I've taken an in-depth look at the stats being provided by GWT(google webmaster tools) for my site going as far back as it will allow and comparing those with my own logs and traffic stats and I've found some rather revealing trends.
Assume my site is about widgets and it has a section for widget help. It also has many other sections and even a forum, an image gallery etc. Over time GWT is telling me that Google began thinking my site was less and less about widgets and more about widget help. Traffic steadily declined on any page with non widget help content and it steadily increased on any page with widget help content.
The current state of what GWT is telling me is that my site is not so much about widgets but about widget help, which is wrong. The site is very much about widgets which leads me to take a broader look at everything to try and find perspective.
The pages which are most helpful to first time visitors are indeed the help pages though they are not the most trafficked or even most popular on the site. These pages, however, provide the most value for Google!
Basically it looks like Google finds that linking to my widget help pages serves the needs of their users best, even if it doesn't reflect the core of my site accurately. The non help pages used to receive the bulk of the traffic from Google and it's that transition that I find revealing.
What can be taken from this? I'm not entirely sure but building pages in a way that is much less editorial and much more functionally helpful seems like a good idea.
Ok, first of all, even saying your site is "very much about widgets" is pretty broad.
Is it about buying widgets?
Is it about repairing widgets?
Is it about the history of widgets?
Is it about where widgets are located?
Is it about the applications of these widgets?
Is it about widget law?
You get what I'm saying here I'm sure. I do a lot of PPC for clients, and the #1 issue I run into is that the site itself, the landing pages and the perceived converting keywords (before I take over) are always way too broad. You have to drill down to the actual user intent. Whatever aspect of widgets that your site is about, that's where you have to place the emphasis so your users (and Google) know what it's really about.
Maybe what you're seeing is side effects of Panda and other changes. Traffic that you used to receive, looking at your widget pages, is now going to Amazon (example). Of course when people buy these widgets, and need help with those widgets, they still come to your site. Perhaps not an increase in help traffic, but a decrease in widget traffic? You haven't mentioned your overall traffic trend during your observation period.