On the Official Google Blog, Amit Singhal wrote about the Google Ranking. Rankings are broken down into three core parts, explains Singhal. First, the best relevant results are provided in what Google calls "no query left behind." Second, Google likes to keep it simple without compromising quality. Finally, Singhal says that there's no manual intervention. The rankings are powered by algorithms that weigh the behavior and linking patterns of people.
The ranking discussion has moved over to WebmasterWorld where one of Singhal's statements stood out at a reader. In response to the "no query left behind" principle," Singhal acknowledges that there may be "less than ideal results for any query" and that Google uses the knowledge of these results for future improvements.
The question that forum members have is "what are less than ideal results?" How does Google know what is and isn't ideal? One forum member speculates that user data may be involved in this. If a user doesn't click on the result at all, that may be an indicator that the result isn't ideal (on the other hand, their curiosity might be piqued and they might click anyway). It's also possible that Google's human evaluators look at the SERP results to find any anomalies.
Forum discussion continues at WebmasterWorld.