**** [google.com...] **** 18.104.22.168 | 22.214.171.124-static.reverse.softlayer.com 23/Jun/11 4:37am
**** [google.com...] **** 126.96.36.199 | host.degaa.net 23/Jun/11 4:04am
**** [google.com...] **** 188.8.131.52 | 66-199-245-35.reverse.ezzi.net 23/Jun/11 3:57am
**** [google.com...] **** 184.108.40.206 | 220.127.116.11-static.reverse.softlayer.com 23/Jun/11 3:52am
**** [google.com...] **** 18.104.22.168 | 216-169-108-193-cust-gw.reverse.ezzi.net 23/Jun/11 3:48am
**** [google.com...] **** 22.214.171.124 | 66-199-245-34.reverse.ezzi.net 23/Jun/11 3:48am
5 of the above IPs found on open anonymous proxy lists. 1 of them listed as a spam IP.
0 Adwords visits from any of the aforementioned nameservers prior to jump in daily click cost.
After competitor copied ad text and costs doubled, aforementioned nameservers begin to appear.
0 visits from aforementioned nameservers on non-Adwords visits before, during, or after these periods.
Average number of visits per day in last 6 months between 3AM and 5 AM: 0.4
Total ~cost incurred on aforementioned clicks: $18
Total ~cost on that day from 12AM to 6AM: $24
Total ~cost on previous day from 12AM to 11:59PM: $50
Adwords automated invalid click rate: 20%
Average invalid click rate after manual investigations: 30%
Adwords actual invalid click rate including similar types of aforementioned clicks: 60%
Adwords response to report on the aforementioned clicks:
"Just heard back from our click specialists. They have conducted another investigation on the account and they have assured me that there was no invalid activity on the account."
If you look at the above information, and would yourself pay $3/click for those visits with the expectation that it may convert because they are legitimate visits, you need to re-examine your understanding of statistical probability.
Further, I have concluded one holistic understanding from this experience:
Google is very aware that there are many industries where the invalid click rate is much higher than others. This is what its methodology is -
1) Click fraud cannot be controlled, only marginally minimised (10-15%).
2) Tracking the source of the click fraud to a competitor is generally impossible (1-5%).
3) Even if this is done, and the competitor's adwords account is removed, it does not prevent the competitor from continuing the click fraud.
Since it is not a readily preventable problem:
4) We must minimise the awareness of both the fraudulent clickers and advertisers to the nature of our succeptibility to click fraud. We accomplish this by:
a) Overestimating the number of clicks we deem to be fraudulent. Therefore, include a small percentage of legitimate clicks (2-5%) as invalid clicks to both assert that we are catching click fraud to the advertiser, and also to offset some of the fraudulent clicks we do actually miss.
b) Provide credits to advertisers if they periodically request it with even mediocre proof. Such credits should not exceed 30% of the total number of clicks received regardless of the number of probable invalid clicks. Anything higher could threaten the viability of Adwords as a reliable advertising medium if publicized.
c) Maintain an invisible barrier between invalid clicks and valid ones by hiding specifics to prevent both advertisers and fraudulent clickers from determining the actual percentage of invalid or valid clicks.
d) Ensure that PR continuously presses the evolution of our click fraud catching technology.
By minimising the global awareness of Google Adwords succeptibility to click fraud, it encourages advertisers to advertise and discourages fraudulent clickers from clicking. Though the technology required to commit click fraud is evolving at a faster pace than our ability to catch it, by utilising the "gray area" methods above, we can maintain sufficient profitability and customer confidence until we find another source of income with which to buoy Google.
Since the click fraud industry will inevitably increase, our efforts are aimed at minimising its growth rather than catching it.