers should be looked at, in Fair Isaac Click Fraud Report Spreads False Alarm.
Preliminary data from a study of click fraud by Fair Isaac Corp (FIC) shows that, in the limited cases it was able to study, 10 to 15 percent of billed pay-per-click traffic was deemed "pathological," indicating a likelihood of click fraud, according to Joseph Milana, the company's chief scientist of research and development.
While many reports are extrapolating this data to say that this should be considered an industry-wide click fraud rate, Milana said this was not FIC's intent, and the data should not be used that way. It's press release was carefully worded, but inevitably taken out of context, as I explain here.
"A limited number of advertisers have been able to provide data of the quality that we're asking for," Milana told SEW. "We're not saying these results are definitive."
The click fraud was detected using a combination of Fair Isaac's profiling technology, similar to the kind it uses to detect credit card fraud, and its "anomaly detection engine," which it uses to detect medical insurance fraud.
In March, Google reported that "under 10-percent" of clicks could be categorized as "invalid clicks," which Google catches before advertisers are charged. The amount of invalid clicks that are not proactively detected and are caught by advertisers is less than 0.02 percent. Also in March, Yahoo reported its "network discard rate," representing the average number of clicks (in aggregate) that its clickthrough protection filters identify, tag and do not bill to advertisers, is between 12 and 15 percent.
In January, Click Forensics put click fraud at 14.2 percent, which would compare to Google's "under 10 percent" number. Click Forensics does not offer a corresponding number to Google's "less than 0.02 percent" figure of invalid clicks that are billed to advertisers. That's the number that Fair Isaac's numbers would correspond to.
I've gotten a statement from a SEMPO spokesperson:
Click fraud continues to be a concern for SEMPO and our primary objective at this point is trying to establish credible scope of the problem. We supported Fair Isaac because of their excellent track record in financial fraud.
While we can't really comment on the initial findings announced by Fair Isaac because of the extremely limited scope of the study, we are actively encouraging others to get involved in the study to give Fair Isaac a large enough dataset to be able to accurately scope the incidence of both invalid clicks and pathological behavior.
If advertisers want to get involved to determine if there's pathological behavior in their dataset they can contact either SEMPO or Fair Isaac to learn the details of the type of data required. There is no cost to the advertiser for this participation.
From Google's Shuman Ghosemajumder:
The amount we filter proactively is consistently less than 10% of all clicks but individual invalid click rates can vary by advertiser, campaign and even keyword. The relatively rare cases we find of advertisers being affected by undetected click fraud constitute less than 0.02% of all clicks. Without knowing more about Fair Isaac's data gathering and methodology it is difficult to comment on their study except to say that a handful is not a representative sample size for our hundreds of thousands of advertisers.
And from Yahoo's Reggie Davis:
Yahoo is deeply committed to fighting click fraud, because our business is built on the high ROI we deliver to our advertisers. Click fraud erodes advertiser trust – trust that is key to the success of our business – which is why Yahoo employs an aggressive Clickthrough Protection System that identifies and does not bill advertisers for 12-15% of the clicks on our network.
We are focused on building the world's highest quality, most effective search and display advertising marketplace, and fighting click fraud is a significant part of that effort. Without having more information about their methodology, and given what seems to be a small sample size, it's difficult to comment on Fair Isaac's research.
While we welcome third-party viewpoints on how the industry can strengthen its fight against click fraud, we strongly believe that the industry needs a single set of definitions and standards around click measurement to ensure that we're all speaking to advertisers in the same language. Yahoo is working closely with the IAB as part of the IAB Click Measurement Working Group to help drive the development of industry definition and standards and looks forward to auditing our systems against those standards in the future.
I've spoken with Fair Isaac, and have more updates and some comments on this data in another post, Fair Isaac Click Fraud Report Spreads False Alarm.