Learn more about the new DA.
Posted: Thu Feb 13, 2025 10:17 am
I'm Russ Jones, Principal Search Scientist at Moz, and I'm excited to announce a major upgrade coming next month to one of the most important metrics Moz offers: Domain Authority .
Domain Authority has become the industry brazil number data for measuring the strength of a domain in terms of rankings. We recognize that stability plays a key role in making Domain Authority valuable to our users, so we wanted to make sure that the new Domain Authority brought meaningful changes to the table.
What is changing?
Below are some of the technical changes behind the new Domain Authority and why they matter.
Training set:
Historically, we’ve relied on training Domain Authority against a random, large set of search results. In fact, this has been standard practice across our industry. But we’ve found a way to fundamentally improve on this, making Domain Authority more reliable. Specifically, the new Domain Authority is better at understanding sites that don’t rank for any keywords than it was in the past.
Training algorithm:
Instead of relying on a complex linear model, we have switched to a neural network. This offers many advantages including a much more robust model that can detect link manipulation.
Model factors:
Domain Authority has become the industry brazil number data for measuring the strength of a domain in terms of rankings. We recognize that stability plays a key role in making Domain Authority valuable to our users, so we wanted to make sure that the new Domain Authority brought meaningful changes to the table.
What is changing?
Below are some of the technical changes behind the new Domain Authority and why they matter.
Training set:
Historically, we’ve relied on training Domain Authority against a random, large set of search results. In fact, this has been standard practice across our industry. But we’ve found a way to fundamentally improve on this, making Domain Authority more reliable. Specifically, the new Domain Authority is better at understanding sites that don’t rank for any keywords than it was in the past.
Training algorithm:
Instead of relying on a complex linear model, we have switched to a neural network. This offers many advantages including a much more robust model that can detect link manipulation.
Model factors: