On Correlation between SEO metrics and Ranking

Brief Review

A recent article by Rand Fishkin Lessons Learned Building an Index of the WWW presents an illuminating summary of the seomoz findings with regard to what SEO metrics have a tangible correlation with search engine rankings.

Basically, the major conclusions are summarised by two graphs. The first one

Shows the correlation of various SEO metrics (on the left) and search engine rankings. Here n+1 means comparing two adjacent positions in SERPs. The correlation is measured in percents above the noise level.

As you can see from this graph, the best data are only 5% above the noise level, which is not much to be absolutely certain in their relevance. What transpires, however, is that among the SEO metrics that correlate most strongly with ranking are number of unique domains linking to the site in question and mozRank (seomoz own metric, available to their paid members).

Among other SEO factors available to general public (i.e. without subscribing to seomoz paid tools) are Google PageRank and Google Backlinks. Those, however, much closer to the noise level and hence are less trustworthy.

The data below the 95% confidence level (red bar at the bottom) are within the noise floor and therefore do not display any reliable correlation.

The other graph from Rand’s article summarises correlation between the same SEO metrics and ranking for two positions in SERPs separated by 10 places (e.g., top 15 and top 5).

As you can see, the correlation here is much clearer, as the largest bars are over 20% above the noise floor, meaning that those are at least 20% better than random guessing.

The leading factors here are the number of unique domains (22%), Google and Yahoo! backlinks (20-21%), as well as Google PageRank (20%).

Surprisingly, in his post Rand only emphasize the number of unique domains and Google backlinks as the most relevant ranking factors, although the above graph clearly suggests that Yahoo! backlinks and PageRank are almost equally important for “distant” SERPs positions.

If we focus only on SEO metrics that exhibit the correlation for both “near by” (+1) and “remote”  (+10) positions in SERPs, than there are only three factors that correlates well in both cases:

• Number of unique domains
• Google Backlinks
• Google PageRank

Summary

Statistical analysis of correlation between various SEO metric factors and rankings in SERPs performed recently by SEOmoz team has revealed rather illuminating information. Namely, one of the most important ranking factors appears to be the number of unique domains linking to the site in question, irrespective (!!!) of their quality (in particular, Trust and Authority).

Among other important factors appear to be Google backlinks (link:www.yoursite.com) and Google PageRank.

There are a couple of issues which have not been clarified by the author of the original article:

1. Are their results relevant only to SERPs in Google or in other Search Engines (Yahoo!, MSN Live etc) as well?
2. Are the Google backlinks quoted refer to backlinks determined by the link operator (link:www.yoursite.com) or do these data come from Google webmaster accounts of the domains in question?

Did you like this? If so, please bookmark it, about it, and subscribe to the blog RSS feed.