Yesterday, I had a conversation with Nick Fox of Google about the most recent quality score changes. I posted the extremely interesting question and answer session at Search Engine Land last night. I was honestly shocked by one of his answers, shocked!
When I asked how Google changed the process of normalizing the CTR scores for ad position, he explained the old way versus the new way. In the past, Google took a lower ranking ad and moved it in a higher position to test to see how that ad's CTR would change. Any clicks on that ad were not charged to the advertiser, plus this was done on a very very small percentage of searches. It was the way Google could get a good metric for what the CTR would be for that ad, in the same ad position.
I was shocked by that, just because it sounds totally weird. It makes sense, don't get me wrong. But I just assumed that Google used, what it is using today, a statistical model to adjust those numbers. Now, they switched to a statistical model to adjust the ad's CTR for normalizing that factor within the quality score.
In any event, you probably want to read the whole discussion at Search Engine Land. I personally felt that specific tidbit was fairly enlightening.
Forum discussion at Sphinn.