Welcome to another edition of Whiteboard Friday. This week we're talking a little bit about Google PageRank. I know this feels a little retro for some of you. You're thinking, "PageRank. Oh my God, That's so 2001-2002." But it comes up a lot. It comes up again and again, particularly with folks who are new to the industry, new to SEO, or have heard a lot because Google does a lot of publicity around PageRank being their original algorithm. A lot of media publications talk about it. Even a lot of marketers still think of it as an important metric. So, I am here to set some of those myths straight and also talk about some ways to actually use and understand PageRank in the way that Google gives it to us, which is in the toolbar. So, let's start by talking about the difference between toolbar PageRank, which is what you can see via Google's toolbar or via every tool on the Web. There is no tool on the Web that calculates a different version of PageRank or that uses some other version of PageRank that they secretly get from Google that Google experts write. The PageRank number is always coming via the toolbar, and tools that report it use something called the checksum to get it from the toolbar the way that Google's toolbar pulls it directly from Google Source.
PageRank, toolbar PageRank, is updated randomly. By randomly, I mean, it's been as long as 11 months in some cases between a toolbar PageRank update, meaning that a site's actual PageRank might have jumped up a bunch or gone down a bunch, but Google won't change the metric that they show for 11 months, and then recently we had a PageRank update that only took about one month. It used to be that it would happen every month. Now it is very inconsistent. I think the last ordering was there was a month between the last update. Then it was six months before that. It was about five months before that. It was like three months before that. So it is inconsistent. It is a little random. It is when someone at Google decides to turn on that meter. So you don't really know when the next update is coming, nor do you necessarily know how recent your PageRank is, the PageRank that is being reported in the toolbar, unless you check when the last update occurred. So that's an important thing to notice.
Also to note, PageRank happens on a log scale. Let's talk a little bit about a log scale, what that means. So, for example, a log scale, this is a log base 2. What I have done is just drawn little lines to try and illustrate. So, zero, there is nothing there. One, I drew one little line. Two has two little lines. Three has four lines. It is doubling, the base is 2. So we're going up. You can imagine, this curve sort of looks like this until we get to PageRank of 10, which would be way, way off the screen, just super high up, and obviously my line illustrations are by no means perfect. So this should actually be double that. It is probably not.
PageRank does not use log base 2. We suspect here at SEOmoz because we calculate our own similar metric by crawling the Web, we calculate a metric called MozRank. MozRank is designed to mimic PageRank. In fact, it is very similar to the original algorithm that was published by Larry and Sergey back at Stanford. But what we've found is that when we need to scale it, the log base is not 2, but it is between 8 and 9. That means that a PR 7 for example would be 8 times, let's say 8.5 times, more important, more linked to, more well linked to than a PageRank 6 page. So there is a lot of variation in here, and that gradation is not observable through the PageRank that you get in the toolbar, which is a little frustrating. It is one of the reasons why when you see MozRank reported, MozRank will say something like 6.42. So you get a sense of, oh, okay, that's where I am in the scale. I am closer to a 6 than a 7. I am right about the midpoint. This helps you to just see those changes month over month since MozRank is reported consistently every month.
PageRank in the toolbar has a relatively low correlation to how well things actually rank. Meaning that if you were to take the toolbar PageRank of thousands of top 10, top 20, top 30 results and compare them and ask the question, "Do things with higher PageRank tend to rank better then things with lower PageRank," the answer would be yes, but only barely. In fact, the correlation is around 0.11, 0.12. It is pretty darn low, and it gets lower and lower the further away we move from the last PageRank update, which makes sense, right? As the scores degrade in freshness, so too does that correlation. You get this sense of like, wow, yeah, PageRank is probably a very small part of the algorithm.
What it is useful for, and what Google talks about it being used for, and not toolbar PageRank but real PageRank, which we'll talk about in a sec, is to help them determine which URLs on the Web to crawl and prioritize and recrawl. Meaning that, things that are more important, they tend to recrawl them more frequently, and PageRank is one of the items that goes into that calculation. I think almost certainly another one is how frequently they are adding new unique content that Google has never seen before, because they want to make sure that they are indexing that stuff.
So, real PageRank is actually a number that is calculated between 0 and 1. It uses a ton of trailing decimal points typically. It's updated multiple times daily. So you can hear Google's representatives talking about PageRank and they say, "Oh yeah, internally we are constantly recalculating and refactoring PageRank. In fact, we now have systems to estimate the calculation of PageRank so that we don't have to run it across the full web graph because it is very computationally expensive to run PageRank so many times." In fact, here at SEOmoz, MozRank calculations are one of the big reasons that it takes a long time for us to process and calculate an index. I think it is almost a day, day and a half, of processing across 50 to 60 billion URLs to try and make the MozRank calculations happen, which are similar to PageRank.
So, all of that theoretical stuff in mind, let's talk about actually using PageRank, where you should and where you shouldn't.
PageRank is useful as a raw indication of link popularity. Meaning that, PageRank doesn't take into account anchor text. It doesn't take into account whether the sites are related to each other. It doesn't take into account whether some of the sites maybe have been flagged for spam. It doesn't take into account whether the page is relevant to the actual search query. All it says is, in terms of the raw link popularity, the how many pages are pointed to this and how important the pages are that link to this, how important is this on the grand scale of the Web? Remember, the most important site on the web will be a 10, and everything else will fall back from that. So, as the most important sites on the Web gain tons of links, Facebook for example, Twitter for example, everything else is going to be scaled down a little bit. You might have seen in this recent PageRank update, the last couple, that a lot of sites took hits. Big important sites. Nasa.gov went down, Yahoo went down, Google.com fell from a 10 to a 9. That's crazy. That's amazing. What happens there is essentially the most important sites, this is my opinion, but the most important sites are essentially pushing out the boundary of what it takes to be a 10 and then those other sites are falling along the curve somewhere.
So, okay, good as a rough indicator of raw link popularity. It can be useful as well to compare it to other metrics. So, if I see something like a MozRank of a 6 and a PageRank of 3, I might get a little suspicious. There are two big things that could be happening in this type of an instance. If I see that MozRank is high, PageRank is low, I might think to myself, well, maybe since Google's last update to PageRank, this page has gained lots of links and the SEOmoz web index, Linkscape, has recognized that and is crediting them with higher MozRank, but Google, while they have recognized it internally, through real PageRank, they have not calculated it with toolbar PageRank because they have not done a push, an update, of that exported data and so we're just not seeing it.
The other explanation and the more dangerous one is when you see a site and you're like, boy, I think they might be selling links, I think they might be buying links, I think they might be manipulating the link graph in some way through some Web spam and you see those low PR numbers and maybe the page doesn't rank so well, but the MozRank is high, remember that MozRank does not have Google spam calculations in it. PageRank, Google will sometimes penalize sites and pages for selling links or buying links, and that's the way they sort of let you know, like, wink wink, "Hey, we know you're selling links, stop doing it."
The other thing is that the history of PageRank can be useful. So if you se that your site a couple weeks ago or a couple months ago used to be a 6 and now it is a 7, you can gain some insight into that. Especially when you are looking at penalizations you might say, "Oh man, this site used to be a PR 6 four months ago. Now it is a PR 2. Something is mighty suspicious." There is a tool on SEOmoz that is free. It is called Historical PageRank. You can search for it and you will find that. It can be very useful. I actually use and like that tool a lot.
Do not use PageRank for understanding why things rank well in the SERPs or don't. You look at a page of search results and you see, oh, the number one result, well, it must have the highest PageRank. No. Like, no, I can't describe to you how frustrating that is for anyone, for search engineers, for professional SEOs, for . . . anyway, that is not how the search results work. I know. Ten years ago, it used to be the case. It was sort of like use good things on your page and get a high PageRank and you'll rank well on Google. That's not the case anymore. I think PageRank is possibly responsible for, I don't know, sub 5% of the ranking algorithm today, and I think most Googlers would tell you that as well.
Do not use it for valuing link prospects. A lot of cases, I'll see someone say, "Oh, you know, I have this great opportunity. It looks like a great page. It is sitting at CNN.com/articles/012345, but it has a PageRank of 0 or 1. I don't really know if I want that link." Oh my God! You want that link! It's on CNN! What? Come on! PageRank of the actual page, please, use better metrics. Think about things like domain authority. Think about things like page authority. Think about things like, does that PageRank well, is it relevant, is it on a good site, is it useful, is it going to pass good anchor text, are good visitors going to come from there? Don't be thinking about PageRank as your primary metric for valuing link prospects. You can use it in the sense of, hey, let's see if Google might have penalized this page or site for selling links. Maybe that would be worth checking out.
Finally, don't use it as a key metric for reporting. PageRank is not a KPI for anyone and never should be. If someone is coming to you to do SEO and they say, "Hey, you know, we'd really like to improve our PageRank," say no, because PageRank will not send you more visits. It will not send you more business. It will not make your website better in any way. It will not improve your conversion rate. It won't bring you more Twitter followers. It is not going to make your life any better in any way. So please don't report it as a key metric and don't request that it be reported as a key metric. It is generally useless as a KPI. You might consider looking at it just to make sure that you haven't been penalized or that things aren't going terribly wrong. But if you see your PageRank go up by 1 or down by 1, look at your visits. Look at your traffic. Look at how many keywords are sending you referrals. Those are key performance indicators for SEO, not PageRank.
The last thing, possibly the thing that upsets me the most, because I think a lot of experienced SEOs should know this, and many don't. I am not quite sure where this miscommunication happens or why, but PageRank is for pages only. You'd think that would be obvious since it's in the name. Right? It's not called domain rank. It's called PageRank. I know, technically it refers to Larry's last name, Larry Page, not page rank. We're lucky Brin didn't use it or, I don't know, God forbid, I'd created Google. It would be called FishkinRank. That would be horrible.
But PageRank measures pages. Right? So, when you see someone say, "Oh, yeah, yeah, yeah, I have a PageRank 6 domain," you have my permission to call baloney on them. You can even use something stronger if you would like, because there is no such thing as a PageRank 6 domain. There is a domain whose home page may indeed have a PageRank score of 6, but that is measuring that home page's link importance and that home page's PageRank, not the importance of the domain as a whole. So, you will find dozens, hundreds, thousands of sites who have a PageRank on their home page that is smaller than what their internal pages might have. For years, SEOmoz had a PageRank of 5 on our home page, but we had a PageRank of 7 on one of our articles and PageRank of 6 on another one. It is not describing how important your domain is. It never is. There is no such thing as domain PageRank.
If you would like, you can use a metric called Domain MozRank, which essentially calculates PageRank over the domain level link graph, meaning it consolidates all pages and looks at domains that point to each other. We think this is something the engines do as well. There are patents and scientific papers and research papers that are written for conferences that suggest that something like that exists, but it is not domain PageRank. Or rather, Google does not report anything that is domain PageRank, and anyone who says otherwise should watch this video.
All right everyone. Thanks so much. Hope you enjoyed this edition of Whiteboard Friday. We will see you again next week. Take care.
Looking for information on some of the other metrics you can use for analyzing links? Check out our earlier series on the topic here: Which Link Metrics Should I Use? Part 1.
