Rand Fishkin, Wizard of Moz | @randfish | [email protected] Google Ranking Factors: Correlations, Testing, & Hypotheses
This Presentation Is Online Here: bit.ly/grankfactors2014
What does it mean? How should we apply the data? Correlation
Correlation does NOT say why these results rank higher than these results More on Rand’s Blog
Correlation tells us what features, on average, the results that rank higher have which the lower ranking results do not have. More on Rand’s Blog
Correlation tells us what features, on average, the results that rank higher have which the lower ranking results do not have. More on Rand’s Blog I’m actually MORE interested in this than I am in whatever Google’s actually using to rank the results!
Via Moz’s 2013 Search Ranking Factors
Via Moz’s 2013 Search Ranking Factors To me, this says individual pages still matter, but there’s a lot of weight on the hosting domain.
Via Moz’s 2013 Search Ranking Factors MozRank used to be higher, and so did linking root domains. Google’s probably getting more complex.
Via Moz’s 2013 Search Ranking Factors $100 says that if we could get more comprehensive brand mention data, this correlation would start to look a lot like links
Good discussion about Google+ correlations in this post Google+ is just too damn high.
Good discussion about Google+ correlations in this post Google: “Most of the initial discussion on this thread seemed to take from the blog post the idea that more Google +1s led to higher web ranking. I wanted to preemptively tackle that perception.”
Good discussion about Google+ correlations in this post To me, that’s Google working really hard to NOT say “we don’t use any data from Google+ (directly or indirectly) at all in our ranking algorithms.” I would be very surprised if they said that.
Good discussion about Google+ correlations in this post That said, all of the correlations with social are high. That tells me the things that make content have success on social probably have a lot of overlap with what makes content successful in Google.
Good discussion about Google+ correlations in this post Domain name keyword matching continues to show decline.
Via Mozcast PMD was as high as 5% two years ago. EMD was almost 6%. Both have fallen precipitously.
Basic introduction to LDA and topic-modeling systems here. We were able to build a better keyword-modeling system in 2013, and correlations were higher than in past studies looking at raw keyword repetition or use in title elements.
More on rankings and page load time here. Response time was interesting, but it’s likely a very small direct factor and relatively big indirect factor (i.e. users like fast-loading pages, and people link to/share what they like) ?
See How Unique Does Content Need to Be. Last, more content still seems to, on average, slightly overperform vs. less content. I’d question any causality here, though.
I hope to see many, many more correlation tests and more things considered! Causal or not, correlation data is incredibly useful.
What can we learn from a recent SEO test? Testing
Hypothesis: It seems like Google is starting to ignore or discount anchor text in links.
Here were the test conditions: #1: Three-word keyword phrase in Google.com US #3: We pointed links with NO query-matching anchor text from 20 unique, not-particularly-on-topic, high DA domains at result A and EXACT-anchor-text match links from the same pages at result A. #2: At start of test, result A ranked #20, B ranked #13.
After 3 Weeks: All of the links had been indexed by Google Result B (with exact-match anchor text) ranked #9 in Google.com US Result A (with non-query-matching anchor text) ranked #18 in Google.com US
Of Additional Interest: Result B (with exact-match anchor text) ranked #4 in Google.co.uk Result A (with non-query-matching anchor text) ranked #19 in Google.co.uk ~5 of the 20 linking domains were from UK sites
Takeaways: #1) Anchor text still matters #2) Geographic location of links matters
I’d love to see lots more testing in the SEO world. Even imperfect tests are fascinating and useful, IMO.
Three guesses Rand has about what Google’s up to Hypotheses
Hypothesis #2: There’s an aspect of mention frequency and mention source in Google’s brand/domain bias More and more, these queries return results that look like what you’d get if you polled people on the street to tell you what brands they most associated with the phrase “men’s sneakers”
Hypothesis #3: Google is using search & visit patterns to connect words & phrases and rank results Why do they list these 3 in the top 10? My guess – it’s because they are most often visited by people who’ve done searchers around “luxury resorts Australia”
Hopefully, these hypotheses can lead to experiments, results, and more sharing ?
Rand Fishkin, Wizard of Moz | @randfish | [email protected] bit.ly/grankfactors2014