DA/PA Fluctuations: How to Interpret, Apply, & Understand These ML-Based Scores

[randfish]

Howdy folks,

Every time we do an index update here at Moz, we get a tremendous number of questions about Domain Authority (DA) and Page Authority (PA) scores fluctuating. Typically, each index (which release approximately monthly), many billions of sites will see their scores go up, while others will go down. If your score has gone up or down, there are many potential influencing factors:

1. You've earned relatively more or less links over the course of the last 30-90 days.
   Remember that, because Mozscape indices take 3-4 weeks to process, the data collected in an index is between ~21-90 days old. Even on the day of release, the newest link data you'll see was crawled ~21 days ago, and can go as far back as 90 days (the oldest crawlsets we include in processing). If you've done very recent link growth (or shrinkage) that won't be seen by our index until we've crawled and processed the next index.
2. You've earned more links, but the highest authority sites have grown their link profile even more
   Since Domain and Page Authority are on a 100-page scale, the very top of that represents the most link-rich sites and pages, and nearly every index, it's harder and harder to get these high scores and sites, on average, that aren't growing their link profiles substantively will see PA/DA drops. This is because of the scaling process - if Facebook.com (currently with a DA of 100) grows its link profile massively, that becomes the new DA 100, and it will be harder for other sites that aren't growing quality links as fast to get from 99 to 100 or even from 89 to 90. This is true across the scale of DA/PA, and makes it critical to measure a site's DA and a page's PA against the competition, not just trended against itself. You could earn loads of great links, and still see a DA drop due to these scaling types of features. Always compare against similar sites and pages to get the best sense of relative performance, since DA/PA are relative, not absolute scores.
3. The links you've earned are from places that we haven't seen correlate well with higher Google rankings
   PA/DA are created using a machine-learning algorithm whose training set is search results in Google. Over time, as Google gets pickier about which types of links it counts, and as Mozscape picks up on those changes, PA/DA scores will change to reflect it. Thus, lots of low quality links or links from domains that don't seem to influence Google's rankings are likely to not have a positive effect on PA/DA. On the flip side, you could do no link growth whatsoever and see rising PA/DA scores if the links from the sites/pages you already have appear to be growing in importance in influencing Google's rankings.
4. We've done a better or worse job crawling sites/pages that have links to you (or don't)
   Moz is constantly working to improve the shape of our index - choosing which pages to crawl and which to ignore. Our goal is to build the most "Google-shaped" index we can, representative of what Google keeps in their main index and counts as valuable/important links that influence rankings. We make tweaks aimed at this goal each index cycle, but not always perfectly (you can see that in 2015, we crawled a ton more domains, but found that many of those were, in fact, low quality and not valuable, thus we stopped). Moz's crawlers can crawl the web extremely fast and efficiently, but our processing time prevents us from building as large an index as we'd like and as large as our competitors (you will see more links represented in both Ahrefs and Majestic, two competitors to Mozscape that I recommend). Moz calculates valuable metrics that these others do not (like PA/DA, MozRank, MozTrust, Spam Score, etc), but these metrics require hundreds of hours of processing and that time scales linearly with the size of the index, which means we have to stay smaller in order to calculate them. Long term, we are building a new indexing system that can process in real time and scale much larger, but this is a massive undertaking and is still a long time away. In the meantime, as our crawl shape changes to imitate Google, we may miss links that point to a site or page, and/or overindex a section of the web that points to sites/pages, causing fluctuations in link metrics. If you'd like to insure that a URL will be crawled, you can visit that page with the Mozbar or search for it in OSE, and during the next index cycle (or, possibly 2 index cycles depending on where we are in the process), we'll crawl that page and include it. We've found this does not bias our index since these requests represent tiny fractions of a percent of the overall index (<0.1% in total).

My strongest suggestion if you ever have the concern/question "Why did my PA/DA drop?!" is to always compare against a set of competing sites/pages. If most of your competitors fell as well, it's more likely related to relative scaling or crawl biasing issues, not to anything you've done. Remember that DA/PA are relative metrics, not absolute! That means you can be improving links and rankings and STILL see a falling DA score, but, due to how DA is scaled, the score in aggregate may be better predictive of Google's rankings.

You can also pay attention to our coverage of Google metrics, which we report with each index, and to our correlations with rankings metrics. If these fall, it means Mozscape has gotten less Google-shaped and less representative of what influences rankings. If they rise, it means Mozscape has gotten better. Obviously, our goal is to consistently improve, but we can't be sure that every variation we attempt will have universally positive impacts until we measure them.

Thanks for reading through, and if you have any questions, please leave them for us below. I'll do my best to follow up quickly.

[Deacyde]

This still doesn't explain why I read that the indexing was less than desirable this time around and then saw a drop in DA ( BY 10 Points! ) While I've improved the SEO of our site, now I'm reading the indexing went great and the Drop in DA is basically our fault. Not sure what's going on, but still, doesn't make much sense, as I've seen half of my backlinks drop from being shown but have verified them still being there, manually.

So even in regards to the above mentioned, seems we're missing data that used to be there, and that this is the result of the drop. Bigger sites didn't see a drop least in regards to sites like EPA and BBB, however the smaller sites did see a drop in DA.

Just trying to understand all of this correctly and not speculate, can you bring me up to speed on what I may be misunderstanding?

Thank you

[randfish]

Hi Deacyde - that shouldn't be the message you're taking away from this post at all! As I noted above, you could improve the SEO, improve your link profile, and still see a reduction in Domain Authority as as a score due to how DA is done (on a relative scale, not an absolute scale). We could find more and better links to your site, and you'd still see a lower DA.

If a number of smaller sites in your field have all seen lower DA score in this index, that's indicative that it's nothing you've done, but rather, an indication that DA has been shifting across the board. If you're seeing rankings stay high and organic search traffic stay high, then there's nothing to worry about, and DA should still work just as well as a relative metric across sites (actually, slightly better given the improved correlations).

[Deacyde]

Thank you for responding, again wasn't upset, but was scratching my head as to the various factors that represent the overall DA score. Thanks for schooling me further and have a great weekend.

Now I know & knowing is half the battle...

[randfish]

Always happy to help, and especially to provide transparency. Thanks for the kind response

[DonnaDuncan]

Rand,

It's very hard to repeatedly explain a 30-40% fluctuation in an exponential metric with savvy clients. I'm finding the metric can't be used, especially with smaller websites or competing sites with large differentials in link or page counts. Will the long term efforts Moz has underway address this concern? Is there a ceiling in terms of website size that you don't recommend using the metric under?

I know you keep getting the same questions over and over. That's perhaps part of the reason.

[randfish]

Hi Donna - yes, that fluctuation should be much larger on average in the tail of the web (sites with DA 0-40) than in the middle or head. This makes sense because with a relative metric, all of the factors I describe above are going to be magnified in the tail, particularly because Google's rankings change so much there and because just a few links can have such a huge impact. For metrics-savvy clients, they should be best poised to understand that since DA/PA are exponential, a few links here or there and a few valuation shifts on those links can have big swings in the 10-40 point ranges of DA/PA, whereas in the 50/60+ ranges, small shifts in link discovery or in link valuation (from us or Google) won't have as much change.

As far as a ceiling - no, we don't have a recommendation there. The idea is that as DA/PA fluctuate, especially as they get more accurate in predicting rankings (correlations & coverage), the fluctuations are generally happening because Google's changing and we're getting closer to tracking how and where (with exceptions I noted above around issues with our crawl/indexing). My biggest recommendation is to keep track of similar-sized competitors (and larger/smaller ones) so you've got a set of benchmarks for comparison.

Thanks for the note and apologies for the frustrations this causes.

[DonnaDuncan]

Putting a range around the DAs where fluctuations will be most felt (10-40), helps and makes sense, but I also saw 3-15 point differences in DA this update at the higher end of the scale.

[Ria_]

This answers sooo many of my questions. Thanks, Rand!

[Andrew_Birkitt]

Thanks for clarifying this Rand it is a question that often crops up from savvy clients and you explain it so much better than I could

[KevnJr]

Rand,

Thank you for addressing the fluctuating DA/PA, to be clear: I have been disavowing bad links for 3 months using Marie Haynes post as a guide: https://moz.com/blog/guide-to-googles-disavow-tool.  I have disavowed 88 bad links during that period which I personally investigated and our DA dropped 6 points (31-25) with this last index.

Am I to understand that was because:

1. The index is reading from as far back as 90 days
2. We are in the 10-40 range and small link changes affect us more.

PS: I have been "cleaning up" the SEO using the Moz guidelines and adding quality unique content (products pages & blog posts)

Thanks,
Joe

[randfish]

Hi Joe - yes, it's most likely the fluctuations are because you're in that lower range. Remember that Moz can't see links you've disavowed in Google Webmaster Tools/Search Console, so we wouldn't be lowering DA/PA based on those (although, it's possible that, over time, as Google stopped counting links like that, our algorithm for PA/DA would "learn" and evolve, but that would take a while).

[KevnJr]

Thank you, I'm off to build a spreadsheet of my competitors PA/DA scores

Joe

[SarahWJATL]

You aren't the only one with the crazy fluctuations. The group of websites I monitor via Moz experienced DA changes from +3 to -17. The industry this cluster of sites is in tends to change mightily every time Moz refreshes - DA within the top players in the industry ranges from 20-65 on average. As long as the fluctuation is holding steady across the entire industry, we aren't seeing it as cause for alarm. We do see the same trend and same show runners each update.

My thoughts - as a whole the industry I am in is slipping behind the curve in keeping up with authority signals. The entire point of the tracking is so we can keep ahead of those things. Being proactive not reactive. Don't look at your domains as a single point. Get their competition in there too. You can see how the entire group is performing.

[SEMPassion]

When communicating with clients about progress I avoid talking about their PR/DA directly but talk in terms of improving (or not) in relationship to their competitors.  They really don't care about the NY Times, but about the sites that are directly competing against them.  If I can tell them that we closed the gap between their site and their competitors that is all they want to know. If a new index throws us a curve ball, we soften the blow by following up the bad news with a trend line showing that we are moving in the right direction even though there is an aberration in this indexes numbers..

Chris

[JosephGourvenec]

Rand,

This type of inconsistency isn't anywhere near good enough!

The metrics are not comparable to many months previous as we track DA and PA, I've spent the last week reviewing them.

When reporting back these drops look like anyone using your metrics are actually doing their job and we're all going backwards. Everyone above is is saying the same in a slightly more indirect point but this is a joke.

The credibility of the metrics of Moz mean everything in a digital strategy as they are new new PageRank metrics to showcase if you are actually doing a credible job for your client.

Despite everyone seemingly having fall backs, the issue here is again we all have to go back to our clients and say "Mr and Mrs/Miss invoice payer, we're doing a really good job doing all the things we need to do but your site statistics have gone backwards, not forwards. As we mentioned we use Moz because "We'll they we're credible" source of data they say that they have cleaned up their act due to a hidden bug in the system that was dormant until a series of issues occurred which they fixed. But Rand and their tech team say they've cleaned up their data and won't happen "hopefully" again anytime soon.

But we are doing a good job, promise please pay our bill even though we've had no update on stats for 2 months and then when we do it shows we're going backwards."

We'll in all honesty I'm very Pee'd off and yes I am having a rant here because you should have factored this in and some sort of levy.

I am just yet again disappointed with the out come and deflated with the results we're getting.

Sorry, but that's my truth and reality.

Cheers,
Joseph Gourvenec

[randfish]

Thanks for the feedback Joseph - I appreciate your transparency and can totally empathize with the frustration.

I think the key here, unfortunately, is in understanding and effectively explaining how the metrics of DA and PA operate and why they're not like standard counts that always go up as things get better. Clearly, we need to do a better job of that.

A good metaphor might be how rankings work for countries in various categories. For example, if Japan is ranked as having the world's best healthcare in 2015, and they improve the quality of their healthcare in 2016, are they guaranteed to still be #1?

Not necessarily.

Maybe the #2 ranked country improved even more and now Japan has fallen from #1 to #2 despite actually improving on their healthcare quality. Maybe countries 2-10 all improved dramatically and Japan's now fallen to #11 even though they technically got better, not worse.

PA and DA work in a similar fashion. Since they're scaled on a 100-point system, after each update, the recalculations mean that PA/DA for a given page/site could go down even if that page/site has improved their link quantity and quality. Such is the nature of a relative, scaled system. This is why I encourage folks strongly to watch not just PA/DA for their own pages/sites, but for a variety of competitors and sites in similar niches to see whether you're losing or gaining ground broadly in your field.

Hope that's helpful and wish you all the best.

[Christy-Correll]

I just have to say I love this particular metaphor. I will be using it frequently from now on when explaining how DA and PA work. Thanks so much, Rand!

[EmilyNudo]

We have seen a 10 point reduction in DA but this reduction seems to be in line roughly with our competitors, im just glad to see others have this and its not just us...

[adampetford]

So basically what you saying then is DA is not accurate and it should not really be trusted as an true measure of where my website in terms of Google rankings?