Evidence that Google can deal with reciprocal links?

I have been running coops ads for a while now and getting great results. However, on the 21st October (on one of my largest sites that itself is advertised on the network) I increased from 1 to 5 ads.

Since Thurs last week these great results have dissappeared!

I have other sites that I advertise on the network that are either a) not part of the network or b) have fewer pages and these are continuing to see great results.

Would anyone care to comment on whether this can be considered evidence in support of a theory that reciprocal linking is an issue on google?


Note: this site has over 30,000 pages - at 5 ads per page thats 150,000 ads on my site at any given time. Perhaps someone with a more mathematical bent could calculate how many of the coop network sites I would be linking to at any given time.

 

Firstly, the math side..
Which particular ads are displayed at any one time is anyones guess.. but truth is, the php that pulls the ads grabs 100 every 24 hours (I think it's 24 hours) and stashes them in a file. Which ones are shown throughout your site is a random selection from those 100. So, at best, you'll be linking to 100 discreet sites across 150,000 instances.

I run 5 ads per page on a site that's over 180,000 pages in size, being 900,000 live ads at any one time.

The reciprocal linking theory.. it's prolly a good idea for *ALL* to read this, as it was a simple test I conducted for exactly this reason, and others may have something to add (or a correction to make about my observations).

To start with, the reciprocal linking theory is bunk. It doesn't make sense that links would be penalized purely because they are reciprocal in nature. If penalties occur, they have to be for additional reasons - possibly volume, theme, or other.. but that's speculation.

Here's what I did. I decided to keep the ads showing (900K) on my primary site. But instead of running ads for the same site, I used it's weight to promote a new site (a test site). In simple terms, everyone retained the benefit of exposure through my site, but there were no ads linking back to my site. This was to test the theory of reciprocation.

The site had been (for around 6 weeks) slipping in the SERPS, shedding around a thousand daily uniques, every 2 weeks. So after the 6 weeks, it was down by 3K visitors per day.

I persisted with the test for around 4 weeks, which also resulted in a slight reduction in IBL's with the last update. I then restored the ads with no notable difference.

So in terms of reciprocation - it's (in my books) a myth. I'm now hunting for alternate explanations, the most likely being the relative number of outbounds to unrelated sites.

I'll be removing the site from the network soon to test whether or not unrelated outbounds are a factor at play. I also think it's important to distinguish between established sites, and new sites.

Now that I'm done with this thesis, I'd love to hear what others have to say..

 

Well done...it does show the factors at play well.

In the "big experiment" some of these were shown to be of importance...to wit, the reciprocal linking scenario.

At around June some unexplained movements occcurred in one of the "controls" that were run alongside, as well as in the main experiment, and these pointed to OBLs.

The explanation that I could apply to it was that the balance of OBLs to IBLs was affecting the results - we did not pursue a test - but more the theme, or relevance, of these, affected what was happening as well. [How did we know this? It is too complicated to tell but, basically, by elimination ]

So in summary it is the balance of relevant links that is important - unfortunately we are unable to "theme" the adnetwork links so a test on this is nigh impossible

 

Thanks john_loch and Foxy.

Given crawl time I guess that the number of sites linked to by this website *could* be significantly higher than 100 across say, a two week crawl period.

If the ads are updated more frequently than 24 hours then this number would increase markedly so if anyone could confirm how often the ads are updated (and the quantity saved) that would be useful.

john-loch - I am running 5 ads per page on other websites and to date have not seen any negative impact related to your OBL point.

Foxy - are you saying that you have evidence that theming plays a significant part? and if so is this in relation to IBL or OBL or both?

Oh, and if you could define "relevant" from a robots point of view that would be helpful.

 

Foxy,

Are you saying that the ratio of IBL to OBL affected the ranking?

If yes, which would your test indicate would rank higher?

20 IBL - 10 OBL
10 IBL - 10 OBL
10 IBL - 20 OBL

 

Relevance [that is from a site that links with similar modus operandi] or theming is the most difficult one - in essence if you linked with a keyword imbedded in text [or other] of relevance to the KW phrase then it showed a greater potence than otherwise.

I have not tested the numbers balance precisely, but you should know that we do not do reciprocals, and so when I installed a themed [relevant] directory with OBLs [only one link to it] our breadth of keyword positions in the first 10 jumped dramatically. Thus I would be inclined to say that the 3rd option is that which I would favour, keeping in mind the relevance.

Edit: On further thinking I decided to add that IMHO Google just "loves" Directories based on observations with "generic" type KWs

 

I'm sure there are other things at play (it wouldn't just be unthemed outbounds, otherwise everyones site would head south !)

I suspect it's a combination of factors not the least of which is themed IBL to OBL ratios.

Though I think I'll test this by reducing the number of ads per page rather than eliminating them altogether.

 

Now that will be interesting

 

Foxy..

Just to make sure I'm not missing anything.

At a page level are you saying that you gain rankings for a page by adding OBL ("off site") using keywords that are not an exact match to your KWP for the page in question but... are related by using some keywords in the OBL?

If yes, is this purely related to the OBL text or are you in fact saying that the page you link to is a factor?

PS john-loch - you seem very convinced that reciprocal linking is not an issue for google, can you really be that sure? It is one possible explanation for the woes of one of my sites.

 

Just to throw out another possible explanation, perhaps the addition of 4 OBL per page increased the PR bleed from your whole site. Adding 120,000 extra OBL would start to add up I would think. This explanation really only makes sense if you had very few links on most of your pages to begin with.

 

Hi Crew

Given the structure of this site (it has ~ 70 links off the home page) I don't believe PR bleed is an issue - particularly if you consider all these links are reciprocated!

From what I can see from the code it appears that each time ad_network.php is called it adds one new ad until there are 100 saved. After that it checks to see if the oldest ad is over 3600 seconds old and if true replaces it with a new one.

So, on a high traffic site (note that spiders will also process the file and all pages on a site call the file) it is possible to have one new ad each second making it possible to have 86,400 different ads on a site in any 24 hour period.Of course these ads will not be unique but I don't know how to estimate how many will be for unique domains.

I am no statistician but this seems to suggest that over a "crawl" period (whatever that may be) it is perhaps likely that a large high traffic/highly crawled site will actually be linking to ALL the sites in the coop network.

If so, all its coop links are reciprocated at a domain level. This must make it possible for us to prove whether google devalues reciprocated links.

 

I think a direct match with your KWs is best. We all have seen instances where you try and on-page optimize a page of yours only to find that a different page with just one link mentioning your KW to your page ranks higher.

It also tells G that you have some links about the subject/KW on your page to others thus being closer to what is an 'authority' site. I don't believe it's the actual landing page that is evaluated, just the anchor text, for now...

If you target [KWPhrase]KW1 KW2[/KWPhrase] a good alternative perhaps would be the anchor text to be -KW1 KW2 Information- or something along those lines.

 

Hi TOPS30

What do you make of my previous post regarding the coop network and reciprocal linking?

 

I'm not with the network.

My guess is that it's too hard for them to find out. The links are of a dynamic nature hence not really reciprocal. Ones you link to don't necessarily link back to you.

I think what you noticed might be down to other factors.

 

It is of course possible that the ranking drop is related to other factors and I am checking on these. I guess I am particularly surprised to see the pages I advertise on the coop network decline so markedly even if the rest decline for other reasons.

I do not believe it would be difficult for google to calculate reciprocation on a per domain or per ip basis.

It is true that the ads rotate. However, if you have sufficient traffic and pages I believe that that the statistics would show (they are beyond my powers of calculation!) it is possible for one domain to be linking to practically all domains in the coop network. Whilst the pages the ads are on would constantly change they would nonetheless be there on a domain basis and would be reciprocal on a domain basis.

In my view it is something that would be useful for us all if we could prove/disprove exactly how google handles reciprocal linking.