After carefully analyzing the analytics for a dozen sites, I am convinced that Google is analyzing time on site, in comparison to your competitors, to penalize sites that don't belong in top rankings. Googles Logic: Time on site is the ONLY "black hat proof" factor that they can analyze. If you are sitting 3rd for a keyword with a 15 second time on site, but the sites in 4th and 5th average 2 minutes for that SAME KEYWORD, they deserve higher rankings. How does google test time on site? Simple. They can get time on site from checking how long it takes a user to return back to the search result page. When you bounce off a website, you usually click the back button and click the next result in the listing, dont you? When a user finds the site they are looking, they take a long time to return back to google. GOOGLE HAS THIS DATA! My Proof: I own 14 websites across 4 niches. (Multiple sites rank for the same keywords). I build similar links to all the sites, you would expect them all to bounce in a similar manner. While some dropped, others held solid. I graphed correlation to ranking changes in Excel, and plotted a bunch of factors, time on site, bounce rate, PageRank, Domain age, and 118 others. I used Spearman correlation, and found that time on site has an extremely high correlation to ranking changes. How you can help: Check your analytics. Verify that the keywords with the least time on site are the ones that dropped the most. See if those keywords that have a long time on site stayed relatively solid, even if they had less backlinks.