The term “Ranking Factors” refers to the criteria used by search engines to evaluate web pages and determine their placement in search results. These factors can be related to content, technical details, user signals, backlink profiles, and other relevant features. Understanding these factors is essential for effective search engine optimization.
Since 2012, Searchmetrics has released an annual “Ranking Factors Study” that involves extensive work from statisticians, mathematicians, and search experts worldwide. This study helps identify what influences website rankings. But what exactly is a “Ranking Factor”? How is it analyzed? What is a correlation? This report aims to clarify the data and its interpretation.
Note: A high correlation does not necessarily mean it’s a ranking factor!
Each year, we examine the top 30 search results for 10,000 keywords, equating to about 300,000 URLs, in our Ranking Factors study. We evaluate various factors, such as the number of backlinks, text length, and keyword content features. We aim to understand what sets higher-ranking pages apart from lower-ranking ones. Do they have more backlinks, text, or keywords?
We calculate the “rank-correlation coefficient” using the Spearman correlation to analyze the relationship between rankings and the presence of specific factors. This coefficient shows how closely two variables are related, such as a page's rank and the presence of a factor.
Differences in search result positions compared to a studied value can be visualized in a graph as a curve of average values for each item.
In the graph, four example correlations and their respective curves are shown:
The y-axis shows the average value for all 10,000 URLs studied at position X (x-axis). Factors with a “zero” value suggest no measurable correlation with Google results. A higher correlation value indicates significant differences between positions. Negative values can often be interpreted positively by considering the opposite statement.
The larger the differences from position 1 to 30, the higher the correlation value. Average values are always used for interpretation. For example, factors B and C have the same correlation value (1) but differ in their curves. Factor A’s average value is 95 for each position, but could be 5, with the correlation remaining 0. This highlights the importance of understanding both correlation values and their interpretation.
Search engines evaluate websites based on topic and relevance, structuring pages in the search engine index for optimal user query results. These evaluation criteria are known as ranking factors.
The vast number of internet documents necessitates an automated algorithm for ranking pages, despite human quality raters. The algorithm is a critical business secret, as public knowledge of ranking methods could lead to manipulation. Only Google knows the true Ranking Factors, but we analyze data to identify possible factors and their weightings, providing a solid foundation for our conclusions.
Initially, Google ranked pages based on keyword frequency, leading to keyword stuffing practices for high SERP positions, even for irrelevant pages. This competition between search engines and SEOs gave rise to the ranking factor myth. Semantic search development introduced technical and non-technical criteria, evolving ranking factors into a complex feedback system aimed at delivering relevant content.
A sustainable strategy focused on relevant quality factors ensures long-term success, emphasizing relevant content and combating spam and short-term tactics.
Searchmetrics analysis is grounded in data interpretation, not speculation. We evaluate web properties with high search result positions to identify factors that might influence rankings, but correlations do not imply causation. While we derive a list of factors from data, correlations do not guarantee impact on rankings or usage by Google. Our role is to interpret these correlations.
Logical fallacies can lead to false assumptions. For instance, increased social signals alone don't guarantee better rankings. Examples like storks and birthrates or ice-cream sales and sunburn illustrate illusory correlations. We avoid such fallacies through careful evaluation and robust database use.
Our analysis uses a large set of 10,000 keywords for Google U.S., excluding navigation-oriented keywords to avoid distortion. The database includes the first three organic search results pages, with over 90% consistency with prior years. This ensures a reliable comparison base while incorporating new high-search-volume keywords.
Our up-to-date database incorporates current keywords like “Samsung Galaxy S5” or “iPhone 6” for relevant analysis.
Factors are categorized as binary or numerical, impacting interpretation. Binary factors, like meta descriptions, are present or absent, while numerical factors can vary. Numerical factors often provide more significant insights for correlation studies. For binary factors, specifying average values supports correlation validity.
Correlation values are calculated for all feature data, with graphs showing 30 values per keyword. Average values exclude the top five percent to ensure smooth curves, allowing for clearer interpretation of median and mean values.
The “brand factor” is a notable data feature, where high-profile brands often secure top rankings despite less optimization. Brands might lack h1-tags or have shorter content, yet have more backlinks and social signals. Google effectively identifies and ranks brands based on user trust and recognizability.
Even Google may not fully understand its algorithm’s makeup due to its complexity. Searchmetrics “Ranking Factors” studies offer a methodological analysis, providing the online industry with a data toolbox for informed decision-making based on extensive research.