Link Structure Beats Google's Network Filter

[Garrett]

If you run a network of sites then there's a new algorithmic speculation you need to be aware of - the network filter.

First reported in WebMasterWorld (http://www.webmasterworld.com/forum3/23983.htm) and then in SEORoundtable (http://www.seroundtable.com/archives/000475.html), the network filter reportedly blocks the passage of PageRank and link popularity.

Some believe this filter looks for "unnatural" linking structures, that is, linking structures that wouldn't occur in the wilds of the web as people discover and link to content. Barry says the filter may be triggered if, "you have 20 Web sites and all 20 link to each other on every page."

Barry wrote a later update to the network filter discovery called Pyramid Linking Strategies (http://www.seroundtable.com/archives/000487.html) where he discusses a method for linking networks of sites that may avoid this filter.

Barry had a conversation with an SEO whose sole practice is link building. This SEO, who wished to remain annonymous so that Google couldn't target him directly, creates networks of portals that link to related and quality sites as well as his clients' sites. He links them in a triangular fashion.

"For example, one might create Site A, B and C. Site A will link to Site B, Site B will link to Site C and site C will link to site A. You will never find Site A link to Site B AND Site B link to Site A."

The portal development method that Barry's friend employs is worth mentioning here. "The portal sites will be on specific topics and contain links to high-level ODP and Yahoo! directory sites, you won't normally find deep-level sites listed in the portal sites," says Barry.

He also links these portal sites to his clients.

Getting on-topic links to his portal sites is much easier than getting links directly to his clients.

Once the portal's up he begins his "complex pyramid linking structure" by linking his portals to his clients, even his portals to each other, but never his clients back to the portals.

Barry also reported (http://www.seroundtable.com/archives/000509.html) recently that large sites with thousands of pages are suffering from a reduced presence in Google's index. The network filter as Barry described it didn't seem to remove pages from the index though - it simply removed the ability to pass PageRank and link popularity.

Barry wondered if Google's duplicate content filter's responsible for the dropped pages from large sites, but could it be that the site in question's linking structure is related somehow?

Here's the large sites discussion (http://www.webmasterworld.com/forum3/24091.htm) in WebMasterWorld.

[Dave Hawley]

Oh no, another unfounded speculation of a Google Filter. IMO, there is likley only 3 filters

Use strict filtering (Filter both explicit text and explicit images)

Use moderate filtering (Filter explicit images only -default behavior)

Do not filter my search results.

I can see no reason why google would "filter" any other type pages. They may not pass on a rank boostings, PR etc (at spider time) if you cross-link, but to filter would sloooowww down the SERP's way too much.

I believe, you are either in the Google index or you are not. It makes no sense for Google to spider and index a page only to filter it out in the SERP's.

[ronniethedodger]

You are confusing post-processing filters with the filters that get applied when they update their index Dave. The filters you described above are easy to filter out based on whether it is Adult related ... it is a simple yes/no flag that is set for the site. Even so ... all of that is pre-compiled before the query takes place and does not involve any more processing time than a normal query.

[AussieWebmaster]

Oh no, another unfounded speculation of a Google Filter. IMO, there is likley only 3 filters

Use strict filtering (Filter both explicit text and explicit images)

Use moderate filtering (Filter explicit images only -default behavior)

Do not filter my search results.

I can see no reason why google would "filter" any other type pages. They may not pass on a rank boostings, PR etc (at spider time) if you cross-link, but to filter would sloooowww down the SERP's way too much.

I believe, you are either in the Google index or you are not. It makes no sense for Google to spider and index a page only to filter it out in the SERP's.Actually this is when these results are being filtered... if the links are reciprocal they are not counted etc... it is the way they are added to the database for that particular site filtered by set parameters... what they exactly are is trying to be determined.

[Peter Holland]

I think a good website with a good linking structure is very helpfull for google to identify the immense internet.
If a website(every website) links only to websites who are in same subject or business, its possible then to put a "profile tag, labeled" on the links, and so a whole bunch of websites can be identified by his "profile tag, label", so then if someone search for horses, then only the websites who are "profile tag, labeled" horses can be showed to the client.
Everybody is talking about beating the google alghorythm but why is nobody talking about helping the google alghorithm.
If every put up his favorit color hat(yellow), and put on his favorit color jack(yellow). then you are easy to identify in a street where evrybody wear red hats and red jackets.

[etart]

If only "unnatural" links are being filtered, corporate sites linking to their franchises would be on the "safe" list as would all of the franchised that link to each other correct? If this filteration did take place, how would one even know if it just prevents the PR from going up?

[etart]

If only "unnatural" links are being filtered, corporate sites linking to their franchises would be on the "safe" list as would all of the franchised that link to each other correct? If this filteration did take place, how would one even know if it just prevents the PR from going up?

[Dave Hawley]

Even so ... all of that is pre-compiled before the query takes place and does not involve any more processing time than a normal query..


Agree. This is why I cannot see there being a "Network Filter". It's the word filter that I'm having a hard time with. I don't think Google would "Filter" these pages, but rather not pass on PR etc in an algorithmic sense.

To me the word "Filter" implies that the pages are ommitted, which I do not believe. IMO they are simply not getting a 'boost' as they would if not networked.


Google have used the word "Filter" on their Preference page, but IMO, this is only to make it understandable for the masses and is somewhat misleading. As Ronnie has pointed out the filtering has occured long before.[/b]

[fathom]

hmmm.... a rare occurence!

I agree with Dave here.

First "I do have a large network" and while backlinks and PageRank drop (and go up and down) all the time, ranks go only up.

That being the case a few things to consider (that we must always consider) now and two years ago.

"Network" is not the factor here as the whole World Wide Web is all but networks...

Heavy crosslinking has always gotten people in trouble - there is nothing new here.

Duplicate content across sites has to - nothing new here

PageRank always drops as Google's knowledge of the web grows - nothing new here.

Backlinks tend to drop consistent with PageRank - nothing new here.

And the biggest problem people have with this so-called network of sites - and thus someones rational for a "new Network Filter" - more of the same is better!

The moment that wide-spread report of lost PageRank which also has an immediate effect on backlinks (thus drops) and a couple of edge-riding people get in trouble - GOOGLE invented something new.

Sorry I don't buy, I don't see and I haven't felt it - I care little above "PageRank" or the number of "backlinks" -- only ranks, and as ranks continue to climb measureably across a large network that doesn't implement "more of the same"...

some are viewing the wrong merits and looking for a rational explanation and others likely implemented inferior strategies and putting 2 and 2 together.

While I personally have not viewed any of the threads and this is merely "an opinion" I don't go overboard - maybe that's the problem.

[jackson992]

Sorry I don't buy, I don't see and I haven't felt it - I care little above "PageRank" or the number of "backlinks" -- only ranks, and as ranks continue to climb measureably across a large network that doesn't implement "more of the same"...


However, you need tons of backlinks right now to rank well in Google so why would you say that doesn't matter to you?

[fathom]

Sorry I don't buy, I don't see and I haven't felt it - I care little above "PageRank" or the number of "backlinks" -- only ranks, and as ranks continue to climb measureably across a large network that doesn't implement "more of the same"...

However, you need tons of backlinks right now to rank well in Google so why would you say that doesn't matter to you?

A "ton of backlinks" or "just a few higher quality ones"?

(note: consistent with the thread topic).

[jackson992]

Aside from links from link farms all links are quality links:)

[fathom]

Aside from links from link farms all links are quality links:)

I wouldn't say that:

1. a link from a home page could be signiificantly different from one 3 level down.

2. a straight url link to a nice anchor text is quite a bit different

3. top of page link is better than bottom

4. link # 300 is pretty much useless

5. a duplicate content page link has zero value to you.

There are another 50 odd examples and none have anything to do with link farms or FFAs.

[jackson992]

I really cannot say I agree with any of those points. Most noticeaby I don't agree that position of link affects anything at all. There are several sites that place the nav at the bottom of the page and rank well.

I also don't agree that home page links are any more beneficial than interior pages. Actually I would much rather have one from an interior page. This makes the link more focused IMO

It has yet to be proven that pages with more than 100 links suffer at all despite the fact that I see this posted often as gospel.

To me there is no such thing as a duplicate content page unless it EXACTLY the same as another page which is very rare


#2 I couldn't quite follow what you were trying to say lol

[fathom]

I really cannot say I agree with any of those points. Most noticeaby I don't agree that position of link affects anything at all. There are several sites that place the nav at the bottom of the page and rank well.

So ask them to "move" the links to the top and see ranking "better". I didn't imply "no benefit" but "less benefit" - important links are at page top - not page bottom.


I also don't agree that home page links are any more beneficial than interior pages. Actually I would much rather have one from an interior page. This makes the link more focused IMO

Preference has little to do with this... a website "link credits" tend to be to the mainpage, and subsequently filter down through the website link hierarchy.

The farther you are from the "credit" the less return you get.


It has yet to be proven that pages with more than 100 links suffer at all despite the fact that I see this posted often as gospel.

Fine with me.


To me there is no such thing as a duplicate content page unless it EXACTLY the same as another page which is very rare

Rare to see in Google - rare in use - NOT!


#2 I couldn't quite follow what you were trying to say lol

http://www.spheri.ca or eMarketing Firm (http://www.spheri.ca)

[jbgilbert]

Just 2 cents worth. Too many links from a single domain is not good. Which filter would that be part of?

[fathom]

Just 2 cents worth. Too many links from a single domain is not good. Which filter would that be part of?

Yes, Yes, Yes!

I was waiting for someone to say that!

jbgilbert this is what Google is doing and much like their stance on duplicated content "one will do".

[jackson992]

This to me almost sounds like cross linking which I do on all my sites. Right at the bottom I have a section where it says Visit our other sites. I cannot see Google penalizing for this when it's there as a service to your visitors


P.S.

Fathom, last night I put some links to my main pages at the top of every page on my site. I expect to see rank increases judging by what you said:P

[fathom]

This to me almost sounds like cross linking which I do on all my sites. Right at the bottom I have a section where it says Visit our other sites. I cannot see Google penalizing for this when it's there as a service to your visitors

Penalty - negates effectiveness of a deploy tactic.

Dependent on - depth and use of tactic. If you "are not" in Google's crosshairs - no problem but "more of the same" pushes you closer to that edge.

Back at the thread start some people had legitimate concerns and rationalized that Google is negating "networks" why?

Crosslinking... and lots of it and just happen to drive themselves over the edge at the same time as many others noticed a major decline in PageRank and put 2 and 2 together and got 10!

You can't be saying - they didn't notice anything - are you?


Fathom, last night I put some links to my main pages at the top of every page on my site. I expect to see rank increases judging by what you said:P

Links to what? Hopefully another website but than not a fair test if that link didn't already exist say at the bottom.

Also note: given the fact that a link is "known" to benefit "a little" (per link) and position (if what I say is true) helps a little more:

What do you expect to see as a fair conclusion to assess whether I am correct or incorrect.

If moving up considerably in positioning is your idea of success - don't bother wasting your time > I will concede you have a better appreciation of the situation.

On the other hand - a position or two - 100%!

[cbp]

Too many links from a single domain is not good.

I thinx this a problem also.

I have a site (forum) that is in Google but has had PR0 for 9 months (its always been PR0) - its in DMOZ, its in Google Directory and several others - has some other >PR3-4 links .... just can not get any PR for it (I don't really care as ranking in SE's is not and issue --- the only issue is that its annoying me as I can find no reason for it!!!).

The only possible reason is that it has >1000 links from one domain (the domain that the forum is associated with) - in each newsletter that goes out, there are several links to the forum to discuss issues raised. The newsletter (and the links) are also published on the website (hence the 1000's of links to the forum)... this is the only reason I could imagine for the PR0 the forum has.

CBP

[ronniethedodger]

Is the forum in a sub-directory or a sub-domain?

[cbp]

The forum is on a totally different domain....(long story why :-)

CBP

[ronniethedodger]

The forum is on a totally different domain....(long story why :-)

The only reason I asked, was when I saw your post ... my first thought was that it was in a sub-domain (not a directory). To go that long without having any PR at all and through all sorts of updates, and even before Florida is a bit unusual.

It almost has to be domain related, perhaps something with your host. You remember last October when Google could not crawl a certain web host's sites? (we don't need to say the name again, cuz people confused that with the Registrar part of the company.)

Something along those lines is my thinking on this. I have seen a couple of sites experiencing this problem, and I think it is a hosting issue from what I can see. Their sites are clean too.

[fathom]

The forum is on a totally different domain....(long story why :-)

CBP

Duplicated content?

I'll assume the newsletter items are on the forum - correct?

And the web-based newsletters point to the forum at or around similar content?

Could be the problem

Here an example I use to illustrate a fair variant of duplicated content - over 500 sites saying exactly the same thing.

"The Art Institutes is the leading educator of students in the creative and applied arts, including design, media, fashion, and culinary programs" (http://www.google.com/search?q=%22The+Art+Institutes+is+the+leading+educ ator+of+students+in+the+creative+and+applied+arts, +including+design,+media,+fashion,+and+culinary+pr ograms%22&num=100&hl=en&lr=&ie=UTF-8&oe=UTF-8&newwindow=1&c2coff=1&filter=0)

Interface wise "google in blind" to graphics so text becomes to major factor.

Here though these websites do not point to each other and thereofre no direct method for Google to filter carbon copies.

While I don't have specifics for your website - I doubt the "links" themselves are the problem in that "links to you cannot hurt you".

[chris_g]

Fathom Wrote:
Dependent on - depth and use of tactic. If you "are not" in Google's crosshairs - no problem but "more of the same" pushes you closer to that edge.

I think this is an important point. its the depth of the tactic that develops the increased probability of being filtered (or penalized for that matter) by Google. For example; It's unlikely that 10 reciprocal links will cause problems but 1000 reciprocal links increases your potential problems significantly over time.


ronniethedodger Wrote:

It almost has to be domain related, perhaps something with your host.

I have seen a couple of sites experiencing this problem, and I think it is a hosting issue from what I can see. Their sites are clean too.

This is a different issue but is there some threads that discuss host issues (re: search engine indexing) in depth. We have done some work in this area since we find that more sites have hosting issues than one would think. This is a difficult topic to get good information on.


Fathom wrote:

Duplicated content?

How granular are the duplicate content filters in your opinion. We have noticed cases where if a sentence contains the keyword in search and the sentence is repeated across a couple hundred websites only one website will appear in the listings. I have a hard time believing there is cause to be this granular.

[Will.Spencer]

I have been doing some heavy pyramid cross-linking.

I have two types of sites in the pyramid:

1. Sites that link out, and are linked to.
2. Sites that are only linked to.

While nothing with this many unknown variables can be considered reliable, the results which I have been seeing seem to support the "One site; One vote" theory.

If this is true, pyramid linking is not going to be amazingly effective.