I found a Usenix paper from 1998 on the subject of hit shaving, which is a type of click fraud where the advertiser underreports the number of times the ad publisher has referred traffic to their site (ie. the number of times the clickthrough was actually recorded at the advertiser's site). They describe a couple of methods for minimizing discrepancies that, as far as I know, are not currently used. One involves digitally signing all the pertinent data from the referred traffic (imagine signing everything that shows up in the web server log per visit of referred traffic) and sending it back to the ad publisher. The other involves the advertiser computing the hash of a large number k times (k is some maximum number of referrals), and for the ith referral, the advertiser sends back to the publisher i and the hash applied k - i times.
The paper cites drawbacks with both approaches. The amount of trust and overhead required to maintain consistent numbers may or may not be cost-effective, depending upon the resources available at both sites. For something like AdSense, where the publisher (the participating web site) takes a cut of the proceeds paid by the advertiser, it adds extra trust and overhead requirements as well, some of which the publisher may not understand or be able to carry out. At any rate, while this technique might (if implemented carefully) ease some tensions between publishers and advertisers when discrepancies occur, I don't think it resolves enough discrepancies that it would be adopted as a standard. Hypothetically speaking, if this is the best you could do in terms of limiting discrepancies, it doesn't change my mind that PPC is still a poor business model.
Since the paper was published in 1998, I think it is solid evidence that click fraud was a well-understood problem at the time PPC was implemented. That being the case, the plaintiffs in the lawsuit against the search engines can argue that the search engines were aware of the risks involved (or should have been). I don't think the search engines are guilty of collusion, but I could see the plaintiffs successfully arguing that many, perhaps most customers experienced some click fraud over the course of their campaigns. If these customers didn't receive discounts or refunds, the plaintiffs could argue they are entitled to them.
The paper cites drawbacks with both approaches. The amount of trust and overhead required to maintain consistent numbers may or may not be cost-effective, depending upon the resources available at both sites. For something like AdSense, where the publisher (the participating web site) takes a cut of the proceeds paid by the advertiser, it adds extra trust and overhead requirements as well, some of which the publisher may not understand or be able to carry out. At any rate, while this technique might (if implemented carefully) ease some tensions between publishers and advertisers when discrepancies occur, I don't think it resolves enough discrepancies that it would be adopted as a standard. Hypothetically speaking, if this is the best you could do in terms of limiting discrepancies, it doesn't change my mind that PPC is still a poor business model.
Since the paper was published in 1998, I think it is solid evidence that click fraud was a well-understood problem at the time PPC was implemented. That being the case, the plaintiffs in the lawsuit against the search engines can argue that the search engines were aware of the risks involved (or should have been). I don't think the search engines are guilty of collusion, but I could see the plaintiffs successfully arguing that many, perhaps most customers experienced some click fraud over the course of their campaigns. If these customers didn't receive discounts or refunds, the plaintiffs could argue they are entitled to them.
- Current Mood:
thoughtful
Profile
gregbo- gregbo
- gregbo's homepage
