Rel Canonical Tags and Ranking Factors

[triggerfinger]

I'm curious if anyone has had any experience with this.

Here is the situation: I have user profile pages that are linked to from many different category pages across the site. Originally I only linked to these profile pages from a single user sitemap, and nofollowed links from the category pages (because of parameters). I decided that I could now use the handy rel=canon tag to consolidate link juice, insure no dup urls were indexed and hopefully improve ranking to these pages and all would be right with the world.

Not quite. So far I'm seeing a significant decline in organic referrals. Seems like the situation is the exact opposite of what was intended (and if I may, outlined by Matt Cutts in his explanation of rel=canon).

A couple factors that might be at play:

The content potentially changes with the parameters. This might have shaken up the relevancy of the pages, but again, this is what rel=canon is intended to address.

Over-optimization - The profiles went from have 1 link to potentially hundreds. Could they have gotten hammered with some form of over-optimization penalty? This seems ridiculous, I mean, its my site and I'll link if I want to. If you aren't going to give me better ranking for increased internal links, then fine, but why would you decrease the ranking?

All in all I'm very disspointed that all the talk of using rel=canon to consolidate link equity didn't turn out to be true, and utilizing the canonical link element ended up hurting ranking. If it doesn't pick up I'll have to return to the decidedly sub-optimal implementation.

It begs the question, do internal links even matter anymore?

[deadsea]

I have worked on large sites with lots of user profile pages. In my experience, these types of pages can be SEO sinkholes.

1. Google is very wary of them because users usually have lots of control over them with little editorial oversight. Google has been burned with so many web spam links coming out of user profile pages.
   
2. If they end up ranking for anything, the user experience of the searcher is very poor. They are not good landing pages.
   
3. There can be a very large number of them on your site leading to wasted crawl time from bots.
   
4. They are linked to so often that a significant amount of a sites internal pagerank can be funneled into them. Often there are few enough links on them that not much pagerank flows back out.
   

Several years ago I experimented with pagerank sculpting around user profile pages. There really isn't a lot you can do. From an SEO perspective, the best thing to do is not have the pages on your site at all. If you must have them, make it so they don't get crawled. Robots.txt is good for them.

It also appears that the measures that Google implemented to combat pagerank sculpting seem to kick in on these types of pages. I have some pretty good guesses as to the algorithms that Google uses to assign page rank via internal links. In the end, such low pagerank pages end up getting discounted by the algorithm. It doesn't matter if you have them or not. Google ignores them for all intents and purposes.

[brinked]

I noindex, nofollow all user pages

They simply have no value, most of them have little content, they get 0 traffic anyway and they simply dilute your site especially as it grows with a larger member base.

I dont even allow access to the user pages to non registered users.

[triggerfinger]

These aren't your grandmothers user profile pages. They are editorial reviewed, have content, rank, drive traffic. The hope was to let them be linked to from category pages to get them some more link juice.

Moving away from the user profile page debate, has anyone had experience with large scale Rel Canon and its affect on ranking?

Thanks

[deadsea]

I haven't used rel canonical for pagerank sculpting but I have done lots of pagerank sculpting using redirects, link removal, nofollow and such. Six years ago such techniques were very effective. I started getting diminishing returns three or four years ago. It got to the point where I could change 20% of the internal links and see very little change in rankings.

I would suggest that "conserving pagerank" is a dead concept. I believe that Google does this for you. For all sites, it scoops up lost pagerank and reapplies it to your important pages.

"consolidating link equity" is similarly dead. It appears to me that Google apportions juice within your site for you in such a way that sculpting doesn't effect it much.

Google created these algorithms to boost sites that don't pagerank sculpt to the same level as sites that do.

Canonical tags can still be useful for avoiding duplicate content issues that cause crawling nightmares and dup content penalties. But as far as pagerank sculpting, not so much anymore.

The only way that I have found to use pagerank sculpting recently is to tell Googlebot which pages on your site are more important than others. A page with 10 internal links from prominent places on your site will get crawled more often and rank better than a page with one internal link. So point your best links to your best pages. But it doesn't appear that you can use pagerank sculpting to boost your site as a whole compared to other sites.

[triggerfinger]

"The only way that I have found to use pagerank sculpting recently is to tell Googlebot which pages on your site are more important than others. A page with 10 internal links from prominent places on your site will get crawled more often and rank better than a page with one internal link."

That is what I thought as well. Previous to this change those pages had 1 internal link. Now that they have dozens or hundreds, I see a 30% drop in traffic. Doesn't make much sense to me.

[deadsea]

Were the pages that got canonicalized actually duplicating each other (same titles and overlapping if not duplicate content)?

How many page are we talking about here?
# pages that now have canonical pointing out:
# pages that now have some other page canonical to them:
Total # pages on site:

[triggerfinger]

To your question deadsea:

Same title, overlapping if not duplicate content (content changes depends on where the user comes from on the site)

About 14k canonical profile pages (not quite sure the scale of non-canonical versions, all have parameters)

Something like 500k pages on the site.

I'm going to try to nofollow the links from a large portion of the sites pages, but not all. I just don't see how 1 link from an HTML sitemap is more optimal then dozens/hundreds of links from targeted category pages.

Thanks as always