Is DA a reliable domain metric?

[Gyorgy]

I use DA and PA every day for reporting, researching and most of the time it's a pretty good metric to compare domains and pages but recently I did an experiment and I was surprised when I saw the results.

Two months ago I "link bombed" one of my old, unused websites with thousands of spammy blog comment links. Before the attack, its DA was 21 with a few hundred links.

In August, after the recent OSE update I checked the website again and I was quite surprised to see DA 61 as a results of 8340 links.

A DA value over 60 is considered pretty strong and it's interesting to see that spammy blog comment links could change it so significantly. Someone who doesn't know the history of the domain might get interested in advertising on such websites because the mozbar shows a high DA value.

I know it's difficult to algorithmically differentiate between spammy and valuable blogs but a future OSE update could focus on this issue.

Let me know what you think.

[Gyorgy]

In the meantime my colleague asked me how I would differentiate between spammy and valuable blogs.

My answer was:

These blogs are easily identified as spammy because there's a dodgy 200 word content and hundreds of comments. Every nick name is a link with "smart" anchor texts. The comments are only a few words long and these pages have roughly 4-500 links. ...so it's pretty easy to algorithmically identify these websites as link farm.

I used a black hat link scraper that uses a list of free-to-comment spammy blogs.

Combine the site analysis with this list and you have a nice negative list of domains.

[RyanKent]

DA is a very useful metric, but the tool's limitations must be understood. There are several:

The main thing to recognize when using DA is it's based upon the Linkscape crawl of the web. SEOmoz is working to improve the tool, but there are growing pains. For example take a look at: http://www.micrositez-seo.co.uk/

If I recall correctly, that site had a DA of 8x. Due to issues with the last crawl, the site's records were not captured correctly and it presently shows a DA of 1. Any issue with Linkscape will cause an issue with DA.

Also understand Google crawls every site, but Linkscape only crawls the top 25% of web pages. Google will almost always show more links then reflected in your DA.

Another important note, DA is mainly a measure of link counts based on the linking page's DA/PA. As you pointed out, Google will discount spammy links and eliminate their value. Linkscape has no way of knowing what links Google, Bing, Yahoo, etc have discounted, nor what sites have been penalized. The DA will continue to show link value.

The outcome of your test was predicable. If you add 100 links from JC Penney at the time when the site was under a penalty, the DA would have improved for the site receiving the links. The DA does not use "spamminess" or penalties as ranking factors.

In summary, yes DA is a very valuable tool in analyzing websites. Like most tools, it is important to thoroughly understand how the calculations are performed without the tool, then how the tool itself works. Using DA in combination with that knowledge and experience can make your job a lot easier. If you try to base your SEO decisions blindly on DA without the required understanding, you wont be happy with the results.

[Gyorgy]

I'm afraid you misunderstood my post.

I know that it's only a link metric but since DA and PA is becoming a standard in the SEO world, future updates should consider more variables in the calculation, because online advertising professionals often use and refer to the mozbar for site comparisons.

"DA is mainly a measure of link counts based on the linking page's DA/PA"

Probably it's more than that. The [DA specification page]( It uses a machine learning model to predicatively find an algorithm that best correlates with rankings across thousands of search results that we predict against.) talks about machine learning and that's exactly where this spammy site analysis could improve the calculation. Algorithmically it should be too complicated to filter spammy sites and lower the value of links coming from those pages.

This post is just a tip for SEOMoz' further improvements, because I do like the team and their tools. A software can be improved by continuos testing and by talking about the weaker parts.

[RyanKent]

Thanks for explaining Gyorgy.

I agree with you there is definitely room for improvement. I know the present focus is the Linkscape tool. If you have any specific suggestions for improving DA, then you may want to submit them to the help desk so they can be considered. The help desk staff does not regularly read the Q&A.

[seanmccauley]

The validity of PA and DA is something which must be of great interest to us all. I have recently looked at a spammy page with PR3 but PA of 72? Another aspect of Link Quality is relevancy, which cant be incorporated in to a metric, as relevancy only exists when the backlink has been built.. That would be interesting if a tool could look at a back link and incorporate relevancy in to an Authority metric. All in all PA/DA has been very useful and I look forward to future refinements.

[Gyorgy]

Do you guys know the Bad Neighborhood tool? It's a pretty simple tool and I use it to quickly check websites for links, dodgy anchor texts and anything suspicious. I would implement a similar method in the DA calculation.