Every two years, our friend Rand at Moz wrangles his team to produce a correlation study of web site ranking factors.
Below we present the 2013 results which are very interesting to examine. You can choose to plot only a few categories of ranking factor and see the correlation to high rankings.
Of course correlation does not equal causation. However, in this data are some strong steers into where search engine marketing is going, and where effort should be focussed going forward.
The SEO industry is currently in a little bit of disarray. The amount of useful information available to SEOs is being reduced as Google increasingly hide referer data. Google’s recent algorithym updates seem to be increasingly focused on penalising any activity that might be “manipulative” beyond creating great content.
The takeways from all of this are pretty much what we’ve been saying about SEO for the last few years – create great content, pay attention to semantic markup, understand social and what channels beyond the web that you can leverage, be mobile friendly.
Moz 2013 survey and correlation data
Domain Level Anchor Text
These features describe anchor text metrics—both partial and exact match—about the root domain hosting the page. For example, for the page www.test.com/A, these features are for anchor text links pointing to *.test.com, not just page A.
Over the past two years, we’ve seen Google crack down on over-optimized anchor text. Despite this, anchor text correlations for both partial and exact match remained quite large across our data set.
Domain Level Brand Metrics
These features describe elements of the root domain that indicate qualities of branding and brand metrics.
For this study we tracked domain name mentions in Fresh Web Explorer. The correlations for mentions are relatively high, falling between 0.17 and 0.20 for mentions of the full domain name.
Domain Level Keyword Agnostic
These features relate to the entire root domain, but don’t directly describe link or keyword-based elements. Instead, they relate to things like the length of the domain name in characters.
Although none of these factors were highly significant, we did find a negative correlation of -0.09 with the length of the domain name.
Domain Level Keyword Usage
These features cover how keywords are used in the root or subdomain name and how much impact this might have on search engine rankings.
The ranking ability of exact- and partial-match domains (EMD/PMD) has been heavily debated by SEOs recently, and it appears Google is still adjusting their ranking ability. In our data, we found EMD correlations to be relatively high at 0.16 and as high as 0.20 if the EMD is also a dot-com.
Domain Link Authority Features
These features describe link metrics about the root domain hosting the page (e.g., for the page www.test.com/A, these features are for links pointing to *.test.com, not just page A).
As in 2011, metrics that capture a diversity of link sources (C-blocks, IPs, domains) have high correlations. At the domain/subdomain level, subdomain correlations are larger than domain correlations.
Page Level Anchor Text
These features describe anchor text metrics—both partial- and exact-match—to the individual page (e.g., number of partial-match anchor text links, exact-match links).
Despite Google cracking down on over-optimized anchor text, we found high correlations with both partial and exact match anchor text to the URL, with a 0.29 correlation with the number of root domains linking to the page with partial match anchor text.
Page Level Keyword Agnostic
These elements describe non-keyword usage and non-link metrics features of individual pages such as length of the page, and load speed.
This year’s data showed an interesting negative correlation (-0.10) to page response time.
Page Level Keyword Usage
These features describe use of the keyword term/phrase in particular parts of the HTML code on the page such as the title element, H1s, alt attributes, and more.
The data measures the relationship between the keyword and the document-both with the TF-IDF score and the language model score. We found that the title tag, the body of the HTML, the meta description, and the H1 tags all had relatively high correlation.
Page Level Social Metrics
These features relate to third-party metrics from social media sources such as Facebook, Twitter, and Google+ for the ranking page.
Social signals were some of our highest correlated factors, with Google+ edging out Facebook and Twitter.
Page Link Authority Features
These features describe link metrics to the individual ranking page such as number of links and MozRank.
Page Authority is a machine learning model inside our Mozscape index that predicts ranking ability from links and, at 0.39, it is the highest correlated factor in our study.