Site Strategics CEO Erin Sparks and Digital Media Director Tom Brodbeck talk to the movers and shakers in digital marketing during each episode of their award-winning EDGE of the Web podcast. In Episode 303, special guest Russ Jones, Principal Search Scientist at Moz shares details about the new and improved Domain Authority 2.0 tool that helps digital marketers predict how websites will rank in Google search results relative to similar sites.
Russ Jones: His Background and Experience
When Russ Jones discovered SEO, he took himself off the lawyer track (his original career path) and never looked back. For a decade he worked at Virante, which later rebranded itself into Angular and more recently rebranded itself into Hive Digital. During his time at that company, he was the Chief Technology Officer. When he sold software to Moz that he had created, he joined the company as its Principal Search Scientist. In that role, his primary responsibilities involve the development of new metrics, proofs of concept, and feature consulting in terms of what Moz should be rolling out next for its customers. A core discipline at Moz is the ongoing evolution of SEO in terms of tactics and recommendations. Toward that end, there a whole team of subject matter experts whose sole responsibility is to stay on top the industry and innovation. And if you look at Search Engine Journal’s 140 Top SEO Experts You Should Be Following in 2019, six of the people on the list are from Moz, and no other organization has that many SEO experts on the list. Moz is the kind of company that gives back to the industry and community through its research, which is a point of pride for the company.
Moz Domain Authority: The What and Why
Domain Authority is a tool to provide a score from 0-100 to measure the strength of a domain to rank on Google compared to similar domains. There was a time when everyone could rely on Google’s PageRank metric to tell them about the strength of their pages, right there in its toolbar. But when Google stopped calculating PageRank or at least not disclosing it publicly (and later removed it from toolbar), it created a vacuum in the SEO industry for a good metric that webmasters could use to gauge site strength and predict if their site would rank in Google.
Moz already had similar tools like Moz Trust and Moz Rank, but both were static, not trained against anything, and did not involve any machine learning. The innovative move with Domain Authority was to take a site’s link information (number of links, types of links, quality of links) run it through a machine learning algorithm to predict the likelihood that a domain would rank. It’s not just how many links a site has, but whether or not those links contribute to ranking.
Right now, Domain Authority is requested over a billion times a year. It’s used for far more than just predicting rankings. Users often use it to determine which sites to do outreach to, they determine which domains they might purchase if they’re creating a new company, and what brands they might go after, and they use it to determine how well they’re performing compared to their competitors. There’s a rich array of use cases coming from one little metric.
The key is that Domain Authority is trained against search engine results pages. Google gets tweaked and updated constantly, but as long as Moz re-trains its model against the most recent Google algorithm, there is a high level of confidence in Domain Authority’s predictive capability. Domain Authority is going to go up and down depending on what any given site is doing (or not doing) relative to its competitors that helps it achieve better rankings.
Domain Authority 2.0: Fundamentally Changing How the Data is Trained
In the Domain Authority 2.0 release, Moz is fundamentally changing the way in which it trains the data. The way most of these tools work is that they consider a number of features or variables, like how many links there are to the domain, how many unique linking domains, how many are site-wide, and so on. Then after collecting a lot of search engine results and running them through a machine learning process, out pops a model that tells you, based on the link metrics, the likelihood a domain will rank for a particular keyword.
For the Domain Authority update, Moz took the time to question all of its assumptions and found implicit bias in the methodology. The best way to understand this is with a sports analogy. If you wanted to measure who is the best NBA basketball player, you could look at a number of relevant statistics such as points scored, number of rebounds and number of assists within a data set that compares all NBA players to those who became all-stars. The people who scored the most points and rebounds became all-stars. But what that approach overlooks are other key things that go into being successful in the NBA – like height. The training set is biased because everyone who manages to get into the NBA is tall. You’re probably not even going to get into the NBA if you’re less than six feet tall, but the model is missing it as a factor.
How does this apply to domain authority? It helps explain why there would be unexplainable anomalies, like a domain with a high Domain Authority score even though it doesn’t rank for any keywords at all! To correct for this, Moz includes a selection of sites in the training set that don’t rank for any keywords so the machine learning algorithm learns not only what factors are responsible for allowing a site to rank, but for allowing the site to even get in the game.
Connect with Russ Jones and Moz
Twitter: @rjonesx (https://twitter.com/rjonesx)
LinkedIn: https://www.linkedin.com/in/russjonesmoz
Blog: http://www.thegooglecache.com
Moz on Twitter: @Moz (https://twitter.com/Moz)
Moz on Facebook: @moz (https://www.facebook.com/moz)
Moz website: https://moz.com
Why Site Strategics Produces EDGE of the Web
Site Strategics is an agile digital marketing, which means it helps companies improve their digital marketing results based on data. Its EDGE of the Web podcast is a way to help everyone, including themselves, keep up with the ever-changing digital marketing landscape in terms of the concepts, strategies, tactics and developments that have a direct impact on digital marketing success. Your company can even become a sponsor of the EDGE podcast. Visit our sponsorship page for more information!