Google Ranking Factors

Over that past year and a half, Google has made several big changes to its search engine ranking algorithm. These changes have had a significant impact on the Google rankings of thousands of websites. This has left many in our industry, including the experts, scratching their heads, wondering which factors have changed. As a result, our industry has become overwhelmed with rumor and speculation.

Often the internet marketing advice we hear is not substantiated with any empirical evidence. This is of course the cause of much of the rumor and speculation. We know there are many great companies out there that use good science and sound logic to direct their internet marketing strategies, but in our experience, there are still a lot of firms that don’t – too many rely solely on gut instinct and hunches. We hope that by publishing our recent findings we can help to dispel many of the rumors that have been circulating and create a real awareness of Google’s most recent changes.


These are the ten factors from our study that showed the highest correlation to ranking well on Google. We were a bit surprised to find that off-page factors still had a much higher correlation to rankings on Google than on-page factors. In addition, it was interesting to see that even with Google’s EMD update, an exact match domain to the search query is still the factor that has the highest correlation with rankings.

Top Ten Ranking Factors

There are two valid reasons why the “EMD” correlation coefficient in our study differs so much from findings in other studies:

  1. The correlation coefficient for “EMD” has fallen
  2. We used a different methodology (our explanation follows)

In our analysis, our variables consisted of two data types:

  1. Ordinal
  2. Dichotomous

It’s common practice in statistics to use the Spearman rho to measure the correlation coefficient between ordinal values and rankings; however, this is not true when it comes to dichotomous variables and rankings. For dichotomous variables, statisticians primarily use a method called “Rank-Biserial.” This method is used when one variable is ordinal or rankings, and the second is dichotomous (meaning it’s either 0 or 1, yes or no, etc.).

The Spearman rho was not designed to measure the correlation coefficient of rankings and dichotomous data. For this reason, we chose not to use this method for our dichotomous variables (please refer to our methodology where this is explained). This is where the difference between our study and others that have been done previously is manifested.

Other correlational studies we’ve seen use the Spearman rho to measure the correlation between “EMD” and rankings. This results in a correlation coefficient that’s different from ours, and also less accurate. For fun, we also measured the correlation coefficient of “EMD” using the Spearman rho, but did not report our findings. The result we got was a correlation coefficient of 0.1493. If you compare this number to other correlational studies done previously on “EMD,” you can see that the correlation is weaker now than in the past. However, had other firms analyzed their dichotomous data using the “Rank-Biserial” method, you probably would have seen much higher correlation coefficients than what they were able to report.

See these papers for a more in-depth explanation on correlations:

http://www.andrews.edu/~calkins/math/edrm611/edrm13.htm
http://www.napce.org/documents/research-design-yount/22_correlation_4th.pdf
http://coedpages.uncc.edu/cpflower/rsch8140/review/correlation.htm

Our gut instinct has been telling us for some time that social media factors have become much more important to rankings. This study proves it. As you can see here, Google +1s for the page have a lot to do with getting a website to rank well. More than anything, our study seemed to indicate that Google’s algorithm updates are forcing internet marketers to do the types of things that “real marketers” should be doing.


While planning which backlink information to collect for our study, we decided we didn’t want to just measure the number of backlinks for each of Google’s search results, but instead thought it would be more informative to collect a variety of backlink features (i.e., the type, origin etc.). We did this for the ranking page, the ranking domain, and the ranking subdomain. The resulting correlation coefficients for the backlink factors of both the domain and subdomain are illustrated below:

Backlink Ranking Factors Image

The highest correlated page backlink factor was the number of domains linking to a page that contain the entire keyword in the anchor text. This variable was closely followed by the number of links pointing to the page that contain the entire keyword in the anchor text. This factor included multiple links from the same site.


It’s becoming increasingly important for internet marketers to measure and understand the influence of social media factors on search engine rankings. Our study showed a consistent positive correlation between social media metrics and Google’s search engine rankings. Google +1’s, a relatively new social media metric, proved to be our most highly correlated social media factor on both the page and domain level. The resulting correlation coefficients for both are illustrated below:

Backlink Ranking Factors Image

On-page factors accounted for 421 of the 491 explanatory variables used in our study. This number is astounding when you consider that only 34 made it to the final results, meaning most of the correlation coefficients we calculated were not statistically significant.

Not only were most on-page variables insignificant, but they were also far less correlated than off-page factors. The resulting correlation coefficients of our 34 on-page factors are illustrated below:

Backlink Ranking Factors Image

Overall, it appears that on-page factors have a much lower correlation with Google’s search engine rankings than do off-page factors. The biggest surprise in this segment (and one that made us check our work a dozen times) was the positive correlation we found when the keyword being searched was discovered in the ranking page’s meta keywords tag. We plan to follow up more on this finding with a more in-depth analysis later.


On every level of our study, Moz scores consistently showed a relatively high positive correlation to Google’s search engine rankings. This serves as a testament to Moz’s overall expertise in this area and validates the credibility of their specialized tools and methodologies. The resulting correlation coefficients for these scores are illustrated below:

Backlink Ranking Factors Image

As discussed in the introduction, our team wanted to look for variances in the correlations we found by segmenting our data into several logical groups. By validating this suspected variability, we could easily justify more in-depth studies, which hopefully result in a much better understanding of how the Google algorithm, or algorithms, are employed.

We decided that keyword segmentation by search volume made the most sense, so that’s what we did. Just to be clear though, no one should be changing their search engine strategies based on these findings. They’re certainly interesting, but there are too many factors at play to know what’s really going on here, and this is only our initial look into this phenomenon. Please do not think we are suggesting that Google uses different algorithms based off of the results below. It is just as likely that several other factors may be causing the variances we found. The resulting correlation coefficients for domain, off-page variables have been segmented and illustrated below:

Backlink Ranking Factors Image

In contrast to the correlations on the domain level, the page level correlations exhibited the shape we expected to see in terms of a graduated scale based on the traffic of the keyword. What does this mean? We have some interesting theories as to what is causing this phenomenon, but we’ll save that for a future paper.


For several years it’s been our belief that Google utilizes several filters to detect both web pages and websites with unnatural backlink and social media profiles. One of the easiest ways to find these unnatural results is by ratio analysis and probability distribution. Since our study already gave us all the information we needed to do our own ratio analysis, we decided to explore this area, and compare those ratios to Google’s search engine results. What we found was very interesting.

The graphs below illustrate the correlations of various ratios to Google’s search engine rankings. In each graph we used the title of each as the denominator for each ratio and the individual variables below as the numerators. For example, for the variable “# of Page Facebook Likes” in the first graph, we computed: “# of Page Facebook Likes” / “Page Authority” first. Next, we used our resulting ratios to calculate a correlation with Google’s search engine rankings.

Backlink Ranking Factors Image

Because previous studies on Google’s ranking algorithm have not concluded whether or not social media metrics are causal, we initially expected smaller correlations using “Facebook Shares” as our denominator. What’s even more interesting is the disparity between “Facebook likes” and “Facebook shares.” In the last graph, we found this ratio to be the most negatively correlated ratio with Google’s search engine rankings. This result seems to indicate that certain social media factors may be more than just correlated with Google’s search rankings. You decide.

satisfaction guarantee

FREE Google Ranking
Consultation


A Netmark representative will be in contact with you within 24 hours

Top