Cyrus Shepard's Beyond Exact Match Anchor Text Whiteboard Friday back at the start of September inspired me to run an experiment. Cyrus pointed out that there has been some evidence to suggest that the exact match anchor text link may not be the holy-grail it once was. I wanted to test the theory out for myself though and try to delve a little deeper into the effectiveness of different kinds of links and also attempt to identify which link type is the most effective. I also wanted to get an idea of just how advanced Google is at recognising the context around a link, i.e. is it even necessary to include a keyword in the anchor text or can it process non-keyword, 'click here' type links if they are placed near to the keyword itself.

Before we get started, I want to say that my test was pretty small (three new domains, three mini-sites, 10 links to each) this means that by no standard are my findings conclusive but I think it definitely offers some food for thought and certainly some areas worth investigating further.

Phase #1 – Keyword selection

I was in the market, so to speak, for a keyword with relatively little competition. My logic being that this would make the test as clean as possible since the only competition would be the other sites within this experiment.

This would mean that we would be largely in control of other variables like on-page optimisation, domain age etc. offering us more of an indication of the impact that the different links were having. A "Laboratory SERP" rather than a SERP that's been plucked from the wild.

I settled eventually on the 'Orange Mega Sweets' because it was really low competition and I needed a keyphrase with at least three words to make the experiment possible.

Phase #2 – Domain purchase

The next step was to acquire three domains. I wanted three brand new domains so that no other factors were coming into play. I considered acquiring the exact match domain to see how it would rank compared with the above three but in the end decided against this as it wasn't really my intention with this experiment.

Above you will see a screenshot taken on September 13th 2011 detailing the hand registration of PurpleSweets.com, YellowSweets.com and BlueSweets.com. I decided on these three to ensure they were all equally relevant, similar lengths and of a uniform style – again to try and make this test as clean as possible.

I also paid for a new hosting plan to ensure absolutely no associations were made between these sites and my other websites. In hindsight, I should have gone further because I should have bought a hosting plan for each to ensure no association was made between the three test websites.

Phase #3 – Website Build

Very simple one-page microsites built on the latest version of WordPress with a standard WooTheme installed.

I made a couple of tweaks like footer attribution link removal, changed the homepage to display fixed content, linked out to a relevant Wikipedia entry to add some clout, made the title text based and added a keyword based tagline. I made the same on-page changes to all of the websites.

Body copy on each of these websites was again different (albeit similarly rubbish) and I also found a different image for each of the sites but kept the same title and alt tags for fairness.

Phase #4 – Linkbuilding

This part of the experiment was the most challenging. Deciding which types of links to acquire was certainly a difficult call to make.

In the end I went with a batch of 10 links for each of these three sites. I settled on some link sources which I happen to know index really fast (read – slightly too hot for client projects). Since this was an experiment, it didn't matter if the domains got burned.

(Link URLs masked to protect the sources - I know you can probably find them if you really want to!)

As you can see, with BlueSweets.com I built arguably organic looking links using non-keyword anchor text such as 'here' 'click here' and so on. My theory with this kind of link was that I wanted to test just how clever Google was at recognising the context of a link as these 'click here' etc. links were all placed within close proximity to the target keyword phrase 'orange mega sweets'.

With PurpleSweets.com, I went exact match anchor text crazy and finally with YellowSweets.com I kept things looking a little chaotic with a mixture of partial anchor text links.

To be as scientific as possible, I created links that were from very similar sources for each of the sites and I built 1 link at a time to each website – all links were completed and live within a 24 hour time period.

Phase #5 – The Results

Here are the rankings for each of the websites on day two of the experiment (14th September):

As expected, none of the sites ranked inside the top 50.

What about on 15th September though?

Interestingly, by day three of the experiment and only day two of these sites being live BlueSweets.com shot straight to number one for the target keyword. If you remember this was the site with the 'click here' type anchor text links.

It was a surprising start given that my feeling was non-keyword anchor text links probably lacked the relevance that Google was looking for but maybe I was about to prove myself wrong and perhaps Google has got smart at recognising the context around a link..? Nevertheless, my feeling that after just one day it was too soon to make any kind of judgement.

September 16th and BlueSweets.com had maintained its position - see screenshot below for Google.com accessed via a proxy.

By September 17th BlueSweets.com was still sitting at position #1 and the other two sites were still nowhere to be seen.

 

Later that day I ran the Rank Tracker again (yes I checked rankings twice in one day…) and things had really been shuffled:

Out of nowhere YellowSweets.com takes the lead and BlueSweets.com disappears off of the radar. YellowSweets.com is the website with a link profile consisting of partial match anchor text links.

The eagle eyed amongst you will note that the ranking URL was http://yellowsweets.com/p rather than the main URL – odd considering I mistakenly installed WordPress there for less than 5 minutes before I realised what I'd done, removed it and re-added to the root folder. If anyone has a theory on how Google managed this or why it happened...I would love to know!

I let things run their course for a couple of days and by 21st September, we see the following:

YellowSweets.com holds #1 and has an indented listing at #2. PurpleSweets.com (that has 10 exact match anchor text links) suddenly makes an appearance, dropping to #3 from #2 that it achieved just a day earlier (20th September).

Fast forward a week and the picture remains almost the same...

By 28th September, YellowSweets.com had lost its indented listing but the site remained at #1 with PurpleSweets.com sitting at #3. BlueSweets.com is nowhere to be seen.

Summary

Small Disclaimer

During this two week experiment, there was one occasion where SEOmoz's rank tracker data didn't match up to the rankings seen when a search was performed manually via a proxy.

Make of this what you will but my feeling is that this doesn't detract from the overall validity of the experiment because I was trying to identify patterns in a site's movement relative to each other rather than specific rankings on a given day.

What I learned from this experiment

Google isn't quite there yet

Google perhaps isn't quite there yet with their assessment of link context – the results of this experiment seem to suggest that Google still needs a hint and places a high degree of trust in the relevance of the anchor text (whether partial match or exact match).

The links built for each of these sites were from generalised sources and arguably from websites that weren't all that relevant. The pages these links were placed on however, made some of the right relevance & signals in terms of relevant largely handwritten content, a decent page title and no nasty paid links to bad neighbourhoods.

For this particular SERP at least, Google still appears to place a huge amount of importance on the anchor text of a link rather than the context or perceived relevance.

Partial match is the better long term strategy

Although YellowSweets.com was the slowest off the mark in this experiment, it made a steady rise to the top and retained its position. This indicates that partial match linkbuilding in practice is the safest in terms of long term stability and potentially also the more effective now and in the future (proving what Cyrus was saying in his ‘Beyond exact match’ WBF).

Think beyond the link

In the end it comes back to Cyrus’ closing statement – "Judge like a human". In practice, a link profile consisting entirely of exact match links doesn't look natural and is easy for Google in the future to de-value a huge proportion of your link profile in one fell swoop.

Assess link opportunities also as promotional, brand awareness and traffic opportunities and look at ways to build a presence online that will survive a future that is likely to be less about links.

Questions

• What made BlueSweets.com jump straight to #1? - And did this trigger some kind of warning/penalty that made it disappear?
• We know about penalties for over-doing exact match links so why did PurpleSweets.com escape unscathed? - It has a link profile consisting of 100% exact match links and no other directly competing site has this many exact match links for that keyword so in theory the site should stick out like a sore thumb.
• This leads me on to my next point that what results would we see in a different sector? - Google undoubtedly monitors some SERPs more than others so what would be considered a dangerous number of exact match links is relative to the marketplace.

I would like to end by saying that all links were built for test purposes only. If you do find the link sources (well done you) please keep them to yourself, I have started to remove the process of removing the links since the experiment has drawn to a close.

By James Agate, Founder and Director of SEO at Skyrocket SEO – a leading eCommerce SEO and Conversion Optimisation consultancy.