A subdomains fix for Panda - Matt Cutts suggestion to HubPages

[willybfriendly]

In May, Edmondson wrote an email to Google engineers...and asked whether he should break up his site into “subdomains,”...In June, a top Google search engineer, Matt Cutts, wrote to Edmondson that he might want to try subdomains, among other things.

The HubPages subdomain testing began in late June and already has shown positive results. Edmondson’s own articles on HubPages, which saw a 50% drop in page views after Google’s Panda updates, have returned to pre-Panda levels in the first three weeks since he activated subdomains for himself and several other authors.

[blogs.wsj.com...]

[walkman]

I had a look at some of these subdomains (on that forum thread, just click on the name of someone who says they are participating) and at a glance they have all the structural problems that the original Pandalised hubpages had.

Alysa, care to explain?


Indyrank,
I'm done following Matt Cutts' advice or that of Susan Moskwa. They live in la-la land or something :) Here's Matt's pre-Panda advice on duplicate content: let's us crawl it, we'll sort it our for you. Panda sure did it sort it [youtube.com...]

[trakkerguy]

Umm... why not just get rid of the material that is NOT "high quality" content?

Then you won't have to worry about "getting relief" for the high quality content by moving it onto individual subdomains or other shenanigans.

Well, one reason would be that.... Almost noone has recovered simply from removing "bad" content.

This way, they don't have the long wait. And get to show advertisers they ARE coming back, and soon.

Also, you then are stuck trying to draw the line on what is good and what is bad, and then having a revolt from unhappy writers who get unfairly booted.

[Whitey]

I'm late in here, but a sub domain strategy sounds as though it's not for everyone. It could give temporary relief but badly backfire for a large number of reasons.

Again, repeating what's already been said, nothing new here about why the HubPage content is visible and ranking.

noone has recovered simply from removing "bad" content.

Some claim to. But effectively it's a complete and costly site rebuild for some. Some won't make it. The alternative is to choose to do nothing and move on to other things.

Sites that fix , will be back - when, who knows. This is a penalty with a manual on/off switch like PR in a sense. for those effected, let's hope it's not years.

[ken_b]

Umm... why not just get rid of the material that is NOT "high quality" content?

Money.

Even after Panda 1,2, 2.1 blah, blah, blah. If you have enough more "low quality" content than high quality content, it might be better money wise to move the high quality stuff and leave the low quality stuff alone. Easier too.

[indyank]

Alyssas, Would you care to explain the structural problems that you see?

[elsewhen]

Umm... why not just get rid of the material that is NOT "high quality" content?

because no one really knows how google defines high quality content. by segmenting the content, hp seem to cause panda to assess each bucket independently.

[AlyssaS]

walkman and indyrank - click on one of those subdomains. Look at the homepages, and look at the category pages.

Then compare to a blogspot.

With hubpages - the structure with the duplicate versions of the category pages (hot, latest and best) on the primary domain, the links between the hubs back to the category pages on the primary domain and vise versa, the links between the good authors and the bad through fanning, the aggresive overlinking through the hubtivity, the amount of links on the subdomain home page - all are elements that I think hurt hubpages. And they've been retained in this new re-incarnation.

We know link structure is an issue from Matt Cutts in his last livechat.

[searchengineland.com...]

They haven't really resolved link structure, apart from moving parts of the site onto subdomains.


I guess at the end of the day it all depends on the relationship between domains and sub-domains.

My understanding is that a bad subdomain doesn't usually affect a good primary domain, but a bad primary domain can infect all the subdomains.

So with blogger (and wordpress.com, typepad.com and others), the primary domain is clean and pristine. Some of the subdomains are bad, but they are discrete and therefore don't affect the primary domain (plus they work to get take down the spammy stuff).

Think of the nightmare for G if a single bad subdomain tanked the entire wordpress.com site. So they must have something built into the algo determining the relationship between subdomains and domains - and they probably use interlinking to determine the relationship - and look how they are treating their own blogspot structure - the subdomains link back once to the domain within an iframe in the nav bar at the top, and the domain does not any pass link juice to the subdomains. That's no relationship at all!

With hubpages, for the reasons detailed above, the primary domain has gone bad, and it's key elements haven't changed, and the subdomains are heavily linked back to the primary domain and vice versa.

I think they'll rank well initially, but when the algo catches up with them, it will conclude that they are not discrete elements but part of a single site, because of the heavy interlinking. If I'm wrong, mea culpa.

If there are others with more experience of the relationship between primary domains and subdomains, please chip in!

[trakkerguy]

Very nice analysis AlyssaS. Will be interesting to see if it plays out that way.

I think they would still have a good chance though, since Panda has more to do with quality of a site (or subdomain), rather than linking, imo.

If you start with a new domain,and you have good quality content, would a few links back and forth to some low quality (but not penalized) site put you in Panda land? I am guessing not.

I agree a bad domain can infect a subdomain, especially if linking back and forth. But think that is more likely for a domain that has a traditional penalty for naughty behaviour.

I am guessing Panda-itis is not contagious disease that you pass on to everyone you get too intimate with - but more like a genetic disorder that you can't pass on to others.

[walkman]

Think of the nightmare for G if a single bad subdomain tanked the entire wordpress.com site. So they must have something built into the algo determining the relationship between subdomains and domains - and they probably use interlinking to determine the relationship - and look how they are treating their own blogspot structure - the subdomains link back once to the domain within an iframe in the nav bar at the top, and the domain does not any pass link juice to the subdomains. That's no relationship at all!

Your explanation makes sense but I guess we'll soon see if G added it or not. It depends if Panda really want to screw cross-linking 'domains' as part of the cluster.

Leaving Hubpages aside this can very easily screw innocent sites. Each folder /subdomain can have a different section or pages with a different purpose and here comes some 'Panda' that crushes you. And not even a private email with suggestions from Matt Cutts. A few days ago I no-indexed my most visited page, that one probably has even has more links than my home but it's different. You visit it, look around and leave. I am seeing if it skewed my site or not. So much for focus on the content /user, huh :) ?

_{Footnote: Edmondson from Hubpages wrote a scathing editorial on Google how they are promoting Youtube and Blogspot while penalizing /Pandalizing others, er competitors. Specifically mentioned how many junk vids youtube had, how the moved up etc. Not to revisit it but many of us have noticed Blogspot junk go on top lately. I wondered about Matt's advice and especially the "among other things" part. I'd like to know too, I got a family to feed as well.}

[dataguy]

I'm having second thoughts on this. Maybe this is a legitimate strategy.

I know that historically Google has treated isolated subdomains largely as individual sites, more so that I thought was proper. Maybe this isn't a hack to bypass Panda, maybe this is actually what Google wants us to do.

For whatever reason, Google has had a hard time determining the author of content when mixed with other authors' content. This can explain why they announced their use of the rel="author" attribute a few weeks ago. It seemed like a stupid feature to most website owners, but apparently Google really is having a hard time figuring out who the author is on multiple-author sites.

Then consider blogspot. Maybe this is the model that Google is most comfortable with. We know that more that 80% of blogspot is spam, but Google doesn't devalue every blog on blogspot.com. They are kinda like silos, and Google obviously recognizes them independently. As mentioned above, wordpress is another good example of this.

As the owner of a website with thousands of users submitting content, my hardest job is separating the spam from the legitimate content. If I separate the authors into their own subdomains, my job becomes much easier. Hopefully Google will just devalue the offending content and leave the good content alone.

I've already started working on building tests for this.

[Freedom]

Or, instead of subdomains, move your ? content to a new domain.

After Panda, I realized I had to have back-up domains indexed and ready to go, post-sandbox, as a proactive preventive measure against google.

If Google ganks me, then switch it all over. - I realize this may not work in many cases, but it's an idea I am seriously considering.

[netmeg]

Considering how much play this is getting in various industry blogs and discussions, even if it *does* work, I suspect it won't work very long.

When you find something that works DON'T TALK ABOUT IT. #fightclub

[dataguy]

I'm leery of anything Google says or does right now, but if Google wants us to use this technique, then maybe they won't smack it down.

Moving content to sub-domains is what MC has been saying all along. Doing this isn't something to trick Google, it's what they're telling us to do.

The chances of it working may be less than 50/50, but if it does work, for many of us it's worth taking the chance.

There should be some differentiation between UGC sites and otherwise, though. Sub-domains make a lot more sense in a UGC site where you can split things up by user.

[potentialgeek]

In June, a top Google search engineer, Matt Cutts, wrote to Edmondson that he might want to try subdomains, among other things.

I don't think we can read much into this, because Cutts basically was answering a question politely. The webmaster was already about to try his idea, not Cutt's idea, and Matt said he could try it and some other stuff. Which is a low-value statement or no-comment comment. This is consistent with his previous public statements on Panda. He has been very helpful on other topics but very cagey on Panda. It could simply be the fact that Panda code was written by engineers in another dept, not the Spam Dept, so he isn't authorized to say much about it.

Is the content mass-produced by or outsourced to a large number of creators, or spread across a large network of sites, so that individual pages or sites don’t get as much attention or care?

This quote is one of the Google Guidelines after Panda [googlewebmastercentral.blogspot.com]. I don't see how putting old content on a new subdomain will override the issue Google expressed in that guideline.

Vanessa Fox, speaking for Google about Panda [webpronews.com], recently said:

Google wants the searcher to be happy and easily find their answer. Let’s say the content and the user experience are good for that page. Then you run into the issue of quality ratio of the whole site. The question then becomes if someone lands on your site and they like that page, but they want to engage with your site further and click around your site, does the experience become degraded or does it continue to be a good experience?

Google has already decided it has assumed responsibility for user experience beyond the landing page via Google search results. That was the motivation behind the decision to allow bad pages to bring down the entire site. A new subdomain doesn't protect users from bad pages. There's no way this subdomain "fix" is going to last. It must be a loophole Google engineers are already working to fix.

[elsewhen]

even if it *does* work, I suspect it won't work very long

i am a little more optimistic about it than you are... hubpages subdomaining strategy isn't really a trick to get google to think that low quality content actually is high quality content.

if a site with 100% crap splits up that crap into subdomains, then all 10 subdomains will get pandalized the next time the data is re-run.

if panda really treats subdomain content separately, then one would expect some of hubpages' subdomains to get out from panda, while others stay pandalized.

all the subdomaining strategy does it so immunize the good stuff from the bad stuff.

[walkman]

November 2011: New stats for Google's search index

4,501,701,451,854,019 pages and around
4,501,701,451,854,019 subdomains, give or take a few.

To be treated like, say, Wordpress, they probably have to keep the subdomains separated, not heavily interlinked and articles in the same major category.

[trakkerguy]

all the subdomaining strategy does it so immunize the good stuff from the bad stuff.

It allows you to immunize good from bad. But it also speeds up the reevaluation considerably, so is much more useful for some situations.

If a site only has 10% "quality" content, you can pull that 10% out to a subdomain and maybe see a quick recovery. So you aren't stuck with trying to weed out 90% low quality content. And don't have to wait months to see if you removed enough bad content to recover.

It is much faster, and easier to test. Gives us a much better tool to reverse engineer. So in that regard, they might shut it down soon as Netmeg predicts.

[rlange]

Splitting low-quality content off to a subdomain just doesn't seem like a logical solution to me. It seems obvious enough that a site could benefit from placing their entire forum on a subdomain, but I can't imagine placing low-quality articles on a separate subdomain. That just seems like more of a mess to me...

--
Ryan

[elsewhen]

I can't imagine placing low-quality articles on a separate subdomain. That just seems like more of a mess to me

in the case of UGC sites that have content from all sorts of contributors, it is difficult for them to know what google considers high quality and low quality. panda openly throws out the baby with the bathwater (see "low quality content on one part of your site, can negatively impact high quality content elsewhere on your site.")

in the case of hubpages, they are separating content by author - some of which are probably great, and some of which are probably terrible. the really good stuff, should get linked to more, get more social shares, and should get better user-engagement, and probably has better results on anything else panda cares about. why wouldn't google feel confident about sending it's users to those subdomains?

[trakkerguy]

Splitting low-quality content off to a subdomain just doesn't seem like a logical solution to me

That is NOT the solution hubpages is appearantly benefiting from.

The solution discussed here is, splitting the HIGH quality content off to a subdomain.

It recovers almost immediately. It doesn't get dragged down by the low quality content is was related to, so should escape any future Panda attacks.

Low quality content may also recover when moved to it's own subdomain, but SHOULD be hit with the next Panda evaluation.

[indyank]

He has been very helpful on other topics but very cagey on Panda..

Are they cagey because they use some strange (unacceptable) methods to crawl web pages? (Just thinking aloud here)

Have been hearing some strange crawl errors and stuff from a lot of people.Most of them could be found in WMT and in your log files.

One of the friends using a particular cache plugin for wordpress noticed strange errors in his logs - [wordpress.org...]

On further analyzing the log files for a few months, many such similar errors are in fact triggered by googlebot.

[epmaniac]

hi there, i have one question...

my site is similar to hubpages, where lots of users post different content of their own...

currently all of this gets displayed in main domain,... i use tags also for seo purpose.

My point is, when i move my authors and create subdomains authors wise like the way hubpages is doing,..

what should i do with tags?

is it ok that from each subdomains, lots of tags point to main domain's landing pages? i.e.subdomain linking back to main domain via tags is ok?

[Alekh]

good news...

[nuthin]

I have been using sub domains for years on a few of my sites.

I'm of the opinion if sub domains are treated as unique domains and most of the authority for domains always end up on the home page, then if you implement an effective internal linking / off-page strategy to parse internal PR onto the sub's then one would assume the sub domains will gain more authority than an internal page.

I'm with netmeg.

When you find something that works DON'T TALK ABOUT IT. #fightclub

If talks about using a sub-domain strategy kill my sites, watch out. :-) Argh......

[epmaniac]

no, i mean, from subdomains link going out to main domain's landing pages.... is it allowed?

example: a link mywidget.example.com/blue-widget.htm going out to

example.com/blue-widgets

[thegypsy]

Regarding this;

I'm leery of anything Google says or does right now, but if Google wants us to use this technique, then maybe they won't smack it down.

This part was not reported correctly in the WSJ article. From what I know that's not how it went down and Google just wasn't in the mood to debate that with the WSJ.

This reminds me of the last time I was here (when I was involved in the Overstock affair with the WSJ). I am sure we all know better than to trust everything we read out there. Both stories contained some incorrect assumptions and this one is no different.

Google did NOT suggest this as a work around and I am NOT either. I get the sense that Panda runs seperately than the 'at the time of discovery' algo set. Which means, that once Panda goes over the new SDs, the same issues might occur.

moderator's note: thegypsy (a.k.a. Dave Harry) has written
more about his involvement here: [searchnewscentral.com...]

[edited by: tedster at 4:24 pm (utc) on Jul 18, 2011]

[indyank]

I get the sense that Panda runs separately than the 'at the time of discovery' algo set.

@thegypsy, that is true and most googlers statements suggest that. Panda is a separate run and there probably is a manual trigger even if there is no other manual intervention as Google claims.

But what does happen is when you have a new url for a piece of content and you 301 redirect the old url to the new one, there are signs of ranking recovery, but it does evaporate over time.

But hubpage CEO definitely seem to say that they see the recovery only for the good content and not for the bad. That one is a mystery! How can the distinction be made by the normal algo if panda is a separate run?

[walkman]

about.com has all subdomains. I know they were hit by another panda but as much. Quality may differ too, of course.

[dataguy]

When you find something that works DON'T TALK ABOUT IT. #fightclub

I have nothing to say.

;)

[trakkerguy]

But hubpage CEO definitely seem to say that they see the recovery only for the good content and not for the bad. That one is a mystery! How can the distinction be made by the normal algo if panda is a separate run?

Yes, that was my question. My first thoughts were that his assessment was not accurate on that point.

But, if it was accurate , ie. low quality authors don't recover when moved to subdomain- it could be that they don't have sufficient external backlinks, and only ranked previously due to the internal linking structure.

Actually, that fits in with my thoughts on the signals G is using to evaluate a site for Panda. A site with lots of crappy pages would tend to have few backlinks to those pages.

So, a "good" author moves to subdomain. He quickly recovers from Panda, and rankings recover because he has some external backlinks.

A "bad" author moves to subdomain. He quickly recovers from Panda, but rankings don't return as he had few external backlinks (that is evaluated as normal algo, and not waiting for next Panda run).

I'm not sure I actually BELIEVE that crappy authors don't recover, or that the backlinks have anything to do with it - just a thought.