WebProNews

After Shuman Ghosemajumder posted a pair of blog entries attacking the click fraud assessments and methodologies of companies like Click Forensics, that company's CEO Tom Cuthbert and a couple of other executives took to the blogosphere to respond.

Believe it or not (and you might not believe it), Cuthbert and Ghosemajumder are not strangers peeking over the tops of opposing trenches, tossing mustard gas and grenades at each other in a virtual sense.

Tom noted in an email to WebProNews that he and Ghosemajumder had been in touch by phone last week, and that Shuman emailed him the morning his blog posts made it online. Given those communications and others previously, I asked Cuthbert about Ghosemajumder's concerns:

My only guess would be that he has not dug into our process. He joined our Network in October and could clearly see that our set up requires Google advertisers to utilize the click ID. I’m not sure I can answer your question but it is something I have wondered myself!

I have a great deal of personal respect for Shuman and all of us at Click Forensics admire Google’s success. We recognize how important it is that Google and the other providers work with third parties to mitigate the risks of click fraud. Shuman has said that on several occasions as well.

Tom and two other Click Forensics executives, Peter Norwood and Kevin Embree, joined the conversation by blogging their concerns about Shuman's points.

Norwood expressed his surprise at Shuman's post, saying, "The posting is so filled with inaccuracies that I can only assume it represents an individual’s view and not that of Google’s executive management." He goes on to note that the accusation that Click Forensics may rely on just a single metric, IP frequency, to determine click fraud is incorrect:

Google knows quite well that our heuristics are broad and deep and involve large numbers of attributes, not a single attribute like “IP frequency”. There are actually multiple classes of heuristics with each class consisting of many separate attributes:

Technical heuristics stem from information about the visitor’s IP, browser and system configuration. This information can tell us many things. For example, is the visitor coming from a location in Eastern Europe even though the ad is targeting US customers – or – is the visitor a Bot, an automated program designed to commit click fraud.

Behavioral heuristics include information about the visitor’s path – the series of clicks that follow the initial click on the ad. The amount of time visitors spend on a site, the specific path they take on the site, and eventually whether or not they make a purchase – all of these behaviors are input to our scoring algorithm.

Economic heuristics are some of our most confidential, but what we can say is that indications of fraud are related to the economic rewards of clicks – fraudsters are after money and that’s information we can use to our advantage.

Community heuristics include information we can gather from across search providers and channels – just as criminals can be identified by patterns even if they rob in different locations.

"One reality that advertisers understand all too well is that it is impossible for Google to police itself, because there is an inherent conflict of interest," said Norwood.

Embree made a careful statement about click fraud, which he likens to the crime of cattle rustling that persists to this day:

I'm now focused on the detection and mitigation of a similar problem - click fraud for online advertisers. Please notice I did not use the word solve.

Solving the problem may be too much to hope for, but increased cooperation from the search engines with third parties should make its mitigation much more effective. Let's hope the IAB's working group on click measurement can match this need with action soon.

---
Tag:

Add to Del.icio.us | Digg | Reddit | Furl

Bookmark WebProNews: