Photo by stevendepolo
Some of my UK SEO friends have been known to argue that irrelevant links have long been detected and discounted by search engines, making related links an important part of your link building strategy. Do you really need large numbers of “relevant” links to get a site to rank for your top keyword?
No. As long you’ve built links on reasonably trusted, authoritative domains, and you’ve thrown in some (sometimes over) optimised anchor text for good measure, you can still rank. That’s not to say relevance plays an important role, but not as much as one as I had been hoping for.
I’ve been checking out some high search volume, relevant key phrases to get optimising my new company website for. One of the most consistent things I’ve found by checking out the back link profile data is that link authority and inbound anchor text optimisation completely outweighs relevance of the links themselves. Though I can’t tell you the keyphrase, or expose the competition’s link building strategy, there are some very cool take aways below.
The data sources
I gathered data from Yahoo Site Explorer and Linkscape. All the usual metrics were there (see below) with an additional series of percentage calculations and counts of inbound link relevancy by profiling link URL’s, titles and domains.
- URL -Google Position -Pagerank -# Yahoo Inlinks -# Relevant links in first 100 Yahoo Inlinks -Yahoo Inbound link Relevancy % -mozRank and other Linkscape metrics from the advanced report CSV -Inbound link URL relevancy (count in Linkscape export) -Inbound link relevance % -Most valuable inbound link mR -Count of relevant anchors -Relevant anchor % of total inbound links -Average URL mR of all relevant inbound anchors -Highest mR of all inbound (relevant anchors)
1) The first 100 links reported by Yahoo showed high, medium or low relevancy counts. There was no strong correlation between Yahoo Site Explorer inbound link relevancy in the first (arguably most authoritative) 100 links and Google ranking for the term monitored. The top ranking site had 1 relevant inbound link, on a non keyword rich domain name.
The second and third results contained much larger numbers of relevant inlinks, though the vast majority of them were nofollowed, comment links from blogs.
2) mozRank of the URLs offers a stronger correlation in some cases, with a significant anomaly in position 6. This ranking happened to be a very strong domain, with an article on the subject in question holding the ranking position with a non canonical form URL.
3) Juice-wise, most of the domains sat in a comparitively similar range (domain juice is the sum of mozRanks for all URLs in a domain. It is shown on the same 10 point logarithmic scale that mozRank uses), so not much joy there.
4) Inbound link URL relevancy *(count) was calculated by looking at the URLs, domains and titles of the most authoritative inbound links in the Linkscape data.
* I would like a way to make this calculation stronger by also including mentions of the key phrase in the body content on the linked-from page, so consider this a rough guide. To make the calculation even stronger, an array of related key phrases could be generated and each one used to count towards relevancy. This would likely increase the relevancy count in some cases!
5) The overall inbound link relevancy expressed as a percentage to make the comparision a touch more meaningful. Again, the highest ranking sites inbound link profiles often produced the lowest (or very low) levels of relevant inbound links.
6) I found the percentage of “optimised” inbound links for our top 10 ranking URLs quite interesting. Could there be a tolerance range in which a search engine finds certain anchor text distribution percentages acceptable? The first 5 results, with a gentle downward slope followed by a increasing percentage for the lower positions looked really interesting and I’m definitely looking forward to doing more tests with other search terms.
7) My favourite measure – a plot of the average URL mozRank of all inbound links that carry relevant anchor text, originally discussed in a post about the Vince update, appears once again. Although the difference (slope) is gradual, and there are a few anomalies, there’s definitely a trend to be found here.
What’s the conclusion?
My conclusion? Sites are ranking regardless of how “relevant” this test percieves their inbound link profiles to be. The top ranking sites in the test had consistently low relevant link counts. Instead, authority factors such as value (mR) passed via relevant anchors and anchor text term distribution percentages seem to play a strong influencing role in final ranking position.
For next time, I’d really like to spend more time analysing and defining “relevancy”, as I mentioned in point 4. That said, I don’t think we would see much of a change in the patterns, with high ranking sites continuing to obtain rankings using heavily optimised inbound link anchors from trusted domains.