We've been investigating click fraud for the past couple of years using various tools, both commercial and home-grown. I also help the programmers who put together most of the currently available commercial services.
Our best analyses have come from our home-grown programs that involve extensive log analysis and filtering.
I've got to say that while I'm certain click fraud activities are on the rise given the increase in Google AdSense publishers and other PPC affiliate programs being offered by the search engines, and given the increasing costs-per-click in some competitive areas, we believe that Google and Overture are doing an excellent job of filtering fraudulent clicks and keeping the cost of those clicks out of our invoices.
I have worked closely with the developers to hone most of the commercial click fraud analysis programs out there, and they frequently mention various statistics including one popular estimate that most PPC advertisers suffer from a click fraud rate of around 30%. We have found that in all cases, the commercial analysis services provided data that was very difficult to understand, even for pros like us, and did not clearly illustrate the service's conclusions about various fraud rate values.
We just haven't seen it in our log files. We estimate that our NET click fraud rate is in the range of 1%-4% each month when the commercial services' results indicated somewhere in the range of 25%-70% (yep ... one vendor said 70% of our PPC clicks were fraudulent!) This is in a field that commands $50 per click and higher for the top terms.
We are keeping an eye on it, but we simply do not see that great of an impact as GGL and OVT are succeeding in keeping them off our bill.
I am definitely interested in hearing others' analyses, particularly from those who are doing their own.
If you are using one of the (few) commercial analysis programs, how are they working for you? Can you understand what their data is telling you? Does it correlate with what your log files tell you?