24
2008
Oct

Click Fraud — The topic nobody wants to talk about!

mad prisonerWe had a small victory in the war on click fraud this month. It feels good to win one but this is only one battle in a very long global war. The problem with click fraud is that it is tough to detect and prove. Any reasonable professional in this field knows that there is a degree of click fraud in the numbers but we must keep it contained.

The process of reporting suspect activity was surprisingly simple although it took a little determination to get the job done. The first report was responded to by a 1st level support person that told us that there was no sign of fraud. We followed this with a respectful “we disagree response” along with more details and documentation on why we thought this was fraud. A few days later we got a response from someone higher in the support chain confirming our position and issuing credit to the account. Google to their credit immediately shut down the Adsense account of the perpetrator and I trust they took other actions.

Google will not disclose their methods of detecting fraud and I agree that keeping the deep details secret is a good idea. However I think some more reputation management is in order for this issue. Maybe they could include some system wide data to the placement report. On our wish list of data to help us detect fraud would include a placement report with the site’s CTR, Bounce Rate, and Conversion Rate. Ebay has built a self-maintained system of ranking sellers and maybe something like that could be implemented in the Content network. I think advertisers would be willing to pay more if they knew that the sites being served to were higher quality.

There is no doubt that if the fraud had fit one of the pattern signatures within their system it would have been immediately resolved. We know this because every month we see credits issued to our clients from Google’s automated detection tools. The problem is that automation cannot find things it was not designed to find. It takes people to see new patterns and those with the most interest in lowering the click fraud level are the most likely to see questionable situations. This situation is a good example because the fraud did not create a known pattern signature. So the big question is how do we as advertisers report new tricks in a way that Google can prioritize and resolve them?

The reason it surfaced in our systems is that we look at traffic with a different perspective. Let’s face it, finding click fraud is not completely in Google’s best interests. After all closing this down reduces their revenue, but Google also realizes that its entire business is based on the trust that the traffic is of reasonable quality. We believe that Google is doing what it can but the violators are getting smarter and detecting fraud is hard. The new challenge is that violators are starting to engineer the user interface to force bad clicks and violate the system.

In this case what they designed was a floating window that was positioned over the ads. The close button was programmed to close with a mouse-over rather than an on-click event. For people with good mouse skills the click is happening almost as quickly as the mouse over event so they click on the close but the window closes before their click resulting in the click traveling to the next layer in the browser, which is where the ad was located. Since this was done by a smart person I would assume that they controlled the CTR by managing the ratio of mouse-over to on-click events. After all if they had too high of a CTR it would trigger a flag in Google’s systems.

While click fraud is a very complex topic we all have to keep our eyes open and report anything that looks questionable.