Judging by our analytics and the volume of Q+A about mozRank and PageRank, I'd say a lot of you are applying metrics to your SEO. And I don't just mean search engine referrals. Given the economy, it's great if you can lay out some hard numbers, connect that to results, and make a strong argument for the work that you do.
I know we're seeing our SEO business continue to grow, not shrink, even in this harsh economic climate. I think a lot of that is because we're able to provide good numbers to back up our strategy recommendations. But even with the expertise on our SEO team and out in the community, I still see a few common questions about mozRank and PageRank, and what these mean for real-world SEO:
- What's the scale for mozRank? And why do I care?
- What does mozRank measure compared to Domain mozRank?
- How does mozRank compare to or differ from PageRank? Why should I use one or the other?
- What does PageRank really tell me about a page? How is it limited and what can I do with this knowledge?
These are great questions. We've got some discussion in our Linkscape help center around these, but it's a little technical and product focused. And we'd prefer to tell everyone (not just PRO members) something about SEO here.
If you'd like, you can jump straight to the takeways. By the way, when I say "PageRank" below I mostly mean Toolbar PageRank, the green fairy dust in your Google Toolbar.
First, on to scales. Both mozRank and PageRank (both academically and in the real world) have very few pages at the top (mR/PR 10) and many, many pages at the bottom (mR/PR 1). Because there's such a big disparity here, both of them have a handy 10 point scale, as illustrated below.
You'll notice the Y-axis is showing a hypothetical Link Juice metric on a log scale. So where we have mozRank 4, you'll see that corresponds to a hypothetical link juice value of 599, and mozRank 5 corresponds to 5,000. This just reflects the relative effort to get these mozRanks, whether this be links, authoritative endorsements, etc.
Take a mozRank or PageRank 5 page: one point above PageRank 4, one point below PageRank 6: one point in both cases. But the work you put in to go from 4 to 5 is quite different from the work you do to go from 5 to 6. For beginners it's frustrating to hit a particular level of PR or mR and feel like you've plateaued. When this happens, it's time to sharpen your pencils because you've got to break out some new techniques.
FYI, we set mozRank so that each level is ~8 times as much link juice as the prior level. To show this in another way, here's the same graph, but we've taken out the fancy scaling and just show the gradations between mozRank 5 and 7.
Suddenly you can see the real difference in SEO effort between a mozRank 5.08 and a mozRank of 6.61, and the work left until mozRank 7. Show this the next time someone gives you a hard time about that link building effort you're making. Or better yet, the next time you're link building, be sure to measure where you are today, and choose the right link building tactics. What worked to get you to mozRank or PageRank 5 just isn't going to cut it if you're going after that elusive 7.
When people say site PageRank, they're really talking about the link profile of the whole domain: what other sites are linking back? mozRank and (the original academic form of) PageRank both measure only links between pages. This ignores any factors about content, anchor text, domain age, authority, or trust. Domain mozRank and the concept of site PageRank are both interested in only links between sites. This still ignores factors about content or anchor text or domain age. While mozRank is scoped at the page level and measures reach by links to that page, Domain mozRank is scoped at the whole domain and measures how broadly the domain is referenced across many different domains. In this case, many links from a single domain don't help, but a few links from each of many different domains does help.
But typically one uses PageRank of the homepage to measure this. PageRank doesn't do a bad job of this, but it's not directly measuring this effect. Inside Linkscape (and exposed on the mozBar), we show Domain mozRank, which does directly reflect this, on the same kind of a scale described above.
Each level of Domain mozRank is about five times the juice of the prior level. This reflects the fact that there are many more pages than domains. But you get the same issues trying to jump from DmR 5 to 6 compared to 4 to 5. Getting more links from the same domains already linking to you isn't going to help your site-wide link profile. So if you're stuck at DmR 5, it's time to reach a little more broadly, engage in some new communities, and partner with some new sites.
So what about PageRank? Why do I keep talking about mozRank if Google isn't using it in their algorithms? That's a very valid question. We are confident, and plenty of expert SEOs agree that Google cares about links. They care about links from authoritative domains more than links from non-authoritative domains. And once you've gotten those links, the links you give out count for more. This is exactly the intuition we capture in mozRank and Domain mozRank. In fact, we've done a lot of studying and comparing mozRank and PageRank and we've found something really encouraging, and something a bit surprising.
* I've included a small amount of noise in PageRank (+/- 0.5) because PageRank is only provided with 10 gradations (e.g., PR 5 or 6 but never 5.34). This causes bunching in graph, which makes interpretation difficult.
This graph visually shows how mozRank compares to PageRank. The x-axis represents toolbar PageRank* of a page, and the y-axis represents mozRank for the same page. I've included the line y=x, which shows what perfect correlation would look like. For you stats junkies the Pearson's correlation coefficient is 0.48, which is good, but not perfect correlation.
We're pleased with this correlation. But by PR 4, mozRank starts to fall below PR, in some cases by at least a point. Our rule of thumb is that mozRank should be within a point or two of PageRank. This gets at data to support a belief many of you have had for a long time: Toolbar PageRank is correlated with site-wide authority and trust effects, beyond just page-level links. This can make things difficult for the metrics driven SEO: how can you measure your current position, and progress against different ranking factors, when the metrics you've got combine effects?
The Pearson's correlation between Domain mozRank and PageRank of the homepage of the domain of 0.71. This is a much more significant correlation than the page-level correlation between mozRank and PageRank. This time, we see much more significant clustering around the perfect correlation line y=x. And this time we see much less of the underestimating we saw with page-level mozRank. This suggests that Toolbar PageRank is showing several factors, including page-level linking, but also site authority and trust. And those factors are combined into a single score. Using PageRank alone can leave plenty of question marks about your strengths and weaknesses.
For the metrics driven SEO, this implies a few things:
- A high Toolbar PageRank for a page might not indicate a widely popular page. In fact, the page might be very lightly linked to, but might instead be reaping the rewards of being on a strong domain (e.g., some Wikipedia pages).
- Analyze the profile of the whole domain during the initial audit process, and not just specific pages. A new or unknown page might receive a high PageRank just by being on a strong domain. PRO members can try out the labs backlinks analyzer and choose "root domain" or "just this page" to see these two profiles.
- Work on site-wide performance, and then focus it. Gain authority for your whole domain, then focus that strength through link sculpting, on-page key word factors, and anchor text.
- Use fine grained metrics. Where appropriate, metrics like mozRank and Domain mozRank along with some comparisons to the competition can give an audit some powerful, targeted conclusions about strengths and about what is missing. We're certainly doing a lot of this in our own consulting.
The online marketing space is filled with measurements: analytics, conversions, cost-per-click. A lot of SEO is something of an art requiring high-level expertise. But there's plenty of room for measurement here too. Check out your site profile, check out your strong and weak pages. Measure the authority of your site. Prioritize your work based on your known strengths and weaknesses. And show your stakeholders not just what you're doing but why, and how that's changing.
With the economy in the shape it's in, the people who can measure their work and validate their assumptions are the people who are going to survive. And they're not just going to survive, but they'll thrive as they pick up the pieces the rest of us leave behind.
Nick - just to be clear: Are you saying that the toolbar PageRank Google reports for the homepage of a site appears to be something more like Domain mozRank - measuring more than just page level link popularity?
Also- based on the fact that mozRank is on a base 8(ish) log scale, is it safe to assume that toolbar PageRank probably has a similar scale? I know folks have been speculating about that since I first got into SEO in 2002/3.
I'm saying that PageRank for all pages across a domain may (see Darren's comment blow) incorporate site-level factors, not just page-level factors.
As for the base, we actually have fairly good correlation between mozTrust and PageRank as well. But for mozTrust our base is closer to 25. So I'm afraid the fact that we've got correlation here doesn't say much about the base.
But I'd love to hear discussion around this, and how knowing the base impacts real-world SEO.
Knowing the base would have some relevance.
If a broker offered Your 3 links , of Pagerank 3,5 and 7, Compared with 3 links at PR 6 and all other factors were equal, how would you decide?
A shallow curve of base 6 would lead you toward the 3 6's , which a steeper base, say 8 , would leave you in no doubt that the latter is statisticaly stronger.
I acknowledge that it's overly complex for 99% of SEO, but it's useful to know.
Brilliant post, great summary and fabulous advice Nick. Linkscape and its metrics continue to add so much insight into backlink and domain value analysis. To "go without" it now would be such a step in the wrong direction. Looking forward to your next post!
The Pearson's correlation between Domain mozRank and PageRank of the homepage of the domain of 0.71. This is a much more significant correlation than the page-level correlation between mozRank and PageRank. This time, we see much more significant clustering around the perfect correlation line y=x. And this time we see much less of the underestimating we saw with page-level mozRank. This suggests that Toolbar PageRank is showing several factors, including page-level linking, but also site authority and trust. And those factors are combined into a single score. Regarding the portion I've bolded above... I don't understand how you arrived at that conclusion, especially since you already said this:mozRank, and (the original academic form of) PageRank, both measure only links between pages. This ignores any factors about content, anchor text, domain age, authority, or trust. Domain mozRank, and the concept of site PageRank, are both interested in only links between sites. This still ignores factors about content or anchor text or domain age. Honestly, this post makes very little sense to me. And I spend 15-30 minutes a day reading patents, so if I can't follow it... the average readers are screwed. I'll try to explain exactly where the post lost/confused me... You seem to be saying: 1.) mR has a pretty good correlation to PR. 2.) DmR has an even better correlation to home page PR. ∴) Toolbar PageRank is showing several factors, including page-level linking, site authority, and trust... and those factors are combined into a single score. This conclusion seems to be making a lot of assumptions about how site authority and trust relate to a page vs. a domain vs. a home page. I just... um... what?
Darren,
You make a good point. I haven't formulated a very good proof here.
More explicitly, we assume we have a fairly good sample of link data that reflects the composition of a search engine index. Given that assumption, we've made some observations, and have some hypotheses.
Observations:
1) Page mozRank ignores domain-level features (such as links to a site's homepage or other pages)
2) Domain mozRank incorporates site-level features (and in fact ignores page-level features)
3) PR is better correlated with domain level mozRank than page-level mozRank
Hypothesis:
This, along with our above assumption, suggests (but clearly, as you've stated, does not prove) that PageRank might be looking at site-level features rather than just page-level links.
There's no QED or hard conclusion here :(
There's a lot of other reasonable explanations: our dataset might be poor, our algorithms might be weak. However this explanation jives with our own experience and trends I've seen in the SEO universe. Plus I believe our data set and algorithms are good :)
As far as patents and papers go, I suspect that Google, et al. is likely doing things differently from what their patents and papers lay out.
Having participated in a couple of patents, I can tell you that a lot of the work on a patent goes into covering methods you aren't using, and obfuscating things enough to give rise to the speculation you and I are engaging in :)
Well - knowing the base does give you insight into what it means when you (or a competitor) rises/falls in PageRank as well as a metric to tell you how difficult it might be to rise in that metric.
Thanks for the post - everytime you guys release data like this, I feel smarter and more informed about search operations :-)
I see, so you're looking for the 8x effort vs where you're at then?
In that case, somewhere between 8x and 25x is probably the right neighborhood.
For the average SEO out there, understand, there is potentially a vast gap between mozRank or PageRank 5 vs PageRank 7. While you might rise quickly from 3 to 4 to 5, by that point it starts getting tricky.
I tried to point out that you'll likely have to change tactics, perhaps several times as you go from 5 to 6 to 7.
3) PR is better correlated with domain level mozRank and page-level mozRankIs that supposed to say than?
yes :) comment edited. Thanks for the catch (again!)
Nick, great post, its always nice to see diagrams as they help the information sink in, my agency is focusing more around using mozbar for our link building campaigns because PR is being so inconsistent and the chance that sites are generating a fake PR, you can generate a fake MozRank...
Some Great takeaways here Nick.
Thanks for all of the juicy tidbits of information. Can't wait for the next set!
How often is mozRank updated?
We update the Linkscape index (which drives mozRank) about once a month. In the future we'd like to do this more frequently, and we'd like the updates to be seamless so it's always "just up-to-date".
Could you let me know if its still 1 time a month? Really curious on this as I am trying to keep a very close reporting eye on something.
Thanks
Yeah, we still do about one month updates. But it's really a schedule driven by how long it takes us to crawl and index the web.
That's not as predictable as it sounds though. Sometimes we turn up the volume and suck down more pages, and sometimes that means we've got more data to crunch. So it's been about six weeks since the last update at the moment.
We almost always do a post about that.
Nick, Okay... I think I have a better idea of what you're saying. Now I can proceed... to disagree with everything! =P If I'm reading this post correctly, your first correlation graph is comparing: x = Toolbar PageRank of any given page y = mozRank of that page ...and your second graph is comparing: x = Toolbar PageRank of any given home page y = Domain mozRank of that home page's domain If this is the case, then comparing the two graphs won't result in any suggestions or valid conclusions... because none of the graphed data remained constant. In order for us to speculate or draw any conclusions about the influence of site-level PageRank (i.e. "domain authority") on page-level PageRank, we would need both graphs to contain the same set of pages. As I understand it, this is not the case, as the second graph contains only home pages. Based on the data in these two graphs, the only conclusion/suggestion I see is: Most domains have a large percentage of their backlinks pointing at their home page.
Nick - correct me if I'm wrong, but it seems to me that your data showed a few things:
Thus, the conclusion is that our data confirms nothing, but lends some greater credibility to the concept that homepage PageRank may be more than just page-level link popularity.
So you're saying the Toolbar PageRank for a homepage is calculated differently than the Toolbar PageRank of non-homepages?
I think Nick's data suggests that "something" is different about either how PageRank is calculated on homepages or about how our crawl biases us to have different correlations with them vs. internal pages. Definitely something to investigate more :-)
Nick - I was so confused reading this post however I've taken away a lot from the subsequent debate between you Darren and Rand.
Thanks guys!
The most interesting thing would to connect this data to Google Penalties. But this is something nobody has even ever tried to do.
Darren,
I see your point about the number of assumptions. But I do think the graphs provide some compelling evidence that the assumptions have value.
Moreover, I think Nick's point is more about application for SEO tasks. As PageRank appears to be a conglomerate of page-level ranking, site authority, and trust, it's not sufficient to give you a game plan for improvement on your page(s) or site as a whole.
Linkscape separates the metrics so you can determine which elements you need to address. Determining whether the issue is domain strength, authority to a specific page, or trust can give you a much clearer idea of what to do to improve your rankings.
gillian, I appreciate the reply, but your comment tells me that I've failed to explain myself clearly. When I said the "conclusion seems to be making a lot of assumptions," I also should have added "...and I don't know what those assumptions are." In other words... yes, the graphs support the premises... but the premises don't support the conclusion. The most confusing part for me is the fact that the second graph is comparing DmR to the PR of a single page (the home page). The correlation between those two doesn't seem to imply anything about site authority and trust, as far as I can see.
Excelent post ajuda muito e relevância de poder saber mais sobre PA e DA e Page hanked
Like some of the recent posters, I'm seeing big differences between Mozrank and PageRank.
For example, I have a page that is over a year old. The current PageRank is 0, but the MozRank is 3.9.
I have almost the same issue. When i check the mozrank of my page, I see a rank of 6.15. But the pagerank is 3. This difference is to big. What could be the reason for this?
I'm having a hard time following (been a while since I've taken statistic-type courses) this correlation idea. For example, with one of my sites,
- opensiteexplorer reports: mozRank = 4.84, domain mozRank = 3.79
- google toolbar PR, after 1/20/11 update = 0
shouldn't both of the mozRank values be much lower?
This is a great write up. thank you.
I've been looking into a bunch of domains and was finding the similarity between pr and mozrank was very similar.
Its great to see what it represents.
Hi Folks, sorry for posting slightly offtopic. I have two questions where I am looking for an answer. I am happy to repost them to a message board or forum if required. Thank you for helping. Thumbs up to the SEO Moz people - great product!
https://www.seomoz.org/blog/whiteboard-friday-how-the-link-graph-works
https://www.seomoz.org/blog/testing-the-value-of-anchor-text-optimized-internal-links
Regards,
JC
"A high Toolbar PageRank for a page might not indicate a widely popular page"
Glad you covered this because I just don't get it. I have seen sites with very little content with a PR of 4 and hardly any pages listed in Google.
I think it proves the point that PR is an outdated indicator of a site's popularity and can't even be used as a guide anymore.
Great post - thanks for pointing out that Pagerank is an outdated indicator of link strength given the combine effects you speculate about it.
mozrank = BS
Incorrect:
mR = a more complete way of checking exactly where you are against a competitor rather than using the outmoded pagerank indicator on the G toolbar.
For instance, if we have a PR6 as do our main competitors, yet our mR(d) is better (ie. 6.8 as opposed to 6.1) its a decent measurement that our link profile is better, at the same time in the opposite direction, it helps to identify where you arent doing as well as competitors and analysing their inbounds you can plug your deficit more effectively.
Would you rather wait until the TBPR of your competitors is higher, then have your boss or client asking why their pr is better than yours?
---------
I would certainly rather be better informed.
mozrank = BS You disappeared for 18 months and then returned to say THAT? What a letdown. Wherever you were all that time... please go back.