Robert_Charlton - 7:18 am on Sep 6, 2013 (gmt 0)
adder - Great post. Yes, you need to align many events on the site with the algo and see what likely caused what. Not all might line up simply with algo changes, though.
In this case, I'm assuming that this is the same site that ichthyous described in this thread...
Keyword albums / tag clouds triggering Panda algo penalty?
At some point the number of indexed pages on my site started climbing from around 6,000 to 28,000 very quickly in December of 2012 and then started heading back down to where it started in May of 2013. During this time my site has lost approximately 70% of its traffic and it hasn't recovered no matter what I try.
As I assumed from the description of the problem, the site had a large number of essentially empty search-generated tag clouds, and I felt that, for a bunch of reasons, they'd be a likely Panda disaster.
I'm wondering how the dates, sections, and traffic drops line up with the presence and activity of the tag cloud software. In this present thread, you posted about dates...
...it pretty much aligns with panda updates in late 2012 and 2013.
I'm also wondering how the A and B sections affected align with the number of tag cloud pages accumulated in each.
The thought that crossed my mind when I read the opening post... completely guessing because I haven't seen any graphs or stats... is that, if the extra links "worked" for a while, conceivably the extra traffic that came into section A because of the links increased the number of tag clouds in that section (ie, of thin/shallow pages in the section likely to lead to user/Google dissatisfaction), and ultimately caused the Panda hit. This is speculation on my part based only on the little bit I know about the site, and there may be many other issues overlayed, but I thought it was an idea worth tossing out.
Keep in mind that there might be a lag in some dates as things need to work through the index.
PS: Yes, Panda does affect different sections of a site differently. One technique used early on to revive the best pages of a site was to shunt weak pages off to a separate subdomain.