Today I'm going to make a crazy claim—that in modern SEO, there are times, situations, and types of analyses where correlation is actually MORE interesting and useful than causality. I know that sounds insane, but stick with me until the end and at least give the argument a chance. And for those of you who like visuals, our friend AJ Ghergich and his intrepid team of designers created some nifty graphics to accompany the piece.
Once upon a time, SEO professionals had a reasonable sense of many (or perhaps even most) of the inputs into the search engine's ranking systems. We leveraged our knowledge of how Google interpreted various modifications to keywords, links, content, and technical aspects to hammer on the signals that produced results.
But today, there can be little argument—Google's ranking algorithm has become so incredibly complex, nuanced, powerful, and full-featured, that modern SEOs have all but given up on hammering away at individual signals. Instead, we're becoming more complete marketers, with greater influence on all of the elements of our organizations' online presence.
Web marketers operate in a world where Google:
- Uses machine learning to identify editorial endorsements vs. spam (e.g. Penguin)
- Measures and rewards engagement (e.g. pogo-sticking)
- Rewards signals that correlate with brands (and attempts to remove/punish non-brand entities)
- Applies thousands of immensely powerful and surprisingly accurate ways to analyze content (e.g. Hummingbird)
- Punishes sites that produce mediocre content (intentionally or accidentally) even if the site has good content, too (e.g. Panda)
- Rapidly recognizes and accounts for patterns of queries and clicks as rank boosting signals (e.g. this recent test)
- Makes 600+ algorithmic updates each year, the vast majority of which are neither announced nor known by the marketing/SEO community
Given this frenetic ecosystem, the best path forward isn't to exclusively build to the signals that are recognized and accepted as having a direct impact on rankings (keyword-matching, links, etc). Those who've previously pursued such a strategy have mostly failed to deliver on long-term results. Many have found their sites in serious trouble due to penalization, more future-focused competitors, and/or a devaluing of their tactics.
Instead, successful marketers have been engaging in the tactics that Google's own algorithms are chasing—popularity, relevance, trust, and a great overall experience for visitors. Very frequently, that means looking at correlation rather than causation.
[Via Moz's 2013 Ranking Factors - the new 2015 version is coming this summer!]
We'll engage in a thought experiment to help highlight the issue:
Let's say you discover, as a signal of quality, Google directly measures the time a given searcher spends on a page visited from the SERPs. Sites with pages searchers spend more time on get a rankings boost, while those with quick abandonment find their pages falling in the rankings. You decide to press your advantage with this knowledge by using some clever hacks to keep visitors on your page longer and to make clicking the back button more difficult. Sure, it may suck for some visitors, but those are the ones you would have lost anyway (and they would have hurt your rankings!), so you figure they're not worth worrying about. You've identified a metric that directly impacts Google's algorithm, and you're going to make the most of it.
Meanwhile, your competitor (who has no idea about the algorithmic impact of this factor) has been working on a new design that makes their website content easier, faster, and more pleasurable to consume. When the new design launches, they initially see a fall in rankings, and don't understand why. But you're pretty sure you know what's happened. Google's use of the time-on-site metric is hurting them because visitors are now getting the information they want from your competitor's new design faster than before, and thus, they're leaving more quickly, hurting the site's rankings. You cackle with delight as your fortune swells.
But what happens long term? Google's quality testers see diminished happiness among searchers. They rework their algorithms to reward sites that successfully deliver great experiences more quickly. At the same time, competitors gain more links, amplification, social sharing, and word of mouth because real users are deriving more positive experiences from their site than yours. You found an algorithmic loophole and exploited it briefly, but by playing the "where's Google weak?" game rather than the "where's Google going?" game, you've ultimately lost.
Over the last decade, in case after case of marketers optimizing for the causal elements of Google's algorithm, this pattern of short-term gain leading to long-term loss continually occurs. That's why, today, I suggest marketers think about what correlates with rankings as much as what actually causes them.
If many high-ranking sites in your field are offering mobile apps for Android and iOS, you may be tempted to think there's no point to considering an app-strategy just for SEO because, obviously, having an app doesn't make Google rank your site any higher. But what if those mobile apps are leading to more press coverage for those competitors, and more links to their site, and more direct visits to their webpages from those apps, and more search queries that include their brand names, and a hundred other things that Google maybe IS counting directly in their algorithm?
And, if many high ranking sites in your field engage in TV ads, you may be tempted to think that it's useless to investigate TV as a channel because there's no way Google would reward advertising as a signal for SEO. But what if those TV ads drive searches and clicks, which could lead directly to rankings? What if those TV ads create brand-biasing behaviors through psychological nudges that lead to greater recognition and a higher likelihood of searchers click on, link to, share, talk about, write about, buy from, etc. your TV-advertising competitor?
Thousands of hard-to-identify, individual signals, mashed together through machine learning, are most likely directly responsible for your competitor's website outranking yours on a particular search query. But even if you had a list of the potential inputs and the mathematical formulas Google's process considers most valuable for that query's ranking evaluation, you'd be little closer to competently beating them. You may feel smugly satisfied that your own SEO knowledge exceeded that of your competitor, or of their SEO consultants, but smug satisfaction does not raise rankings. In fact, I think some of the SEO field's historic obsession with knowing precisely how Google works and which signals matter is, at times, costing us a broader, deeper understanding of big-picture marketing*.
Time and again, I've seen SEO professionals whom I admire, respect, and find to be brilliant analysts of Google's algorithms lose out to less-hyper-SEO-aware marketers who combine that big picture knowledge with more-basic/fundamental SEO tactics. While I certainly wouldn't advise anyone to learn less about their field nor give up their investigation of Google's inner workings, I am and will continue to strongly advise marketers of all specialties to think about all the elements that might have a second-order or purely correlated effect on Google's rankings, rather than just concentrate on what we know to be directly causal.
-----------------
* No one's guiltier than I am of obsessing over discovering and sharing Google's operations. And I'll probably keep being that way because that's how obsession works. But, I'm trying to recognize that this obsession isn't necessarily connected to being the most successful marketer or SEO I can be.
Just a few additions:
1) I read this piece - SEO Has Evolved - It's About Popularization Now - by Tad Chef yesterday and I think it closely relates to some of my logic here. Knowing what competitors did to become popular and thus earn their ranking signals might be more useful than simply knowing what signals are needed to rank.
2) I also read, over the weekend, this piece on Google's evolution from machine learning to deep learning, which I think points to the trend of even the search quality engineers in building 43 no longer knowing what elements make up the ranking algorithm (only the machines know now!). When that's the case, optimizing for a fixed set of inputs doesn't make nearly as much as optimizing toward the elements that correlate with what the deep learning machines seems to be biasing toward in your niche/vertical/set of SERPs.
I figure this piece is going to generate some controversy, as many in our line of work have been so opposed to correlation data and research, but in this context, I'm talking less about whether Page Authority has a 0.38 correlation with higher rankings and more about what the domains that consistently rank on page 1 in your vertical are doing that you (and others who aren't consistently on page 1) are missing.
Thanks Rand. That's a useful piece you've written up there. Initially I thought that where is this man going with it but after reading the full post it totally makes sense.
The second article that you shared reminds me of this piece by Aeon magazine. I think you will very much like it.
https://aeon.co/magazine/technology/is-technology-m...
Oh, and while I'm at it, I've noticed that you included the Fermi paradox in one of your presentations.. So I wondered- have you ever read anything by Ray Kurzweil? Singularity, stuff like that? Cause once you do, you see Google's algorithm through different eyes. Just saying :)
*editing this for the second time... Btw, the guy now works at Google so...
"Sites with pages searchers spend more time on get a rankings boost, while those with quick abandonment find their pages falling in the rankings."
Hi Rand,
It seems as though you're making a few assumptions here that perhaps you shouldn't be, and overlooking the ease at which Google could use to join together different signals that could easily reinforce each other.
Regardless of how long or short a time someone spends on a page that they may have selected from search results, it's pretty much easy for Google to look at the query sessions associated with those selections, and determine whether or not those searchers continued to search for more information on the same topic, even though they may have spent a long time on the page that they looked at before, or if they appear to have fulfilled their situational or informational needs even though they spent very little time on the page they may have just looked at.
In that second scenario, which could possibly be mistaken for a quick abandonment of of a page could be instead shown as a very positive success if a lot of people who visit that particular page change the focus of their query sessions to something else immediately after visiting it.
A Google patent that I looked at and wrote about last year (Propagating query classifications) describes Google possibly collecting information about searchers' selections of page for a query that might have more than one meaning, and collecting that information in a triplestore of data about those queries as search entities. The information about the query use, the page selection, and the time spent on a page could potentially be used to determine classifications for the queries based upon different meanings. It's interesting in that it's one of the few patents I've seen from Google
The information collected could be expanded in other ways, such as based upon where the query was performed (geolocation can determine whether something like soda might be queried as "pop" or "coke", for example). It could also easily indicate the next step of a searcher after visiting that page, such as "starting a completely new search," or "continuing the query session searching for similar information."
When Google intelligently combines signals like those, and finds a way to track them that might play off this initial approach and go beyond correlation to a more reasoned causation approach, I actually like seeing it happen (from the patent):
Each record (herein referred to as a tuple: <document , query, data>) comprises a query submitted by users, a document reference indicating the document selected by users in response to the query, and an aggregation of click data for all users or a subset of all users that selected the document reference in response to the query. </document>
Before we quit on causation to accept correlations as strong hints of "something" (what ever it might be), I do like the idea of digging deeper to see if there's more that we could potentially consider that Google might be looking at also.
Hi Bill - yes, perhaps it wasn't clear in the post, but my intention was to present those as examples, not concrete facts of how Google is/might be doing something. And yes, I do agree that digging into ways Google may be doing something (through varieties of sources, patent applications included), can be a way to make oneself more knowledgeable about the potential causal factors. However, just as with correlation data, I'd urge us not to assume we know root causes simply because we have a patent application to point to (or a strong correlation between a data point and higher rankings).
Hi Rand,
When I look at a patent application filed with the USPTO, I know for certain that at least
(1) a number of people inside of Google have discussed an issue, and
(2) filed for legal protection of a process, and
(3) their jobs were at risk in terms of getting it right
- so it's not just some casual fanciful act on their part to have filed a patent on a subject..
I'm am assuming that those actions mean something much more than a very random coincidence about the length of a mustache or a mirror breaking or a black cat crossing a street, and how any of those might influence rankings, and I don't think it's a mistake for me to do that.
When I see people viewing the kinds of ranking signals mentioned above as individual signals that couldn't be used together (length of stay on a page in a search result + continuation or change in topic of the immediate query session), I get a sense of dread fear that forces me to ask why? What I said is quite reasonable, and the response "don't foolishly give too much weight to patents" isn't a helpful one.
I've seen many whitepapers from search engines that discuss things like query sessions and many patents as well. It's reasonable that kind of signal could and would be used in that manner.
I included the mention of that patent not as any kind of proof that what I say absolutely is true, but as an example that clearly shows that the concept is an extremely reasonable one, much more than say the influences of different phases of the moon upon rankings.
I
Why SEOs Need to Care About Correlation as Much (or More) than Causation
So, I read that headline and thought, "That's the most-intelligent trolling headline I've ever seen!" But then I read the article. :)
But today, there can be little argument—Google's ranking algorithm has become so incredibly complex, nuanced, powerful, and full-featured, that modern SEOs have all but given up on hammering away at individual signals. Instead, we're becoming more complete marketers, with greater influence on all of the elements of our organizations' online presence.
Rand, I cannot do an embed here, but this essay seems to be a lengthy follow-up to this tweet of yours from some time ago. (That tweet blew my mind so much that it became a personal mantra of mine and served as the inspiration for quite a few Moz blog posts and a Mozinar.)
Given this frenetic ecosystem, the best path forward isn't to exclusively build to the signals that are recognized and accepted as having a direct impact on rankings (keyword-matching, links, etc). Those who've previously pursued such a strategy have mostly failed to deliver on long-term results. Many have found their sites in serious trouble due to penalization, more future-focused competitors, and/or a devaluing of their tactics.
Over the past few years, we've seen (current and former) Googlers say things such as "don't do SEO" (see here for one example). While I -- and probably most everyone here -- think that Google's definition of "SEO" likely does not match ours, I think Rand's second sentiment bolded above is what the Googlers have actually meant. Stop obsessing about the algorithm and rankings and trickery.
You found an algorithmic loophole and exploited it briefly, but by playing the "where's Google weak?" game rather than the "where's Google going?" game, you've ultimately lost.
I have a confession: After several years of thinking about traditional "SEO," I stopped thinking about links and rankings in late 2013. The more I saw what worked for myself and for clients, the more I saw that I had been prioritizing the wrong things.
Here's where the realization hit. A client was hired my prior agency for both "SEO" and "PR" -- but the PR work had started a month before the SEO work began. Do you know what I saw? Long before I had even touched the website -- there were countless errors in the Moz Analytics report -- the PR work bounced them from who-knows-how-far-below-50 to 15 in the SERP for "mobile diabetes app" in one month. (That's the beauty of links that are 100% natural and earned.)
In any marketing budget, I'd spend the majority on PR and less on what is called "traditional SEO" because of where Google is going.
I suggest marketers think about what correlates with rankings as much as what actually causes them.
Here's my personal mission statement moving forward: Build, promote, and publicize a site that delights its target audience. Everything else -- the rankings and more -- will fall into place.
I once explained it like this: Me and my team at a prior agency focused on technical SEO, content, and PR (via social media and other channels). We didn't even think about links and rankings. The goal was to achieve our clients' marketing and business goals -- and the rankings and more were just nice by-products of doing the process well.
Yes, all of the on-page and technical SEO is vital. (I'm preaching to the choir, so I won't go into details.) But everything else is just coming down to doing good marketing and public relations. It's building a brand among human beings -- and Google is getting scarily good at reflecting actual human sentiment. (See Hummingbird and the returning of SERPs that are relevant to the topic but do not contain any of the wording of the search query.)
That is where Google is going. It's not 100% there yet, but it is getting closer and closer. By focusing on that end destination, you will win. Maybe not tomorrow, maybe not this month or this year. But you will win.
I think you have struck on a really interesting and valuable point. Nearly all methods used in machine learning are dumb in regards to causal direction. They learn from the independent variables described by the individual doing the analysis. More often than not, these variables are predictive and not causal. For example, a site that scores well on Panda surveys might be more likely to have high user engagement metrics. These user engagement metrics do not cause the page to be a good page, but they predict that the page is a good. Google then transforms the predictive value of these independent variables into a causal relationship with the SERPs.
One highly important takeaway from this is that as Google distances itself from the human selection of independent variables and, rather, depends on machine learning or deep learning techniques to identify potentially predictive metrics, a new opportunity arises in SEO to identify echo-chamber metrics that are self-fulfilling. A perfect example of this is the rise of behavior shaping techniques like click through manipulation which we have seen discussed on prominent blogs in the last few months. These techniques are not new, they were used a decade ago when I was first getting involved, but they are far more effective now because the line between prediction and causation has been blurred by these advanced learning models.
Thanks Russ - I think this is one of the most valuable comments and points in this discussion, and one I didn't explain nearly as well as you have.
Thanks for the kind words!
Brave of you, Rand. But I think you are right and somehow this post of yours resonates with my "Stop SEOcentrism" call to action and my endless underlining the importance of developing a strategical mindset.
Because what you are talking about is exactly about that. It's about not fooling ourselves with short breathed tactics, but designing an user-centric strategy, which obviously must be developed in always evolving and tested tactics.
I like to think that SEO is more than plain simple SEO, and that if we embrace - within of our sphere of influence - an holistic vision of marketing, not only we will have success in what we do and we will be less dependent by the designs of the Google's gods, but also we will able to have that influencing impact we always blame we don't have amongst businesses.
And so say we all :-)
Thanks Gianluca.
I wasn't simply trying to say that we need to think or act more broadly as SEOs - that's a given :-) I was trying to express that knowing what correlates with sites/pages that have success may actually be more interesting and worthy of our investigation than just what directly impacts rankings (which has been so much of our focus in SEO for so long). I realize that might be a controversial viewpoint, but it can help us recognize the second-order effects that are so massively influential in the success of so many sites these days.
Sure! And what we say are not in contrast.
What you suggest about correlation studies produces those premises that justifies the strategical view I always try to apply to my vision of SEO, and that I try to educate people about.
Now that I am employed by a full service advertising agency I can say without a doubt that our clients who engage in offline or traditional advertising make my job in digital marketing much easier. When I was freelance SEO consultant I was hyper-focused on SEO, and certainly SEOcentric as Gianluca would say. I don't think SEO's who work at an agency like mine would think this is a controversial point at all.
Rand is essentially right. I would actually go further to say there is no such thing as causation. It is epistemically inaccessible to us. All there is, and all we are looking for (and what we in SEO need to look for), is stronger observances of excluding correlation. Where a situation that is actually AQWER?>B is not construed simply by the impatient observer as A>B (that's the cause/correlation fallacy - known as post hoc ergo propter hoc in the Latins) and where we need to 1) notice / imagine / find what else may be going on there, and then 2) remove the possibility of A, or Q, or W, or E, or R, or ? to know what out of that, if any of it, in whatever combination, might cause B. There was some recent kerfuffle over the disavow... or something... as I vaguely recall, that might have had something to do with someone trying to point that out already... ;p
"There was some recent kerfuffle over the disavow... or something" :-)
This is a really interesting point of view, Rand.
Your title and reading the post made me think... it's funny how as SEOs we scream for finding correlations - which yes, we absolutely should - of success with something (Google's algorithm) that has an unknown equation that determines what results are "spit out." Ironic, no?
I'm sure some will balk at your points or read the title and jump down to the comments, but as a high-level thought piece I think this is really smart. Maybe, just maybe, strong correlation is enough sometimes...? Or, maybe there are times we claim causation when there's only strong correlation anyway..!
As you said, while we should always keep trying to obtain more information, test, and find legitimate causation, there may be times where correlation is enough.
P.S. - Loved the graphics! Animated Rand has a killer mustache, too.
Correlation is a critical aspect of recognizing potential patterns and possibly forming an hypothesis. But hypotheses are a long step shy of becoming theories. As long as that fact is kept in mind, with the attendant knowledge that taking actions based upon correlation alone may or may not provide the desired result, then correlations can be valuable. The danger, one that I see so often in our industry, is people presenting correlation as causation, claiming results from it, which inevitably correlates with (see what I did there?) a flood of copycats.
I think that's an excellent point Sheldon, and one I should include in a future post on things to takeaway (and not) from correlation. In the hypothetical I used in the post where TV ads positively correlated with competitors' rankings, for example, the response can't be "buy some cheap TV ads," but instead should be "what are the most effective ways that I can build brand knowledge, attention, and interest to achieve their kinds of success?"
Great clarification.
I was pretty rage filled when I started reading this, but it did subside as I read on! This was mainly because to me it felt you started off making one point on a more data-driven scientific point of view which I vehemently disagree with, but then the piece evolved into more about how the best SEOs don't just know old-school SEO, which I very much agree with.
This is part of the reason why I advocate in-house SEO as you get a broader exposure to all aspects of online and offline marketing techniques which can compliment your skillset. I do as much PPC and split testing as I do SEO and I sit next to our brand team who create our company mascot and book our space on billboards, along with all the other aspects which we undertake so I have exposure to all kinds of marketing. So I know a little about CRM systems, and I know the data we can collect, and the emails we send out to potential customers, which might influence how people might start searching for our website, products & services. An agency consultant who preaches about "SEO best practice" in a vacuum will struggle to see the bigger picture from their funky agency offices churning out reports and getting annoyed because we can't implement their recommendations perfectly!
I still have serious concerns though how Moz still allows spurious correlations and tiny data sets to be presented in a way which implies a single truth to those who are new to the business. I don't think it is helpful, for example, to run a survey with 128 responses from people who we have no idea if they are really experienced enough to answer and call it a ranking factors survey. That's just my opinion though of course! Others will disagree with me here.
It's just that given the recent guest Whiteboard Friday debacle, I just think the credibility of what is discussed here has been shaken a little and sometimes there needs to be a clearer understanding of proper data research methods, and what is pure opinion, personal experience and conjecture. Both are indeed useful, they are just very different beasts which aren't always that clearly labelled to the masses on here.
Hey Martin: 3 things
1) I'd love to hear more about the point of view you disagree with. Do you disagree with the idea that correlations between say, tweets that use images get more retweets or pages that receive more Google+ shares tend to rank higher are mathematically rigorous? Or with the idea that this information can be interesting/useful? Or maybe something else...
2) I think the Ranking Factors Survey is really well named, because it's a survey of opinions that are aggregated and presented precisely in that fashion (it's right there in the title! we surveyed people about ranking factors!). I think we also do a pretty careful and accurate job showing, explaining, and using naming conventions that explain the difference between the survey aggregation and the numerical correlations. They're split into separate sections, with unique titles and descriptions that explain which part is which and how the data might be useful. We also include a lot of caveats about how to approach it.
3) On the guest WB Friday - I love showcasing the opinions of folks with whom I disagree or where they have dramatically different experiences than I do. I believe Moz would be doing a disservice to our field if we only ever published our own opinions or only ever published pieces that we had personally verified. Exploratory and controversial viewpoints bring out the detectives and analysts in all of us, and I think that's a great thing. I do think we could have done a bit more to make Moz's own experiences and differences of opinions more clear and visible, and Trevor and Cyrus noted in follow-up comments.
Just a note for claryfinng this phrase you wrote in your comment:
I don't think it is helpful, for example, to run a survey with 128 responses from people who we have no idea if they are really experienced enough to answer and call it a ranking factors survey.
The correlation studies and the "expert survey" are presented in the same Moz Ranking Factors document, but are two very separate things:
Finally, in the case of the Ranking Factors research, Moz gave plenty of information about the research had been conducted and offered all the dataset for all who wanted to review the results: https://moz.com/search-ranking-factors/methodology
Excelent post Rand specially the part that instead of thinking "where's Google weak?" we should thinke "where's Google going?"
Rand, this is pretty on point with a blog post I wrote about social media correlation. Your post, however takes that to the next logical conclusion, it's not just social media, it's everything we do that correlates to good rankings. It's not about doing what SEOs do to get rankings but about building up the entire web presence to build rankings. This is mostly why optimization alone has so little value with today's algorithms. One can only do so much manipulation. Eventually, those that are just simply doing web marketing right are going to win.
This, of course, brings me to something that I've been telling clients more and more recently, (and coincides with the tweet Samuel Scott linked to in his comment above: Rankings are just the trophy for doing good web marketing. We often treat rankings as the race we are running, but too many web marketers are running the entirely wrong race. When we do web marketing right, Google rewards us with the prize.
This is a very mature-industry piece, something I'm not sure everyone will be ready for (though the thumbs up/down and the reaction to it says maybe you timed it perfectly.)
I think you're right. I I love the quote everyone else also loves:
"by playing the "where's Google weak?" game rather than the "where's Google going?" game, you've ultimately lost."
But I have to say something that often gets missed: it's VERY profitable to rank for some keywords today even if it means you won't rank tomorrow. Those playing the long term game are *still* waiting in a lot of ways for Google to clean things up.
I remember reading Wil's "Google Makes Liars of the Good Guys in SEO" post quite awhile ago but I refer to it often because it's still true. We know "casino bonus" for instance is a BIG money keyword. The first result, at least here in AU, is casino bonus 2 (doing my best not to link it) dot co
If you visit that site, it's nearly pure spam. It's also making them A LOT of money today. And tomorrow. And yesterday. So if the goal of your business is to make money, playing the "where's Google weak" game has a LOT of income potential.
For your plumbing business? Try to win & play the long game. Fitbit sales & selling your home? If you don't rank today, there's no point ranking next year.
There's gold in them thar hills for those ranking today.
Great post Rand
One thing that's probably worth mentioning here is User Experience. This ties in with your example of the website which provides the information faster in a more pleasurable manner.
It's been said a million times before but build and design for Humans and not Bots and you can't go far wrong!
Trying to trick the smart people at Google is never worth it!
This is a really timely article from Moz that is perfect for my evolving view of SEO. For example, I previously thought that meta descriptions don't improve SEO. However, a blog post I read here recently pointed out that if a good meta description encourages more people to click on a result, then the click-through rate will go up and SEO will improve. Is this causation, or correlation? I'm not sure, but either way it's good for driving traffic to your site.
Another example is the focus on topical keywords and entities instead of trying to rank for specific keywords and keyword phrases. I used to worry about fitting in my exact keywords into the content, trying desperately to make it sound natural. Now, I can just focus on writing well! As long as I've strategically mapped out the correct topics and entities, good writing naturally includes keyword variants, entities, and good old fashioned context.
I don't think I'm at the point yet where I'm capable of predicting where Google is going next, but at least I'm not so reactionary anymore!
"By playing the "where's Google weak?" game rather than the "where's Google going?" game, you've ultimately lost." Great insight. Google is smart. They figure out ways to overcome their weaknesses. Relying on this tactic means eventually it will catch up to you.
Good post Rand.
I just wonder whether I am being a little bit slow or there's actually a mistake in the Ranking Factors infographic, with the "Domain Level, Link Authority Features" being repeated twice (at 5.21% & 20.94%).
Was a typo we fixed it :)
I remember when we used to focus on certain tactics to rank sites higher. And Rand, I like the way you say instead of going after "where is Google weak" focusing on "where Google is going." It's like playing chess, you have to think 3 to 4 moves ahead. Furthermore, there's not one certain "thing" that you can do to make sure a site ranks higher. It's about the collective effort of all the factors. It's not one link that will do it or one title tag that will do it or just simply adding some Schema (I've seen big impact in Local SEO though). We have to look at the big picture.
I think it's an excellent article, Rand, and it's a great way to think outside of the box.
Am I the only one that finds it terrifying that a survey of "SEO professionals" think that social media is more directly relevant to rankings than domain-related factors?
Great contribution Rand, this sentence ("That's why, today, I suggest marketers think about what correlates with rankings as much as what actually causes them") really hits me! As a senior SEO consultant [link removed] I spend a lot of time explaining concepts like this one (or this one "playing the "where's Google weak?" game rather than the "where's Google going?" game, you've ultimately lost") to my clients, trying to make them understand that SEO is way more complex of what they think or read... maybe in a blog post from 2005 (true story).
Again, must-read-post, congrats!
Cheers from Spain,
Paolo
Excellent article Rand. The graphs speak for themselves. Thanks !
I agree Rand. It's basically DED (Domain Equity Development). It's something that is often forgotten: "Don't fight Google, embrace Google". Trying to fool or trick Google into getting better rankings will always be a short term strategy, because eventually Google will catch up to what you are doing. If you embrace Google, you don't have to worry about algorithm updates, because you are both trying to deliver the same thing: Quality Experience and/or info for the user/searcher.
Where's Yoda when you need him? "You must unlearn what you have learned".
Fascinating article. We need to open our eyes!
Thats an impressive infographic Rand, but I think social signals play more than 15% role as a ranking factor. I have seen dramatic results of things going viral and they even start beating the best in those verticals.
It is observed many bloggers don’t give much importance to all SEO requirements or they go to any extreme by either stuffing the keywords or completely ignoring them. Ultimately they have a blog with huge contents with deserted look where a few visitors hardly come and go away without any conversion. Your last point is really important. Many newbie bloggers take SEO one-off task to do while publishing a post and that’s all. It is a recurring process and as more you do it specially off-page SEO the more bigger would be your resultsThanks for sharing such wonderful info buddy.
I've always been a huge fan of Moz (esp Rand Fishkin & Cyrus Shepherd).
So please forgive me if this comment is too long, as I point out 2 things:
You said, "successful marketers have been engaging in the tactics that Google's own algorithms are chasing—popularity, relevance, trust, and a great overall experience for visitors. Very frequently, that means looking at correlation rather than causation."
If by any chance you mean successful marketers like Red Bull, Tesla or even Moz, then this isn't really that helpful either.
Please don't get me wrong here Mr Fishkin, I love this article, and will bookmark it and save it for time to come. But the question is, are we still talking about SEO in terms of discovering the most popular goal-intended keywords, and trying to rank for them?
If so, then what exactly can successful marketers really do? Other than buy a lot of AdWords and get a lot of useful keywords to target customers instead of wonderers? (not the anti-Google argument :-) but seriously) And then spend tons of money on a dedicated social media teams, an Ad agency, a PR firm, and all the digital marketing tools and CRMs you can afford?
Because as for social signals and tv ads and PR campaigns, they will cost a lot of extra time and lots more money for some company looking for just anSEO. And that again underscores the fact that "successful marketers" are simply RICH clients!......Or rich SEOs (digital marketers) who have the balls to demand premium charges.
Correlations are definitely great, but I don't agree that you need to look at "successful competitors" too much. You'll most likely end up seeing what you already knew you'd see. They probably spent much more money!
The big picture of marketing boils down to spending more money on marketing, which is what the kind of clients that turn to SEO are looking to not do!
Also from one of your point, "Meanwhile, your competitor (who has no idea about the algorithmic impact of this factor) has been working on a new design that makes their website content easier, faster, and more pleasurable to consume". Well what I see here are still causative factors, not correlative.
Or am I wrong about that?
Traditional SEO tactics clearly pushes SEOs to improve page loading speeds and user experience, right? I was told that that was causal, isn't it?
I totally agree with you on the idea of Big Picture marketing. In fact I think it is high time that digital marketers (starting with you guys in the developed world) begin to boldly challenge the justification for why traditional advertisers can command the kind of service fees they command.
I also agree with Bill Slawski's comments about the gains in understanding patents. "individual signals that could be used together (length of stay on a page in a search result + continuation or change in topic of the immediate query session)"
If anything at all, it helps us to learn how not to try to trick the algo, and just become quicker in anticipating the kinds of practices that could "help" us next (these won't exclude tv ads, billboards, print, social)
In short Mr Slawski's comment is in the spirit of this entire post.
Which is (and i'll try to bind it up in one sentence):
that SEOs should become more expansive with their marketing strategy because Google will combine much more than one signal (e.g. a tuple: <document , query, data>) to validate its reason to rank or not rank your website!
:-)
Rand, I agree with the sentiment entirely, though the statistical sounding description of working to correlation sits uneasily (as you admit). I'd suggest another way of phrasing the same proposition: don't worry about the technicals and focus on the fundamentals - the essence of what you're saying.
But to keep a sharp edge to proactive digital marketing, the patterns and flows of digital will always allow for areas to find and exploit competitive advantage (a la 'not all patents have been applied for'), to do things differently and better than others - still room to gain from being clever (if perhaps not quite as out and out clever as Google).
I just wish, that more people would at least try to understand the difference between causation ad correlation. I found it very hard sometimes to explain it to customers and (some) other SEOs who couldn't quite get the meaning of both words.
Rand, thanks for this piece. Thinking about this a bit, i believe you are not actually talking about the difference between causation and correlation, just about long-term strategy vs. short-sightedness. To clarify:
Take the example of the mobile app:
Concluding a causation from the positive correlation ist not valid (Mobile App -> Rankings = wrong). Why? Because you cannot exclude the possibility of the causal realtion being Mobile App -> Z -> Rankings (Z could be anything, even moustache length ;) )
You suggest the guidance of the correlation nontheless in order to form a long-term success strategy.
However it would be much more scientific (and useful for strategic planning, IMO) if one could derive a method to discover Z.
Feeling validated that I've chosen to train new team members with a less is more policy. You don't need to be able to dissect 600 algo changes to work in Search.
As SEO's we are focused on getting a website to rank but this article does help us look at the big picture of overall marketing. We need to look at whatever we can do to bring traffic and conversions to a website. Like a mobile app or tv ad as mentioned.
I think that the guy who isn't afraid to go against the correlation (some of the time) is often the one who really does well because he can sense "changing tides" or "emerging trends". Some people are much better at smellin' this stuff than others.
So, what I would really enjoy seeing in the ranking factors is "who is sayin' what" rather than the overall numbers.
It is no different than Moz Q&A. If certain people say things I can be strongly inclined to take their advice or ignore it.
If you only go with the overall trend or some "consensus" then your results are going to be "average". To really bust out you gotta be able to divine what's happening and then act to get "first mover advantage" or better yet.... "only mover advantage".
I think that SEO is a lot like fishin'.... where 10% of the fishermen catch 90% of the fish.
That's another great point EGOL. If correlation is showing you that everyone on Facebook is winning with clickbait headlines right now, the best marketing strategy may be to go the other direction and be the exception to the rule. And maybe if you see that the SERPs in your field are dominated by sites that engage in high volume content marketing (that's mediocre in quality), you should try investing in fewer pieces of content and more in other areas you think can earn you those types of signals.
That's the marketing value of disruption :-)
Oh... thank you Gianluca. I now have a better understanding of disruption. :-)
Each time when someone claim to found "Google loophole" i remind about this site Spurious Correlations and i remind about it one more time.
Breath of fresh air Rand! You tell it like it is!
When I do research and talk about Google LOCAL search theories I have, I keep saying:
"I know correlation is not always causation BUT..."
You explain that big BUT better than I could Rand and how that's sometimes the best thing we have now given Google's complex algorithm.
This makes a lot of sense. For "SEO tweaks and tactics" can fail to deliver at times. Mostly, teaching us a lesson on how we know nothing about the complexity of the algorithm (though we like to think we do). There has to be some sort of balance between signals beyond crossing them off from a checklist one by one. I've heard an interesting theory, that Google might, for example, calculate the ratio between branded searches and acquired links. So there is high probability that it's not about crossing off guidelines form a checklist but rather creating something similar to what Google values. And this, more increasingly, happens to be what searchers value.
"...but by playing the where's Google weak? game rather than the where's Google going? game, you've ultimately lost."
Great statement Rand!
Great article Rand! It seems there is a mistake in the last infographic. 5.21% is attributed to "Domain-Level, Link Authority Features" and it should be "Domain-Level, Keyword-Agnostic Features" according to https://moz.com/blog/visual-guide-to-keyword-targeting-onpage-optimization.
Yep was a typo on our part we fixed it. Thanks for pointing it out.
Interesting perspective. Thank you! Following the "big picture" can also be much more fun, since you work for better online presences, i.e., the user, and do not fight against the ranking algorithm.
If you argue with this thinking, please become an "seo" in a niche I operate in :).
Well pulled together and explained Rand!! True marketers in the SEO world saw this years ago (as Moz did..notice what they've been focusing on with their own site over the last 3-5 years - including a rebrand away from SEOmoz!) and adjusted. It's starting to truly pay dividends.
The next step, in my mind, is to be aware of how RAPIDLY Google is re-working their SERPs to drive more ad revenue from their existing products and open the doors for new ones, while simultaneously increasing their "Google is the only place you need to go" image. The continued expansion of the data they are able to collect and analyze through "free" software is quite impressive. There is a reason they fear Amazon, who may not seem like the most natural competitor.
This is a dangerous point and the next 3-5 years will be incredibly interesting as some brands begin to realize they've been helping grow their own worst enemy IMO.
Build your brand, enjoy organic traffic while it's still relatively easy/cheap to capitalize on, and be aware that the loyalty and eco-system you're building now may be a big part of how your business is sustained in the future without having to pay to play much of the time.
Final thought on those chasing Algos/running risks. It's easy to bash them, but the fact is the only reason we have jobs is because this stuff makes businesses money. Sometimes, short term gains are well worth the price you pay (JCPenny anyone?) - although that's becoming much more rare.
Rand, you are spot on, though I hate to admit it. And scanning the comments here it appears my feelings are shared by others. Why do some SEOs hate this thesis? Because, I believe, it requires a LOT more work and engagement in many other facets of marketing that they didn't have to deal with in the past.
This isn't as easy as it once was. Do "X" on page, submit "Y" off page and Voila! Instant rankings and something we could easily share with a client or boss to validate their investment in us.
Now there are so many "correlative factors" that it is hard for the client/boss to connect the dots and associate value with the SEO's contribution.
So you're saying the longer my mustache the more sales I may get? ;)
Whilst it's fantastic to aim for the where is Google going trying to explain that to a client can be difficult especially if like in your example there is a drop in rankings. even harder if the loop hole can last for a few months or more.
Great post however, look forward to reading the comments on other peoples thoughts.
You got it! Mustache = rankings :-)
In the time I've worked for an Internet marketing agency, I've seen a lot of shift from "here's what we show Google" to "here's what people want." It seems to me that Google's changes over the last few years have forced us to focus on marketing to real people instead of just manipulating what the bots see.
With that in mind, I loved this line, Rand: "You found an algorithmic loophole and exploited it briefly, but by playing the "where's Google weak?" game rather than the "where's Google going?" game, you've ultimately lost."
And this is the kind of post that caused me to be a Rand Fan to begin with back in the day. I have always appreciated the fact that you are more about looking at the big picture / long term to know what to do to achieve long term results. It is that approach that causes me to look forward to the large algorithm updates rather than dread them. I am done with the days of trying to explain to a client why their rankings just tanked. "um, yeah we built your rankings and consequently the business coming from it on a thin, shady branch - and Google decided to trim that branch off the tree - sorry" ;)
Echoing the sentiments of so many already:
Great post Rand. Always great to see magic in action - especially when it is rooted in common sense. As good as understanding the current inner workings of the beast that is Google sounds, it is definitely a better move from a strategic perspective to know where the future lies.
I have some neolithic coworkers who will have to read this blog and address your points or face my fairly limited wrath. Moving from "total" SEO to general marketing is a change that is long overdue and one I am pushing for - always great to know that I'm not alone in this struggle.
On-site and technical SEO is still a factor, but creating genuine working relationships is quickly becoming even more important (if it hasn't already). I for one am all for it.
Many thanks again for your experience and insight.
Thanks Rand this is one of those articles where you can express and engage in a good conversation with clients. Specially with those who bring in someone new/old who knows direct causation of the ranking - mostly links,onpage etc which are part (small now as the evolution) of the big picture but not the picture itself.
Thanks rand. good post.
I like this post by Rand, this is actually a different perspective of looking at things. I mean, these days, it is obvious that we need to take a holistic way of doing SEO and traditional SEO strategies do not really work that well in the long run.
But as Rand is trying to say here, why shouldn't we focus on correlations as well (in some cases, may be more than causations)? Looking at the websites which are doing well can certainly provide us a lot of useful insights and can further improve upon our SEO performance. So, why not include this aspect in your overall SEO strategy?
I mean, you can just deny this perspective, but I think it stands quite logical. Doesn't it?
Thanks Rand, like always great post..
Thanks for sharing the article. It was great going through it and have learnt many new things.
This post is incredibly well done. Evergreen content.