Google has agreed to pay $90M to settle an advertising click fraud case; evidently on grounds that it didn't catch the fraud. Google's defense had been that it does reimburse for fraudulent clicks, and the WSJ claims that the settlement apparently leaves that defense intact.
$90M is a lot even for me, but for Google, frankly, it's pocket change. Their latest quarterlies have them at almost $4B in cash & equivalents. That's down from $5.5B the prior quarter, but the difference has gone to short-term investment, not burn rate.
Google uses some strange proprietary formula based on the number of clicks, number of ads, priority, phase of the moon, and position of the owners' yachts in the latest test race. It was opaque to both the advertiser and to me, and the added layer of uncertaintly almost certainly makes for a less efficient market.
Personally, I think BlogAds has the better model. I never really understood the click rates except as a proxy for eyeballs, and nobody pays for any other kind of advertising that way. You don't see people charging for urban billboards by how many people changes busses just to hit the Macy's sales.
I've only had a couple of BlogAds, but it's easier for me, since I set a fixed rate up front for a set period time. You know, like a normal product, or normal advertising. In some sense, it's based on how many people see the ad, but in another sense, it's also based on the sales the advertiser sees rather than the visits. Market signals are much more direct this way.