expertSEM

In a recent post on the ClickForensics site, the claim was made that in Q4, click fraud rates climbed to 16.6% of all search traffic. This would represent a two percentage point increase in fraud from the previous year, if it were true. The thing is, beyond my assertion that click fraud is wildly overestimated; there seems to be some funny math going on.

Bear with me, I never was a math specialist, but take a look at these figures.

Now, the highest percentages of click fraud are coming from India (4.3%), Germany (3.9%), and Democratic Korea (3.7%). By the law of averages, if those were your highest average rates of fraud, and you had a few low rates, 1%, 2.2%, etc, you’d come up with an average somewhere below the highest rate of 4.3%. This is not a subjective argument, it’s a mathematical law.

So how did they come up with 16.6% click fraud? I absolutely don’t know, so I put it to you dear readers, how did they come to this astonishingly high figure?

They don’t really divulge much on the site. Anyone who knows Tom Cuthbert, CEO of Click Forensics is strongly encouraged to ask him.

As I’ve held for years, I think click fraud is between 1% – 5%, depending on your industry and if you are on content marketing networks. My opinion carries some weight on this issue, I have personally investigated several dozen suspected click fraud instances at the request of clients. Now, any statistician will correctly point out that several dozen is not statistically significant. But if it’s true, that click fraud is pervasive, where are the anecdotes? Where are the fraud billionaires? Search is now a several hundred billion dollar a year industry; if 16.6% of hundreds of billions is being diverted illicitly, where are the mansions?

You always have to look for motive, and only those who will gain by 1) driving you out of business, 2) clicking on your ad, would have an incentive to perpetrate click fraud. If you don’t use content, you cut out number 2 entirely, and number 1 is going to be rare, and with new tools you can limit it once identified.

My last point. I don’t think I’ve ever seen an unbiased estimate on click fraud from someone I would trust on the matter. ClickForensics wants to sell you a fancy detection package:

But simply doing a reverse IP lookup off your server logs for any source with more than a few clicks will tell you everything you need to know. Don’t trust these guys, do your own work, it’s not rocketry. I’d be interested if anyone has any anecdotes suggesting anything near the 16.6% that Click Forensics wants to scare us into believing.