Even experienced search marketers can sometimes jump to wrong conclusions. Noted search marketer Andy Beal thought he had a scoop earlier this week when he did a little rough calculation on a presentation made by Google click fraud point person Shuman Ghosemajumder and pegged the actual occurrence of click fraud at 2% on Google. There was actually a little miscommunication between Beal and Ghosemajumder (since corrected on Andy's blog). I chatted with Ghosemajumder this week and here's Google's side of the story, largely ignored by the mainstream press.

Where Do These Numbers Come From?

BusinessWeek's article said "most academics and consultants who study online advertising" agree with the 10% to 15% number. Yet there has been no independent study done with reliable methodology to accurately scope the size of the issue. The study most often cited is a particularly damning one done by Outsell in May of 2006. In the study, 407 companies were asked what percentage of their search buy they believed to be fraudulent. They then averaged the responses and extrapolated it across the industry. Many of these advertisers weren't even tracking ROI, definitely a prerequisite for accurate identification of actual fraudulent behavior. As Ghosemajumder pointed out, "it's like asking a random group of people what they estimate the average salary in the U.S. to be, when they have no numbers to judge it on, and they don't even know what their own salary is." Yet, this is the number that seems to be accepted as fact by reporters determined to blow the issue into cover story status.

What's Fraud, and What's Attempted Fraud?

One fact that seems to be easily overlooked is what actually qualifies as click fraud. Fraud is only perpetrated when damage is done--in this case, if money passes hands. If no money changes hands, it's attempted fraud. Yet this simple distinction seems to be overlooked by many "investigators" into the question of click fraud. Everything tends to be included in the same bucket, usually accompanied by a whopping percentage designed to scare the hell out of online advertisers.

The 2% number quoted on marketingpilgrim.com came from Andy Beal, not from Google. It was computed by looking at the relative size of some graphics on a slide deck that was prepared to show Google's click fraud filtering systems.

Google has coined the term "invalid clicks" to refer to all those that advertisers are not charged for. This category also includes more benign examples, such as multiple clicks on the same ad that can happen when a visitor "pogo sticks," or clicks on an ad, hits the back button, and then clicks through on the ad again. Ghosemajumder does confirm that the number of invalid clicks represents a "single digit" percentage of all clicks across the network,

The "vast majority" of these clicks are proactively filtered out by Google in real time before any money passes hands, he says. It's as if the clicks didn't happen. The advertisers don't pay, and the publisher where the click originated doesn't get paid. The invalid clicks that slip through the real time filter then go for offline analysis, primarily focused on the AdSense network. Advertisers here are affected, but get refunds from Google without their having to take any action. In this case, Google does have a procedure for going back to the sites where the clicks originated. If anyone is out of pocket for these clicks, it's Google, not the advertiser.

Now we get to the 2% number. It refers to the clicks that make it through the proactive filters that the advertiser has to bring to Google's attention. The official word from Google is that this number is a "negligible percentage" of the total number of invalid clicks. My sense is that it's probably much less than 2%. Remember, this isn't a negligible percentage of all clicks, but a negligible percentage of "invalid" clicks, which in turn is less than 10% of all the clicks happening on Google.

The Impact in Dollars and Cents

So, let's talk about actual fraud, where the advertiser is the one out of pocket. Let's assume there is an advertiser with a $100,000-per- month budget. Let's further assume that the clicks this advertiser receives are representative of the total Google network.

Using the assumed 9% number as the number of invalid clicks, this means about $9000 of the budget falls into this category. From this, the "vast majority" are filtered out in real time, so there is no impact to the advertiser. A smaller percentage is refunded to the client without its having to take any action. Finally, there's the percentage that slips through the proactive filters. Even if we go with 2%, that would make the amount that would impact the advertiser $180. If you're doing your math, that's 0.18% of the total monthly spend, a far cry from 10% to 15%.

But It's Not that Simple

These are the estimates from Google, which has invested heavily in fighting click fraud. The same diligence in policing click fraud is probably not present in all advertising networks. Click fraud is definitely more prevalent in some sectors and on some networks than others. Finally, everyone acknowledges that we don't know what we don't know. If click fraud goes undetected through Google's filters and the advertiser never challenges it, it won't be identified. Google uses the ROI and conversion data that some of its advertisers share with it as an overall indicator of click fraud activity throughout its network. Its executives feel confident that there's very little slipping through all of these cracks.

Yes, this is Google's side of the story, but as the mainstream press seems to be more interested in focusing on a couple of egregious cases rather than providing a realistic picture of the issue across the entire network, I think it's important to pass it along. In the absence of real numbers for the short term, shouldn't you at least balance the numbers being touted by the press with those coming from the people fighting click fraud on a daily basis?

This commentary is insightful. I recommend it to others.