The controversy surrounding cl? fraud comes up every year, but it reached a fever pitch during December's Search Engine Strategies conference in Chicago when participants voiced concerns over experiencing fraudulent cl? rates ranging from 20 to 40 percent, threatening the entire paid search industry.
At the time, Google's Business Product Manager for Trust and Safety Shuman Ghosemajumder tried to calm advertisers' fears explaining that Google was currently "...examining ways to make its fraud-fighting efforts more transparent without revealing crucial information that might help swindlers elude detection." Ghosemajumder did, however, express concerns over revealing too much information, fearful it would give away algorithm secrets to competitors.
Paid Search Revenues Continue to Rise
While the major search providers have always insisted the cl? fraud rate is a gross overestimation, a 2005 Outsell survey found that cl? fraud was a $1.3 billion problem for publishers. At the time, many advertiser respondents (27 percent) said they planned to cut back and/or eliminate paid search campaigns in 2006.
Outsell respondents may have intended to cut down on paid search, but they certainly didn't follow through. SEMPO's year-end search marketing report showed that North American advertisers spent $8 billion on paid placement programs in 2006, amounting to 86 percent of 2006's total search spend ($9.4 billion). Seventy-one percent of SEMPO respondents said they used paid search campaigns, illustrating that there were not many defectors.
Despite advertisers' insistent claims that the search engines don't do enough to eliminate cl? fraud, paid search revenues continue to fill the coffers of Google, Yahoo, Microsoft and many second and third tier search engines. Additionally, there is a huge gap in the professed prevalence of cl? fraud between the search providers and the advertisers and cl? fraud advocates.
Google Cl? Fraud Estimate
The rate of cl? fraud changes depending on whose numbers you believe. Cl? fraud detection agencies put the cl? fraud rate hovering around 14 percent, while others believe at least 20 percent of all clicks are fraudulent.
Late last month, Google issued a statement on the Inside AdWords blog that insisted invalid clicks consistently remain under 10, typically in the single-digits, and that virtually all malicious activity is found by Google's filter. Ghosemajumder claimed the percentage of clicks found by advertiser-initiated investigations account for just .02 percent of clicks. All other accounts, he said, are grossly overestimated.
Alchemist Media President Jessie Stricchiola takes issue with Google's assertion that it refunds advertisers promptly for fraudulent clicks, stating that "Google has been the most stubborn and the least willing to cooperate with advertisers".
Google Cl? Fraud Filters
In February, Google outlined the three-layer filtration process it uses to combat and eliminate cl? fraud. They described the system which uses both proactive and reactive filters as follows:
1. Proactive Filters: Automated algorithms analyze and filter out invalid clicks in real-time without billing advertisers for these false clicks. This accounts for the vast majority of invalid cl? detection.
2. Proactive Offline Analysis: Post billing, Google uses automated and manual analysis to identify fraudulent clicks that somehow made it through the first layer of filtration. Special attention is paid to clicks occurring on the AdSense network. This is done pro-actively and without any involvement from advertisers. When false clicks are found, advertisers' accounts are immediately credited via Cl? Quality Adjustments.
3. Reactive Investigations: Investigations take place when an advertiser approaches Google concerned about suspicious activity on their account. Each complaint is investigated, though Google says refunds are relatively rare. Google claims that the vast majority of fraudulent clicks, more than 99 percent, are found and thrown out within the first two stages of filtration. The third stage only includes the .02 percent of clicks where advertisers are affected by undetected cases of cl? fraud.
Cl? Fraud Detection Agency Estimates
In April 2006, The Cl? Fraud Index reported an industry-wide average cl? fraud rate of 13.7 percent. The cl? fraud rate was broken down as follows:
- Tier 1 search providers -- 12.1 percent
- Tier 2 search providers -- 21.3 percent
- Tier 3 search providers -- 29.8 percent
Some of the newer cl? fraud prevention firms like Cl? Assurance and ClickLab off?algorithm-based programs to limit bad clicks. These programs estimate the statistical likelihood of a cl? being fraudulent based on behavioral variables and IP address.
Gap in Prevalence of Cl? Fraud
As noted above, Google admits to a < 10 percent cl? fraud rate, while advertisers and cl? fraud detection agencies believe it is more like 14 to 20 percent. Ghosemajumder explained this gap saying that many advertisers and cl? fraud detection agencies are looking at the wrong signals, mistakenly classifying valid clicks as fraudulent. Additionally, he believes many advertisers request refunds for clicks already thrown out during the first two layers of the filtration system.
For example, misclassification might occur when counting reloads of an advertiser's landing page. Say the customer clicks through to the landing page, views a product page, and then hits the back button, returning to the same landing page. Without proper tagging, that one cl? and five page re-loads could be misclassified as 6 clicks from the same visitor. Google argues that there are hundreds of different signals that must be monitored to detect cl? fraud, signals that are a closely guarded company secr?and known only to the Google cl? quality team.
A Solution for Cl? Fraud
Like other experts, we believe the only solution to cl? fraud is for independent auditors to evaluate the system using accurate data provided by the search engines and advertisers themselves. It is the only way to get a neutral calculation -- the current cl? fraud detection agencies may not be entirely neutral, and certainly the search providers are not neutral. We need an independent agency that has no incentive to ?rease or decrease the cl? fraud rate. One solution could be to use a technology firm like Fair Isaac, which is currently conducting cl? fraud research for SEMPO.
One thing is certain, until advertisers are willing to provide campaign info, and the search engines are willing to share cl? fraud data, we'll nev?know the actual prevalence of cl? fraud or how much advertisers are losing as a result.
About The Author
Nick Guastella, SEM Analyst and PPC expert at Bruce Clay, Inc., has been active in search engine marketing for the past six years. Nick has worked both sides of the PPC fence. He has insider knowledge gained from a stint starting in 2000 at GoTo/Overture, the original Pay Per Cl? advertising pioneer. He then crossed over as Account manager for ValueClick.com.
