Measuring Link Building – Whiteboard Friday

SEOs have strong metrics to measure the success of their strategies, such as domain authority (DA) and site authority (PA). But how best to use them? This Friday at the Whiteboard, Tom shows you how to think about these metrics as part of a holistic approach to your link building analysis.

Click on the board image above to open the high definition version in a new tab!

Video Transcription

Happy Friday, Moza fans, and today’s Friday on the whiteboard is dedicated to measuring link building. Obviously this is a very big and very old topic in the SEO space in which Moz as a company invests a lot, right? I like Domain authority in Authorization of the site are two of our very popular products that are commonly used just for this purpose.

Now, however, this will not be advertising. I could stand here and say that obviously these are the best metrics in the world and things like that. That’s not what I’m here for. I am here to give you some nuances on how and when to use these measurements and how to think about them and how to use them in addition to other measurements instead of having just one tool and saying that the solution to all problems is not necessary honestly.

Google’s PageRank

To do that, I’ll actually start by going back to 1998 and Google’s model of PageRank. Now I know that a lot has changed since 1998, both in the world and at Google. But that was Google’s original way of thinking about links, and in many ways it’s still the best thing we need to continue. Many current SEO best practices and dogmas are still based on this original understanding, only there are a few things we’ve somehow picked up along the way that don’t really have a basis in anything Google has said or done, so I want to point them out.

Thus, PageRank was originally a way of using links to estimate the likelihood that a user is on the site, and this is already quite interesting, as it shows that this is a model that is popular. So when we talk about it now, we often talk about things like trust and authority and things like that. I’m sure they’re important, but it’s worth remembering that originally this was just a way to effectively assess a site’s popularity.

Note that I also said for the page, nor for the domain. Imagine a world where there is one page on the internet, that is page A, which I have highlighted here. If there is one page on the Internet, it is not so difficult to assess the possibility of a random browser on that page. They are certainly on this side. If we introduce the other side, it’s still not that hard and we just assume it will be 50-50 and so on and so forth.

Probability of connection

This is a kind of basic probability that we have to work with. But then we can take some sort of tangent or a little spice to add to a situation where one side connects with the other, and that’s obviously what we’re actually interested in. So if A connects to that other page and at a time when there are still only two pages on the internet, ignore these other fields, they will come later, there are only two pages on the internet and A links to the other page.

We say that this probability is transmitted 0.85 times. Now 0.85 is a pretty arbitrary constant. It comes from an old Google document. Probably not such an exact value, but it’s fine for illustrative purposes and is the best we can continue.

So why did we say 0.85 in this case by the way? Why didn’t we say that all users click on this page? Well, that’s because we anticipate that some will go and make their own, stop browsing the internet, do something else. This damping factor has proven to be very important in a world where pages actually connect to each other in the big web and not just one link in one direction.

So everything is nice and right, right? What if we had a second connection and introduced a third party to the internet? So this is still a very simplified model. We have three-page internet and two connections, and the connections only go one way.

It’s very, very simple. But in this case, we are saying that both sides cannot get full probability. No, users do not click on both. They click on one of them. So we get half 0.85 A. But then this one too.

Again in a more complex model you could say, oh, one of these links is more likely to be clicked, so it gets more likely or something. But in this simple version we say that it is divided in two ways. Now, in this case, we’ve learned something interesting again, as adding another link has reduced the value of existing links, and that’s something we almost never think about in the context of link building.

But that’s what we think about when we talk in technical SEO conversations about not having too many links in the above navigation and the like. We try to direct our power where we want it most. Then, finally, I promise [indecipherable] it will stop soon. Finally, what if we had another leap in this system? Well, in this case, this 0.85, this damping is repeated.

So 0.85 times 0.85 is about 0.72, so it’s less. It is basically 0.85 times larger than this side above it, so it has become even lower. That’s why we as technical SEOs sometimes get caught up in things like chain redirects and things like that, why we think that’s important.

This is where such dogma comes from. So I’m not going to go any further with this kind of simplified explanation of PageRank. Here are a few things to point out to you. One is that there is a lot about the specifics of the page here, which affects the value of those links, such as the number of links the page has sent out, and also things like what is related to a particular page.

Note that I did not say anything about domains here. It can be on four different domains. It can be on one domain. We only talked about the peculiarities of the site here. Over time, Google has been somewhat hesitant about how they think about pages and domains. But they generally say they are interested in pages, not domains. So this is interesting, right, since everyone could be on the same domain, but this site could potentially be much weaker and carry much less power than this.

Measurements to build connections

So this is interesting, and this is something we don’t usually think about when building connections. So, if we go back to the topic of what I said I was going to talk about, the actual measurements for building connections, we’re looking for some properties.

Quick

Now, what I haven’t talked about are the first two. We want fast measurements. We want it to be available as soon as possible so we can report to our client or our boss or something like that, as well as being busy people. We don’t want to waste time.

Ubiquitous

We want measurements that are ubiquitous, so when I say to my boss, “Oh, I got you a connection that had a DA 90,” there’s a good chance he or she knows what that means. But if I say Tom Capper had a 38B rating, they’ll say, “What are you talking about?” Therefore, I need to use a measurement that is reasonably well understood.

Features of pages and links

But there are also page-level specifics and links that I just talked about. So when I think of a metric like domain authority, it works very well with these first two, and it’s fine with this third one because it’s somewhat qualified for rankings, which is some of what sets that up.

So there are some benefits. It takes some of these things into account, but in the end it’s a domain-level measurement. Therefore, by definition, it must treat all pages in one domain equally. This gives some advantages and disadvantages.

Sharing measurements

So what I want to do is want to put some measurements on such a chart and suggest how we could use them side by side.

I actually have it here as a vertical axis. So the closer it is to what we’re actually trying to measure, which is Google’s view of link value in essence, the longer it will be up. But then I also have this fast / slow measure of convenience. So a metric like domain authority is probably somewhere here. It’s very fast.

It is very ubiquitous. However, it lacks some of that nuance because the measurement is at the domain level and answers a slightly different question. YES is designed to answer the question, “How likely is it that a page on this domain, if all the same, will rank well?” This is a slightly different question than how valuable the connection is. But if I say, oh, I want YES, but not necessarily at the domain level, you could say, “Oh, well, Moz has a measure for that and you should know, and it’s called Page Authority.”

Well, yeah, that’s a good candidate. Like most industry-level metrics in the industry, including Google and our own, the Page Authority is initially informed by some domain-level and page-level factors. We did correlation studies and the like.

As you would expect, it is much closer to measuring the value and classification of a particular page than a domain authority, as it is a more accurate measurement and captures some of these nuances. But you can really go a step further. Now, page authority is slightly slower than domain authority because you have to wait for Moz to discover the page and search it.

We do our best, but not immediately. However, if you’re willing to wait even longer, you can use a metric like referral traffic. Sorry for my absolutely awful writing there.

So with reference traffic, we’re interested in how many people actually click on the link I built to my site. This is interesting because this is what Google was actually trying to measure. So if we can measure that, we’re getting closer to everything they wanted.

So, regardless of the sophistication they’ve built in, we kind of catch that shade. Now this has some obvious shortcomings. One is that a lot of link building campaigns are not good at this measurement, and you can judge for yourself. The second is that you will obviously have to wait quite some time for this data to be available, and even then there may be problems with customer analytics or something like that. Anyway, that’s what I wanted to share with you today.

Basically, I would suggest that you use all of these measurements and some others that could be ranked on this chart. So I wonder what measurements you would use and where you would draw them on this kind of chart. I gave these green lines as a sort of guide because I think you could look for a search in this first section, for example, before you even build a link, and then initial reporting to the customer.

Then this section would be more post-campaign when you want to learn from it and think about what kind of links you would build in the future and whether you would do the same again. But yes, I would love to hear your ideas. Thank you very much and good Friday.

Video transcription from Speechpad.com