Ranking Factor Studies In The Era Of Machine Learning

Just admit it, SEO is scary. Between the inherent complexity of what we do and Google not exactly being the epitome of clarity, the ground that is doing SEO can be a bit shaky at times. That’s pretty much why we’re obsessed with what works and what doesn’t work and are vigilantly on the lookout for content that offers a bit of light at the end of the tunnel.

In the not too distant past, I wrote a piece highlighting how machine learning has impacted rank volatility (in that rank is considerably more volatile). At the time, we touched on what machine learning means for understanding how ranking works and how the process directly influences rank. Here, we’ll get into the nitty-gritty of it all by analyzing the holy of holies of optimization information, ranking factor studies, particularly niche ranking studies by asking one very simple question
…. 

Do ranking factors studies still apply in a world where machine learning and intent reign supreme, and if so, to what extent? 

Recap of Machine Learning’s Impact on Rank

The increase in rank volatility aside, in what for all intents and purposes was “Part I” of this post we discussed how machine learning impacts rank qualitatively, i.e., what rank “looks like” as a result of RankBrain and the like. Since I’m a nice guy, let me recap (and expand on) what we said there so that you don’t have to comb through the last piece trying to glue together all of the pieces to the puzzle. 

Machine Learning Sets Site Proportions

In serving up results that align to user intent, Google uses machine learning to determine the proportion of sites to meet that intent or those intents. OK, Mordy, say that in English, please?! 

If you’ll remember, in the last post I took a very straightforward search term, buy car insurance, and showed that Google sees two (or really more than two) intents embedded in that phrase: to buy an actual insurance policy and to get information about doing just that. 

How should Google handle these two intents? Should 50% of the results on the page be geared to sites offering car insurance? Is buying a car insurance policy the essential intent and learning a sub-intent? If so, should 75% of the results be sites that offer you the opportunity to buy car insurance? You get my point, there has to be some way to determine how many sites should be allocated to each intent on the SERP shown below. 

When machine learning does this, it either limits or enhances ranking opportunities, depending on how you look at it or which kind of site you happen to be. In our case, 60% of page one (or 6 out of 10 results) was allocated towards sites where you could buy an actual insurance policy. Meaning, sites like Geico and Allstate were limited to just over a majority of the only page of the SERP that actually matters. On the other hand, sites like NerdWallet were given the chance to rank on 40% of this page that without Google’s ability to understand intent would not be possible (all things being equal).

==> Check out our interview with Lazarina Stoy about getting started with machine learning for SEO 

Machine Learning Determining Actual Rank Positions

 

That can’t be… can it? “Search your feelings, you know it to be true.” (You can scream now if you want, no one will judge you, this is a safe place.) 

Look back up at the SERP above. See Geico there in the top spot? Think that was an accident? I don’t mean that Geico ranked over StateFarm or whatever other insurance company, I mean that an insurance company ranked over a site that only helps you go about the buying process (such as found in the second result). 

That second result, How to Buy Car Insurance, it can never rank in the top spot (as things currently stand). Not because it lacks in optimization or whatnot, but because it doesn’t align to the essential intent here, to buy a car insurance policy. 

Boom! Machine learning can determine rank down to the position (in certain instances)!

Machine Learning Precedes Ranking Factors

So how about those ranking factors, huh?! They really impacted rank at this stage right? Of course, I’m being a bit facetious. We never discussed
ranking factors above, which is my point. It is possible to discuss rank, to see rank unfold in a way, without initially discussing ranking factors. Meaning, to a large extent, or more especially to a “primary” extent, rank is not determined by any given signal. In today’s world of intent, machine learning is the primary driver of rank in that it creates the very context out of which rank unfolds (as illustrated above). 

That is, there are two questions that machine learning’s “ranking presence” forces us to deal with: 

1. How does machine learning impact the optimization process per se?
2. How does machine learning impact the point of optimization departure, or, how does it change the entire way we view what determines rank?

We’ll get to both of these below, but did I really need to say that? 

Is RankBrain Actually a Ranking Signal? 

 
The underlying premise to what I’m saying is that RankBrain is not the third most significant ranking signal, because it’s not a ranking signal at all. Let’s take a step back for a second and ask what is a ranking factor fundamentally? Well, it’s a signal that tells Google that a site may very well achieve what it set out to do. RankBrain is a major factor in determining what it is that both users and sites have set out to accomplish. As such, this machine learning process has a major role in determining what factors, or what signals are relevant to a given site. This is why you hear people talk about RankBrain in the context of determining how to apply Google’s various algorithms (and why we need to answer question #1 in the section above).

How then can the thing that is determining the signals, vis a vis their relevance, be one of those signals? It’s logically impossible. The ‘thing’ telling the elements of the algorithm what to do cannot be one of those elements at the same time, much the way a light switch cannot be both on and off simultaneously. RankBrain is really an extension of Hummingbird, which is the algorithm. Think of RankBrain as a way of optimizing Hummingbird, or making it more efficient and effective. RankBrain, as such, relates to the underpinnings of the algorithm per se, it exists beyond the signals/factors, it’s a meta-element. 

Yes, I am well aware that Google has said that RankBrain is the third most important signal. I’m also well aware that Google recanted, and said no top signals even exist. Instead, Gary Illyes said, “… it (the signals) very much depends on the query and the results which signals count more.” Which of course sound suspiciously similar to what someone else whose name rhymes with Gordy Loberstein just said two paragraphs ago. Of course, Gary’s statement got a lot less attention than when Google indicated the top few ranking signals. Which just goes to show that we crave definition and information and loath obscurity. However, our want and desires don’t dictate what Google does and does not do. 

What Does Machine Learning Mean for Ranking Factor Studies? 

 

There’s actually a pretty big implication to the above in that each “intent” deserves its own weighing of the factors that determine rank. In other words, since a ranking factor is a signal that a site is likely to achieve what it seeks to do (and therefore satisfy its users), and since different sites are intent on achieving different things, the extent to which a given ranking factor applies hinges on the very purpose of the site.  

 

With that, let’s ask the two questions from above but from the context of ranking factor studies: 

1) Where Google is able to parse intent as well is it can, and where each intent demands its own weighing of the applicable ranking factors, just how relevant are general and even niche ranking factor studies?  

2) How relevant are ranking factor studies when they don’t deal with the true underpinnings of rank, i.e., machine learning’s ranking impact? 

A Look at Intent and Ranking Factor Study Relevancy 

Let me be clear at the onset. I am not advocating that we throw some really good ranking factors studies that are out there in the trash. This is not some sort of sensational post aimed at rocking the boat simply for the sake of causing waves. Rather, this is a comprehensive look at how the underpinnings of rank work in regards to the various levels of intent that intrinsically exist within any keyword or piece of content, so as to determine the role of the various categories of ranking factor studies. So let’s do just that. 

General Intent for General Ranking Factor Studies

Any given query phrase or web page is subject to multiple layers of intent at both the general and specific levels.
Intent is very much a spectrum. For example, when searching for any piece of content, a user almost always intends to be served content that is, among other things: 

As such, you can imagine that “factors” such as a keyword in the title, or links, or being HTTPS are generally relevant to almost any site to varying extents. When you see a “general” ranking factor study indicate as much, the sentiments/data are all things considered, entirely accurate. And, if you want to “get ranked” you need to make sure your pages more or less comply with these sentiments. However, that is not to say that doing so will result in ranking well on the SERP. Rather, it means that doing so may allow you to enter the
ranking fray.

What’s the difference? 

As opposed to being that which determined your rank, ensuring your page meets the most basic of standards is what allows you to even be considered for top-level rankings.

For illustrative purposes let us say there is a raffle to play first base for the New York Yankees (or whatever sports team you like, or if you hate sports, imagine a raffle to star in your favorite movie… everyone happy now?). You and 1,000,000 fanatics line up outside the stadium for the raffle and sure enough, your name is called. Super excited, you run to claim your prize, except you’re a bit disappointed since this stage of the raffle only allows you entrance into the stadium to participate in the next round of bidding.

Optimizing your page so that the content is of good quality, or that it’s safe for users to navigate only gets you into the stadium, it does not get you onto the field.   

Niche Ranking Factor Studies for More Specific Levels of Intent 

Moving right along down the funnel of intent, and we get to the niche level. Niche ranking factor studies rely on, and are applicable
only, because there is a commonality between the sites that exist within any given niche. Generally speaking, a commonality of intent exists between the sites within a specific niche. The easiest case I can think of are sites related to finance. Here, and again, generally speaking, users want, or in our lingo, users intend that the financial sites they interact with be safe and secure. More so than say a site where they look up a recipe for tonight’s dinner. For the “finance niche” an added degree of site security is an implied user intent and as such, something like HTTPS might be a more influential factor than perhaps a keyword in the title (I’m just supposing, not proclaiming). 

Now, does this apply to every site within the finance niche equally? No. Do you care that investopedia.com be as secure as the site where you do your online banking? And even if you are an online banking site, does being HTTPS mean you’ll hit the top of the SERP, even if your competitors are not?

Not sure? Let me ask it this way, is the essential intent of an online banking user to have a safe experience, or is to manage their money?  

 

Think about it like this, general ranking factor optimization gets you into the stadium, but only so you can sit in the cheap seats. Optimizing for niche specific factors (all things being equal) take you from the bleachers and puts you in a better position to see the field… but you’re still not on the field. 

Catering to Query Specific Intent 

While there are multiple layers of intent, it is the essential intent that differentiates one page, or one piece of content, or even one search term from the next. In other words, what factors are the most influential? The factors that most closely relate to the essential intent of the site/page. If a ranking signal is an indicator that a site has a good chance of achieving what it seeks to do, or what users want from such a page, then the most important signals are those that relate to what gives the page its identity (not to get too philosophical here, that’s for later). 

Let me give you an example from the health niche. Consider the query do I have cancer, which is what one might consider
to be a pretty standard or traditional query related to the health niche (and one that you will hopefully never have to search for). In such a case, and I would imagine for the niche in general, having well-organized content so that users can digest the information would be vital (think longer content, H1s
and H2s, etc.). 

Let’s, however, consider another query within the same niche, desserts for diabetics. Would you believe that for such a query the top recipe sites do not
rank. Rather, health sites like diabeticlivingonline.com occupy page one of the SERP. That is, despite relating to food, this query exists within the health niche. And do you think that content organization is as important here, for a what is all but a recipe, at least to the same extent as someone looking to see if their symptoms meet the criteria for possibly having cancer? Or the inverse, do you think a page relating to informing users as to the symptoms of having cancer needs an image to the same extent as a recipe site? Here, despite the recipe site being in the health niche, having that image of what the complete concoction looks like is essential to keeping users on the page (much like a standard recipe site). That is, even if this diabetic recipe site needed to have the same level of content organization as a site offering the symptoms of cancer (which of course it does not), would that be the difference maker? Is that the most dominant signal needed for a what is all but a health-oriented recipe site?

Here are two sites, both in the same niche, with each site having two very different factors that influence ranking. In this case, to what extent is a niche ranking factor study helpful? 

Take the travel niche for example. We’re talking about hotels, resorts, etc. Images are going to be a dominant factor here, I don’t even need to do a study to know that. Consider though a site reviewing hotels. Images are still going to be highly relevant, but so will offering a longer sort of content experience. Whereas a tourism site may get away with a quick snippet of content for the top attractions a city has to offer, would users want the same from a site where they expect a detailed review of a hotel? 

So yes, in the travel niche there is a general expectation to see images, and as such, it may be (or may not be, I have not studied it in-depth) the prevalent factor across the niche, but not at the site level. 

If a niche ranking factor study gets you close to the field, understanding the factors that are more essential to your page or to a query is what gives you actual playing time.  

Summing Up Where We Stand – General and Niche Ranking Factor Studies

I know this can be a lot to digest, and that’s because the underlying concept is one that has filled up the pages of philosophy masterpieces for the past thousand or so years. This is like some sort of crazy cosmic merger of the world of SEO and the world of philosophy. So let me take a play from my old teacher’s playbook and summarize what I said above in slightly different terms. 

Here are a few sentences that if you were to take them out of context would show me to be a simpleton: 

Humans, as a general rule, have two eyes. We usually have two arms, two legs, two ears. We have toes and fingers. We have hair (for some of us less each day, which makes looking down at the shower drain a dreadful nightmare of an experience). 

Do any of these things give us our identity? Not really. Sure, in a general sense the human form is part of who we are relative to squirrels, but it’s not who we are. 

Ok, how about this: 

Humans, as a general rule, are compassionate (I know, fool, he still believes in humanity), we all have dreams and ambitions, likes and dislikes, and so forth. Are these things “us,” do our desires and ambitions define us? Well, they say a lot about us, but they sure don’t make me… me (if my mood defined me, I’d be in big trouble). 

What in the heck am I doing here… this is SEO, not Aristotelian logic 101?! 

Actually, funny you should mention the Greek freak, because all of these things, our bodily structure, our emotional makeup… are in Aristotlean terms called secondary substances (don’t ask me why I know this). Meaning, they are a part of us, but a secondary part, not the essential component of our identity (I’m taking a bit of leeway with the concept but who cares, this is SEO, not logic class).  

The common intent found across all pages (see above) is like the commonality we all have physically. Sure, it’s a part of what makes us human and so forth, but no one would identify as their body per se (relative to something deeper, such as the personality, and so forth). The ranking commonality, or the ranking correlations, shown at the top level of intent, while legitimate, don’t get anywhere near the core of intent, and as a result are extremely hard to apply at the page or even the site level. 

The same within the common intent that exists at the niche level. Yes, this layer of intent is deeper than at the general level. Most users of pages within the finance niche are more sensitive to security issues, but does that “intent” define the user’s aims and aspirations? Niche level intent, though deeper, still misses the
intent mark. Much like our emotional selves, though deeper than our physical selves, do not define us, niche level intents do not touch on essential intent. Thus, basing your optimization strategy on niche ranking factor studies, though not a bad place to start, cannot offer insights on the weighing of the various factors at the site or page level…. which is the realm in which you exist.   

Much like people, each individual page has its own unique qualities that differentiate it from all other pages (in a perfect world where people don’t publish the same dribble over and over and over…). Each site or page is intent on achieving a very specific goal, one that is far more refined than any niche level “purpose.” Jump back to the health industry… could you say that every site is aimed at helping people be more healthy? Sure, that’s not a terrible way to put it. Now apply that to the page level. Take the page offering my diabetic uncle some tasty treats, should we say its essential aim is to make people healthier? That sounds a bit too general, no? Then why would you engage in an optimization strategy that is based on just that?! 

Remember, what is a ranking factor? It’s a signal that tells Google that a site is likely to achieve its aim of satisfying user intent. Which intent? The general or niche level intent, or the user’s actual intent? Just like with people, the essential intent is the most specific, the one that differentiates the thing in question from all other things. For us as people, it’s that ineffable quality that makes me different than you, and just about everyone different that grandma Edna (I don’t have a grandma Edna per se, but we all have a figurative one in our families). For a page or for a site, or even for a user (approaching this from another vantage point), the essential intent is that which makes the site, or the page, or the search term different from all others (or most others). As such, why would Google measure success and weigh its signals on anything but that highly specific essential intent (especially now that it can do as such)? The answer is they don’t. So then why are we as an industry still bank on studies that are intrinsically far too broad for the world of intent as ruled by an always improving machine learning system? 

Query Level Ranking Factor Understanding 

Great Mordy, you threw out all of that data I’ve been relying on for God knows how long… thanks a lot!… No problem, you’re welcome. 

Let me be clear (again), I am not saying that you should trash ranking factor studies. For the most part, whether general or niche, they are well written, well-researched pieces of SEO knowledge written by some folks that I have a tremendous amount of respect for. 

You should read these studies, you should learn from these studies… but you should also go far beyond them. 

It’s important to know what gets you into the ballpark these days, it’s important to know what ranking trends your niche shows, but if you want to know what will be the difference maker for your site, for your pages, for the keywords you are targeting… go query specific, I beg you. 

Thanks Mordy, but how do I do that???

Optimizing for Intent, Not Niche 

Like I said, don’t throw out niche ranking factor studies, they’re a great place to start. Take a look at the studies and see what applies to the niche, and then ask a very simple question… do the most influential factors make sense for my site? 

If a study says that longer content is the most influential factor in the health niche (I’m just making up an example) and you have a “health” recipe site, does that really make sense for you? Do users really want to read through a 1,000-word recipe? Wouldn’t you be better served with some bullet points and well-placed images of the succulent delights you are offering? 

I know “Google” and “ranking” and “signals” are complicated topics, but it’s not rocket science either. You literally have to think about what makes sense and what does not make sense for your site/pages based on what the user would expect, because if the user expects a given experience, you can bet that Google does too. 

Of course, this is not an exact science, and experimenting with this formula is a very, very, very good idea. If you think optimizing for “A” is really important, track your progress, experiment with optimizing for “B” instead and see where it goes. If you’re totally clueless, have a look at the niche study and start with the most influential factor and take it from there. But experiment.  

By the way, this may mean looking at the more influential ranking factors from other niches. Running with this case of a site offering diabetic dessert recipes… I would have a look at what works for the food and drink industry first, not the health niche. 

Whatever it is, think query and page specific, and don’t get trapped in data. Meaning, don’t get so obsessed with incorporating data that you don’t allow yourself to think critically about what makes sense for your site, not your niche!  

Forget About Ranking Factors (Initially), Focus on Intent Analysis First  

There is a whole other way to go with all of this. Yes, optimizing based on how Google views you through the lens of intent makes far more sense than doing so based on the nature of the niche you exist within. Still, like the 1960’s, I’d like to expand your mind, by telling you to forget about ranking factors, at least initially. Rather, try to make sure you first are able to fill all of the intent slots on the SERP. 

In other words, do you really want to spend all of your time optimizing for this and improving the page for that, and whatever… or do you want to make sure you aren’t limiting yourself to just 40% of page one? What’s the point of optimizing something (based on page specific signal analysis of course), if that page only has access to a limited amount of the SERP? Perhaps, adding on to that page, or even writing all new pages that
tackles the keyword from a whole new angle would give you a better chance of ranking?  

In the era of machine learning,
ranking is not fundamentally dependent on incorporating ranking signals, but in aligning to user intent. 

How to Uncover Multiple Layers of User Intent 

This is where it all comes together. You might be asking, “Why do I need to know all of this theory? Can’t I just get insights?”. No, because if you don’t understand the concepts, the tips will be meaningless. I know, I know, everyone wants quick “actionable” tips. The problem is, if you don’t understand something, you both cannot run with it, and cannot act on it since you don’t really know how important it is. This is all the more applicable in this instance. What I am about to show you
has so many variations that depend on the nature of your site, of the keywords you are targeting, etc. The idea here is not that I will offer you the answers, but the foundation of how to go about query specific analysis so that you can take it and run with it! 

So here we go:

Method #1 – The Marketing Mind

 

The entire idea here is to understand the multiple intents embedded within a keyword term. One really easy way to do this is to apply a marketing approach to your keyword targeting. Get into a user’s head the same way you would when running a marketing campaign. Only in this instance, don’t ask why a user would buy a certain product, but why they would search for a certain term or set of terms.  

In other words, what are the multiple reasons why a user would run a certain query? What are the implications of that? What subtopics or ancillary information might a user want or need depending on the reason behind the search? 

Take the query content marketing pro. I can think of four intent tracks off the top of my head: 

1. To find a content marketing professional for hire
2. To learn how to become a content marketing pro
3. To find a list of content marketing pros (as in to follow on social media or something of the sort)
4. To learn what a content marketing pro is

If you want to rank for this term, and don’t want to limit yourself to one “bucket” of intent, and therefore only one portion of the SERP, considering these intents could be a powerful tool. Again, this is just a simple example that I literally conjured up in less than 3 minutes. 

While this method is not as profound as what I’m about to offer you below, it’s a good way to break into the habit of centering your content strategy around meeting user intent at multiple levels. Baby steps.  

Method #2 – Digging for Intent on the Google SERP 

This is the “money” method (at least until SEO tools start to break into parsing intent from the eyes of Google). 

I could describe what I call ‘digging for intent on the SERP’ or I could just show you what it is:

Earlier, we discussed the multiple intents of a very simple keyword, buy car insurance. I like this keyword (which is why I’ve been sticking with it for both
Part I and Part II of this mini-series). On this page, when I did my search 4/10 sites were informative in that I could not buy an actual insurance policy. In other words, 40% of that page was dedicated to learning about car insurance, not buying it. Is this a fluke, or is this equation pervasive for the term ‘buy’?

To answer this, I ran another 150 “buy” keywords. I took simple things like buy socks and keywords for more complex items like buying a house, a car, or even a boat and plugged them into Google. Out of the 1,500 or so results that I got back (150 keywords x 10 results per page on average = 1,500), 26% were informative sites. Meaning, roughly 1/4 of the results I got back from these “buy” searches. 

That was a bit lower of a proportion than what I saw for buy car insurance, so I took out some of the less complex items from the dataset and sure enough, the number started to rise. Seeing this, I decided to hone in on one item, one “complex” item… software. I ran 100 ‘buy’ keywords for all sorts of software, eLearning software, tax software, asset management software, and so on. Guess what percent of these 1,000 or so results were ‘informative’ sites… 40%. 

Great, so there is a very real connection between the term ‘buy’ and learning. Meaning, when you type ‘buy’ Google understands it as ‘learn’, 25% of the time for less complex items (on average) and 40% of the time for more complex items (again, on average). 

However, that’s not where this ‘intent’ story ends. Within my ‘buy software’ dataset, 70% of the informative sites, or 70% of the 400 or so sites that did not allow me to buy software, but to learn about buying software, were what I call “best of” sites. You know these sites… they’re the kind with titles
like: Top 5 Photo Editing Software Solutions You Need Now. What of the other 30%? They were sites that taught me how to choose between the various ‘best’ software products, what considerations you should weigh before buying a given software solution, and so forth. 

Let’s count up our intents here: 

1. To buy software

2. To learn about buying software

2a) To learn which product to buy

2b) To learn how to choose between the various products, what to consider before buying 

What You Get for Your Intent Analysis 

OK, so what did we do here? We took a keyword that any given number of sites may be vying for (buy car insurance) and found an intent patten (buying sites and informative sites appear together). We then determined that this pattern was pervasive, i.e., this is how Google generally understands the word ‘buy’ in
search. We also qualified this by determining the proportion of this understanding depending on complexity. Finally, and perhaps most importantly, we uncovered multiple layers of intent by analyzing the types of sites that appeared on the SERP. 

Great, let’s apply this to a real-world scenario. Say you have a client who has an e-commerce site where they sell a whole bunch of products, complex products, who is targeting a whole host of ‘buy’ keywords. You would probably recommend they start a blog so that they don’t preclude themselves from ranking on 40% of the SERP. You might even suggest they write about both which products to buy, and how to go about choosing between all of the wonderful products available to them. 

If you know what intents Google sees in the keywords you are targeting, you have the ability to design a content strategy that hits on all of those intents, and sub-intents, and sub-sub-intents… you get my point. 

Get In the Game, For Crying Out Loud, Get in the Game! 

Ranking factors are obviously important, and the studies that analyze them are wonderful… but hyper-focusing on them won’t get you out of the stands and onto the field. The fact is, you have to optimize on a per site/per page basis. Signal weights are contingent on the purpose of the page and its content, that’s what a ranking factor is, a method of determining if a page will succeed or fail in what it has sought to do relative to user intent.

General levels of intent and purpose are prerequisites, they are not difference makers. Having a quality page with decent links is not going to make your page highly relevant to a user, it means, for lack of more eloquent language, that your pages ‘don’t suck’ so there’s the chance Google will show them on the SERP when appropriate. Is that what you’re really aiming for? So while the data of a general or niche level ranking factor study can be helpful, they simply don’t deal with the crux of optimization for your site and your pages. 

Forgetting all of that, if you’re not aligned to Google’s lens of intent, your hours of optimization are for naught. Don’t get caught in the trap of worrying about which factors are the weightiest for your pages when you haven’t even jumped into the **** yet. If you don’t consider how your content fits into the various buckets or slots on the SERP, then forget ranking factors, your entire foundation and starting point is on shaky ground. By the way, I don’t mean some sort of quick general analysis, how Google parses up a SERP is very nuanced, to the extreme, so give it a good run.

The point of departure in today’s world of SEO is not “ranking signals.” A deep, nuanced, and ever-changing understanding of how Google sees user intent for your content and targeted keywords is what puts the ball in play. And we all want the ball in play.