It's time to update everybody, and it's good news.
We had a small jump in referrals on Monday 20 August, and over the last 2 weeks traffic has been steadily increasing, with yesterday at 90% of where we were before panda 3.5. Fingers crossed that it continues.
Overall we bottomed out in early June at 45% down (there was something that hit out site mid June, around 19 June that nobody has talked about). We did a total of 4 things to the site, though the 4th was only rolled out on the 18 August, so too soon I think to have had any influence on the recovery.
1. a couple of days after my last post here I did what I asked about in the OP, all moved to a subdomain and noindexed, including the jump page. There was a bit of a hiccup with the 404 page (details here [webmasterworld.com
...] but in a couple of weeks g had got it's teeth into them and within 4 weeks a large number of the PDFs were dropping out, all done in 6 wks more or less, even though there are still a couple showing up in a site search.
2. Revamped the about us section, with completely new copy for the category and 2 of the 10 subcat pages, including some industry membership logos, mentions, and all written without any thought at all to SEO or KWs. Added lots of facts and figures, such as dates, employees etc. The worst hit page was in this section FYI, though it wasn't one of the re-written pages.
3. Improved the call to actions on our money pages, which are in 2 categories, toning down lots of keyword work. It wasn't excessive at all, maybe 2-4 mentions of the main 3 word KW in 500-1500 words and another 3-4 related ngrams in the text. We cut this back to 2 kws plus trimmed back ngrams and ‘supporting’ kws from 3 or 4 words to 1 word, ie 'great big widgets' to 'widgets'. About 10% also had their title trimmed back a couple of words, touching up the description meta and cutting back the KW meta (yes, I know) from 6-7 to 3 core KWs. We did the same also to all other pages, including those hit hardest by panda (this was all we did to those pages). They now read much better, and there was also perhaps a dozen sentences trimmed out of the text that were superfluous and paragraphs broken up.
4. Lastly, we add an extra navigation option - as well as the main menu, at the bottom of each page there's a 'next' option, and we added also a 'back' option to the previous page in order. Our thinking was that people felt pulled through the page with few choices, and we're effectively doubling their options once they've read the entire page (we're normally at around 3 mins on average, 3 pages and 55% bounce). It seemed to be a point where we losing visitors. Also, these next and back options were configured to lead visitors up the silo, so after reading all 5 sub-sub-category pages, the last next option leads 'up' to the sub-cat page, after reading all 5 of those, up to the cat page, etc. and it's noticeably boosted the number of people exploring our site more widely. It’s also worth noting that we toned down the internal anchor text in the ‘next’ link, which were very kw focus, to shorter and sweeter text anchors, with the reasoning that after reading a page of widgets it’s logical to say ‘what about big fat ones’ instead of ‘big fat widgety widgets’
I don't think point 4 actually helped with the recovery though as it was rolled out too late and only to half of our site so far (the English version only, our site is bilingual) but might explain the ramp up in recovery over these past 2 weeks while g is figuring out how the internal links have changed.
Does anything in my description sound familiar? I hope it helps people out, and gives a bit of encouragement at least!