De-indexing millions of pages - would this work?

[TalkInThePark]

Hi all,

We run an e-commerce site with a catalogue of around 5 million products.

Unfortunately, we have let Googlebot crawl and index tens of millions of search URLs, the majority of which are very thin of content or duplicates of other URLs. In short: we are in deep. Our bloated Google-index is hampering our real content to rank; Googlebot does not bother crawling our real content (product pages specifically) and hammers the life out of our servers.

Since having Googlebot crawl and de-index tens of millions of old URLs would probably take years (?), my plan is this:

1. 301 redirect all old SERP URLs to a new SERP URL.
2. If new URL should not be indexed, add meta robots noindex tag on new URL.
3. When it is evident that Google has indexed most "high quality" new URLs, robots.txt disallow crawling of old SERP URLs. Then directory style remove all old SERP URLs in GWT URL Removal Tool

• This would be an example of an old URL:
  www.site.com/cgi-bin/weirdapplicationname.cgi?word=bmw&what=1.2&how=2
• This would be an example of a new URL:
  www.site.com/search?q=bmw&category=cars&color=blue

I have to specific questions:

1. Would Google both de-index the old URL and not index the new URL after 301 redirecting the old URL to the new URL (which is noindexed) as described in point 2 above?
2. What risks are associated with removing tens of millions of URLs directory style in GWT URL Removal Tool? I have done this before but then I removed "only" some useless 50 000 "add to cart"-URLs.Google says themselves that you should not remove duplicate/thin content this way and that using this tool tools this way "may cause problems for your site".

And yes, these tens of millions of SERP URLs is a result of a faceted navigation/search function let loose all to long.
And no, we cannot wait for Googlebot to crawl all these millions of URLs in order to discover the 301. By then we would be out of business.

Best regards,
TalkInThePark

[TalkInThePark]

Thanks a lot, Tom. Time will tell...

Just one last thing:
what damage are you (and Google) thinking of when advising against removing URLs on a large scale through GWMT?

Personally, I think Google says so only because they want to keep as much information possible in their index.

[tomcraig86]

Thanks for the PM, I can now appreciate the problem a little more.

I think it's something that you should not rush. What you've done seems the best thing you can do for now.

Longer term, I'd look at your CMS options!

[TalkInThePark]

Yes, I have put a conditional meta robots "noindex" on all pages whose URL contains more than 2 GET elements. It is also present on URLs containing parameters of little or no SEO value (e.g. the "price" parameter).

Regarding the nofollow directive, my plan is to not put it in the head but on the individual links pointing to URLs that should not be indexed. If we happen to get a backlink to one of these noindexed pages, I want the link value to get passed on to listed product pages.

My big worrie is what should I do if this de-indexation process takes forever...

[tomcraig86]

If you could put a conditional meta tag in to the source code, that will show the nofollow tag if the URL contains more than 3 GET elements, then that might help?

You seem to have already thought hard about your options, and they sound ok. Let's just wait to see whether any Gurus are about to shout stop!

[TalkInThePark]

Thanks for answering that quickly, Tom!

We cannot robots.txt disallow all URLs. We get quite a lot of organic traffic to these URLs. In july, organic traffic landing on results pages gave us approximately $85 000 in revenue. Also, what is good to know is that pages resulting from searching and browsing share the same URL - the search phrase is treated as just another filtering parameter in the URL.

Keeping the same URL structure is part of my preferred, 2-step solution:

1. Meta Robots "noindex" unwanted results pages (the overwhelming majority)
2. When our Google index has shrunken enough, put rel=nofollow on internal links pointing to those results pages in order to prevent bots from crawling them.

I have actually implemented step 1 (as of yesterday). The solution I was describing in my original post is my last resort solution. I wanted to get a professional opinion on that one in order to know if I should rule it out or not.

Unfortunately, I cannot disclose our company name here (I have a feeling our competitors use Seomoz as well :)). But I'll send you some links in a private message.

[tomcraig86]

If I were you I'd keep the same URL structure. You're correct in thinking this won't be a quick fix.

First, use the robots.txt to disallow robots access to the search pages.

Don't remove all results just yet from GWT, this will be a long task and might damage your sites performance.

Could you provide some links to your site? I'll have a closer look.