It is interesting that this escalated to legal proceedings. Typically when an advertiser is overcharged for an ad buy the publisher will refund their money (if they cannot make-good through additional ad inventory). Perhaps this case is different because the advertiser may feel that Google intended to cheat them. Or perhaps the parties are disagreeing on the actual amount of fraudulent clicks. Is there really a fool proof system to determine fraudulent clicks? A general number could be found by comparing log file data to Google/Yahoo click numbers. However, this may not be 100% accurate.

Advertisers running pay-per-click search campaigns can begin to protect themselves by taking the following steps:

1. Know the click fraud policy of the search site where your campaign is running

* how do they define click fraud?
* how do they prevent click fraud?
* how do they handle instances where they cannot prevent click fraud?

2. Set-up your own unique tracking codes for your search campaigns
3. Compare your tracking code data to that of the search engine
4. Look at unique user data, IP addresses, etc (anything that will help you to identify real user traffic)
5. Address discrepancies with the search engine

A few technologies have emerged to help advertisers address this problem but I have no direct experience using them. I welcome any other suggestions from the community.

Comments On This Entry

"A general number could be found by comparing log file data to Google/Yahoo click numbers."

If only it were that easy! That method really won't do the job. If you and are competitors, and my goal is to deplete your budget with extra clicks, your web server log files will match Google's perfectly. Problem is, the clicks won't be from bona fide users.

Your #3-#5 will detect some kinds of click fraud. But still not all. How to detect a sophisticated fraudster who uses distributed proxies (located around the country, perhaps on hacked user machines that have been infected with malware), who uses good software that generates realistic-looking HTTP headers and other HTTP traffic (perhaps instantiated IE). It's not an easy problem.

All that said, I remain surprised by how little the search engines have said about this problem and what they do to detect and stop it. Greater disclosure could help inspire greater confidence among their advertisers.

Posted by: ben Edelman at April 5, 2005 11:39 PM

Unfortunately you are correct Ben. A Savvy fraudster can disguise their efforts most of the time. While many companies are emerging to address this problem, it is unclear as to how effectively they perform.

One such company is http://www.whosclickingwho.com/
And in their FAQ section you will find this interesting blurb.

Don't the search engines track fraudulent clicks?
Sort of. The search engines do track fraudulent clicks, but from what we have seen their controls are not terribly strict. If someone clicks on your ad 5 times in 20 seconds, you probably won't be charged for all 5 clicks. BUT, if someone clicks on your ads once or twice per day for a month, our experience has shown that you will probably be charged for nearly every single click! We don't think this is right. Additionally, if someone is clicking across multiple search engines in order to hide their fraud, WhosClickingWho? will detect this! If you really want to know who is clicking your ads, you should be using WhosClickingWho?

Posted by: Peter Figueredo at April 6, 2005 08:29 AM
http://www.revenews.com/peterfigueredo/archives/000572.html