Welcome to the Q&A Forum

The problem is that there is little to no content on these pages, aside from the the "boilerplate" stuff - header, navigation, sidebar, footer, etc - which appears on every page. What that means to a crawler/bot, is that these pages look very similar. If there aren't plenty of unique words on the page, there's always a chance that the page will look like duplicate or thin content. How many words is unique? Hard to say, but a general rule of thumb has always been 150-200 words for me. It's possible that the number could vary depending on the ratio of boilerplate to unique text, though these days Google is pretty good at distinguishing boilerplate from content. The issue is that thin content can be a problem in a post-Panda world, although on the scale you're looking at Google can most likely deal with it just fine.. For more info, to help you make an informed judgement,  I recommend this [great 'thin content' article from Dr Pete](. http://www.seomoz.org/blog/fat-pandas-and-thin-content). Could you add some descriptive text on the client directory pages, and redirect the empty articles somewhere else (or fill them with content?) Another option for the client pages could be to consolidate (and redirect) all of the categories into one client page - maybe using Javascript to hide/show each category on request, to keep things tidy.

Having rel=canonical is recommended for all pages of your website, even if it's the original version. It's because for example if your url is: http://domain.com/url and you have that set as canonical on the page, and someone references to your page as http://domain.com/Url http://domain.com/URL http://domain.com/url?ref=feed The canonical URL will still be http://domain.com/url and hence won't count as duplicate pages.

As another data point, we are still including 2-3 keywords in title tags and have not suffered for it yet. We often do something like KW1, KW2 & KW3 | Brand or KW1 - KW2 | Brand depending on whether the keywords are very close variations or somewhat dissimilar terms.