Large site with faceted navigation using rel=canonical, but Google still has issues

[PeteGregory]

There are a large number of urls like this because of the way the faceted navigation works and I have considered no index, but somewhat concerned as we do get links to some of these urls and would like to maintain some of that link juice.   The warning shows up in Google Webmaster tools when Googlebot finds a large number of urls.  The rest of the message reads like this:

"Googlebot encountered extremely large numbers of links on your site. This may indicate a problem with your site's URL structure. Googlebot may unnecessarily be crawling a large number of distinct URLs that point to identical or similar content, or crawling parts of your site that are not intended to be crawled by Googlebot. As a result Googlebot may consume much more bandwidth than necessary, or may be unable to completely index all of the content on your site."

rel canonical should fix this, but apparently it is not

[PeteGregory]

Yeah that is why I am not real excited about using robots.txt or even a no index in this instance.  They are not session ids, but more like:

www.example.com/catgeoryname/a,

www.example.com/catgeoryname/b

www.example.com/catgeoryname/c

etc

which would show all products that start with those letters.  There are a lot of other filters too, such as color, size, etc, but the bottom line is I point all those back to just www.example.com/categoryname using rel canonical and am not understanding why it isn't working properly.

[AlanMosley]

OK, sorry I was thinking too many pages, not links.
using no-index will not stop PR flowing, the search engine will still follow the links.

[PeteGregory]

thx

[AlanMosley]

Because rel canonical does nothing more than give credit to teh chosen page and aviod duplicat content. it does not tell the SE to stop indexing or redirect. as far as finding the links it has no affect

[sichristie]

Did you find a solution for this? I have exactly the same issue and have implemented the rel canonical in exactly the same way.

The issue you are trying to address is improving crawl bandwidth/equity by not letting Google crawl these faceted pages.

I am thinking of Ajax loading in these pages to the parent category page and/or adding nofollow to the links. But the pages have already been indexed, so I wonder if nofollow will have any effect.

Have you had any progress? Any further ideas?

[Dr-Pete]

I'm working on a couple of cases now, and it is extremely tricky. Google often doesn't re-crawl/re-cache deeper pages for weeks or months, so getting the canonical to work can be a long process. Still, it is generally a very effective tag and can happen quickly.

I agree with others that Robots.txt isn't a good bet. It also tends to work badly with pages that are already indexed. It's good for keeping things out of the index (especially whole folders, for example), but once 1000s of pages are indexed, Robots.txt often won't clean them up.

Another option is META NOINDEX, but it depends on the nature of the facets.

A couple of things to check:

(1) Using site: with inurl:, monitor the faceted navigation pages in the Google index. Are the numbers gradually dropping? That's what you want to see - the GWT error may not update very often. Keep in mind that these numbers can be unreliable, so monitor them daily over a few weeks.

(2) Are there are other URLs you're missing? On a large, e-commerce site, it's entirely possibly this wasn't the only problem.

(3) Did you cut the crawl paths? A common problem is that people canonical, 301-redirect, or NOINDEX, but then nofollow or otherwise cut links to those duplicates. Sounds like a good idea, except that the canonical tag has to be crawled to work. I see this a lot, actually.

[sichristie]

Thanks Dr Pete,

Yes I've used meta no-index on pages that are simply not useful in any way shape or form for Google to find.

I would be hesitant noindexing my filters in question, but it sounds promising that you are backing the canonical approach and there is a latency on reporting. Our PA and DA is extremely high and we get crawled daily, so curious about your measurement tip (inurl) which is a good one!

Many thanks.

Simon

[Dr-Pete]

I should actually add that Google doesn't condone using rel-canonical back to the main search page or page 1. They allow canonical to a "View All" or a complex mix of rel-canonical and rel=prev/next. If you use rel-canonical on too many non-identical pages, they could ignore it (although I don't often find that to be true).

Vanessa Fox just did a write-up on Google's approach:

http://searchengineland.com/implementing-pagination-attributes-correctly-for-google-114970

I have to be honest, though - I'm not a fan of Google's approach. It's incredibly complicated, easy to screw up, doesn't seem to work in all cases, and doesn't work on Bing. This is a very complex issue and really depends on the site in question. Adam Audette did a good write-up:

http://searchengineland.com/five-step-strategy-for-solving-seo-pagination-problems-95494

[PeteGregory]

As for my situation it worked out quite nicely, I just wasn't patient enough.  After about 2 months the issue corrected itself for the most part and I was able to reduce about a million "waste" pages out of the index.  This is a very large site so losing a million pages in a handful of categories helped me gain in a whole lot of other areas and spread the crawler around to more places that were important for us.

I also spent some time doing some restructuring of internal linking from some of our more authoritative pages that I believe also assisted with this, but in my case rel="canonical" worked out pretty nicely.  Just took some time and patience.

[sichristie]

Yes that's a different situation. You're now talking about pagination, which quite rightly, canonicals to parent page is not to be used.

For faceted/filtered navigation it seems like canonical usage is indeed the right way to go about it, given Peter's experience just mentioned above, and the article you linked to that says, "...(in part because Google only indexes the content on the canonical page, so any content from the rest of the pages in the series would be ignored)."