Click fraud detection is a set of procedures and actions aimed at detecting and filtering fraudulent clicks which occur on pay per click (PPC) programs.
This definition focuses on fraud detection, see click fraud for more details on different types of fraud.
Fraud detection is done by PPC network operators, i.e. Google AdSense, advertisers and third party vendors.
Network operators are not directly affected by click fraud because fraudulent clicks are billed to the advertisers. On the contrary, over the short term, the higher the fraud is, the higher their revenues are. However, well known actors fight and filter invalid clicks to maintain advertisers’ trust.
To detect click fraud PPC network operators observe and analyze:
duplicate or repetitive clicks
time repartition of the clicks
publisher behaviors (i.e. Google Adsense partners)
post click behavior (when networks access to this data)
An example of IP address based fraud detection (Source Trademob Whitepaper):
An example of fraud detection based upon daily click pattern (Source Trademob Whitepaper):
Advertisers have also theorically a front-row seat to observe and detect click fraud by observing browsing and conversions on their websites. They can isolate traffic source segments which widely underperform compared to the average conversion rate or, even worse, traffic taht never convert.
Thus, it is relatively easy for a merchant to identify poor affiliate fraudsters or malicious AdSense partners. When the conversion is not an order, the fraud may be more difficult to detect because humans or bots can complete webforms.
Third party services can also provide click fraud audit and detection services by using the same means. They can also have fraudster databases. They allow to automatically produce potential fraud reports for asking for refunds from PPC search engines and networks .
The Click Report methodology:
An example of click fraud detection pricing:
A warning message for competitors’ fraud:
How Google detects and processes invalid clicks on Google Adsense :