Weighting the Clusters of Ranking Factors in Google’s Algorithm – Whiteboard Friday
Posted by randfish
One thing we collect for our semiannual ranking factors survey is the opinions of a group of SEO experts (128 of them this year!) about the relative weights of the categories of ranking factors. In other words, how important each of those categories is for SEO relative to the others.
In today’s Whiteboard Friday, Rand explains some key takeaways from the results of that particular survey question. In addition, the pie chart below shows what the categories are and just where each of them ended up.
Whiteboard Friday – Weighting the Clusters of Ranking Factors in Google Analytics
For reference, here’s a still of this week’s whiteboard and a fancy version of the chart from this week’s video!
Video Transcription
Howdy, Moz fans, and welcome to another edition of Whiteboard Friday. This week I’m going to talk a little bit about the ranking factors survey that we did this year and specifically some of the results from that.
One of my favorite questions that we ask in our ranking factors survey, which happens every two years and goes out to a number of SEO experts. This year, 128 SEO experts responded, sort of folks who were hand chosen by us as being very, very knowledgeable in the field. We asked them, based on these sort of thematic clusters of ranking elements, things like domain level link authority versus page level keyword agnostic features, weight them for us. You know, give a percentage that you would assign if you were giving an overall assessment of the importance of this factor in Google’s ranking algorithm.
So this is opinion data. This is not fact. This is not actually what Google’s using. This is merely the aggregated collective opinions of a lot of smart people who study this field pretty well. This week what I want to do is run through what these elements are, the scores that people gave them, and then some takeaways, and I even have an exercise for you all at home or at the office as the case may be.
So interestingly, the largest portion that was given credit by the SEOs who answered this question was domain-level link authority. This is sort of the classic thing we think of in the Moz scoring system as domain authority, DA. They said 20.94%, which is fairly substantive. It was the largest one.
Just underneath that, page-level link features, meaning external links, how many, how high-quality, where are they coming from, those kinds of things for ranking a specific page.
Then they went to page-level keyword and content features. This isn’t just raw keyword usage, keyword in the title tag, how many times you repeat on the page; this is also content features like if they think Google is using topic modeling algorithms or semantic analysis models, those types of things. That would also fit into here. That was given about 15%, 14.94%.
At 9.8%, then they all kind of get pretty small. Everything between here and here is between 5% and 10%. A bunch of features in there, like page-level keyword agnostic features. So this might be like how much content is in there, to what degree Google might be analyzing the quality of the content, are there images on the page, stuff like this. “How fast does the page load” could go in there.
Domain level brand features. Does this domain or the brand name associated with the website get mentioned a lot on the Internet? Does the domain itself get, for example, mentioned around the Web, lots of people writing about it and saying, “Moz.com, blah, blah, blah.”
User usage and traffic or query data. This one’s particularly fascinating, got an 8.06%, which is smaller but still sizeable. The interesting thing about this is I think this is something that’s been on the rise. In years past, it had always been under 5%. So it’s growing. This is things like: Are there lots of people visiting your website? Are people searching for your domain name, for your pages, for your brand name? How are people using the site? Do you have a high bounce rate or a lot of engagement on the site? All that kind of stuff.
Social metrics, Twitter, Facebook, Google+, etc., domain-level keyword usage, meaning things like if I’m trying to rank for blue shoes, do I have blue shoes in the domain name, like blueshoes.com or blue-shoes.com. This is one that’s been declining.
Then domain-level keyword agnostic features. This would be things like:
What’s the length of the domain name registration, or how long is the domain name? What’s the domain name extension? Other features like that, that aren’t related to the keywords, but are related to the domain.So, from this picture I think there’s really some interesting takeaways, and I wanted to walk through a few of those that I’ve seen. Hopefully, it’s actually helpful to understand the thematic clusters themselves.
Number one: What we’re seeing year after year after year is complexity increasing. This picture has never gotten simpler any two years in a row that we’ve done this study. It’s never that one factor, you know, used to be smaller and now it’s kind of dominant and it’s just one thing. Years ago, I bet if we were to run this survey in 2001, it’d be like page rank, Pac-Man, everything else, little tiny chunk of Pac-Man’s mouth.
Number two: Links are still a big deal. Look here, right? I mean what we’re essentially seeing in this portion here is domain-level link authority and page-level link features, all of them. You could sort of think of this as maybe page authority being a proxy for this and domain authority being a proxy for this. That’s still a good 40% of how SEOs are perceiving Google’s algorithm. So links being a big important portion, but not the overwhelming portion.
It has almost always been the case in years past that the link features, when combined, were 50%. So we’re seeing that they’re a big deal both in the page and domain level, just not as big or as overwhelming as they used to be, and I think this is reflected in people’s attitudes towards link acquisition, which is, “Hey, that’s still a really important practice. That’s still something I’m looking forward to and trying to accomplish.”
Number three: Brand-related and brand-driven metrics are on the rise. Take a look. Domain level brand features and user usage or traffic query data, this is comprising a percentage that actually in sum exceeds page-level keyword content and features. This is really kind of the branding world happening right here. So if you’re not building a brand on the Web, that could be seriously hurting your SEO, maybe to the same degree that not doing on-page optimization is. Actually, that would be a conclusion that I personally would agree with as well.
Number four: Social is still perceived to have a minor impact despite some metrics to the contrary. So, social you can see up here at 7.24%, which is reasonably small. It’s the third-smallest factor that was on there. And yet, when we look at how do social metrics correlate with things that rank highly versus things that rank poorly, we’re seeing very high numbers, numbers that in many cases exceed or equal the link metrics that we look at. So here at Moz we kind of look at those and we go, “Well, obviously correlation does not imply causation.” It could be the case that there are other things Google’s measuring that just happen to perform well and happen to correlate quite nicely with social metrics, like +1s and shares and tweets and those kinds of things.
But certainly it’s surprising to us to see such a high correlation and such a low perception. My guess is, if I had to take a guess, what I’d say is that SEOs have a very hard time connecting these directly. Essentially, you go and you see a page that’s ranking number nine, and you think, “Hey, let me try to get a bunch of tweets and shares and +1s, and I’m going to acquire those in some fashion. Still ranking number nine. I don’t think social does all that much.” Versus, you go out and get links, and you can see the page kind of rising in the search results. You get good links from good places, from authoritative sites and many of them. Boom, boom, boom, boom. “I look like I’m rising; links are it.”
I think what might be being missed there is that the content of the page, the quality of the page and the quality of the domain and the brand and the amplification that it can achieve from social is an integral part. I don’t know exactly how Google’s measuring that, and I’m not going to speculate on what they are or aren’t doing. The only thing they’ve told us specifically is that we are not exclusively using just +1s precisely to increase rankings unless it’s personalized results, in which case maybe we are. To me, that kind of hyper specificity says there’s a bigger secret story hiding behind the more complex things that they are not saying they aren’t doing.
Number five, the last one: Keyword-based domain names, which I know have been kind of a darling of the SEO world (or historically a darling of the SEO world) and particularly of the affiliate marketing worlds for a long time, continue to shrink. You can see that in the correlation data. You can see it in the performance data. You can see it in the MozCast data set, which monitors sort of what appears in Google and doesn’t.
Our experience reinforces that. So remember Moz switched from the domain name SEOmoz, which had the keyword SEO right in there, to the Moz domain name not very long ago, and we did see kind of a rankings dive for a little while. Now almost all of those numbers are right back up where they were. So I think that’s (a) a successful domain shift, and I give huge credit to folks like Ruth Burr and Cyrus Shepard who worked so hard and so long on making that happen, Casey Henry too. But I think there’s also a story to be told there that having SEO in the domain name might not have been the source of as many rankings for SEO-related terms as we may have perceived it to be. I think that’s fascinating as well.
My recommendation, my suggestion to all of you, if you get the chance, try this. Go grab your SEO team or your SEO colleagues, buddies, friends in the field. Sit down in a room with a whiteboard or with some pen and paper. Don’t take a laptop in. Don’t use your phones. List out these features and go do this yourself. Go try making these percentages for what you think the algorithm actually looks like, what your team thinks the algorithm looks like, and then compare. What is it that’s the difference between kind of the aggregate of these numbers and the perception that you have personally or you have as a team?
I think that can be a wonderful exercise. It can really open up a great dialogue about why these things are happening. I think it’s some fun homework if you get a chance over the next week.
Until then, see you next week. Take care.
Video transcription by Speechpad.com
Sign up for The Moz Top 10, a semimonthly mailer updating you on the top ten hottest pieces of SEO news, tips, and rad links uncovered by the Moz team. Think of it as your exclusive digest of stuff you don’t have time to hunt down but want to read!
Continue reading →