Search Algorithm Release Process at Google
During the panel discussion at the Churchill Club Danny Sullivan asked Matt Cutts, Ben Gomes, and Amit Singhal about the recent algorithm changes including the famous Panda. Although the answers were not particularly informative they did give insight into Google’s quality algorithm release process.
According to Amit Singhal, once identified the problem is solved strictly by algorithm (Google still claims they do not use manual human intervention for much more than testing the quality of results and manual penalties in very few instances). Candidate algorithms are then implemented in code and tested using old query logs. There results are then analysed for quality. In the next stage small human control group is used on sandbox results which show A-B in random fashion. The data collected is used towards a limited live release where a relatively small number of users may see new algorithm in action. These results are further analysed and assessed for quality before making the algorithm permanent. It’s likely that these mini releases happen often and may be responsible for sometimes surprising search results many SEO professionals are seeing in SERPs.