- Oct 9, 2013
- 3,115
- 12,589
This is a slightly more advanced topic that a lot of you probably won't know about.
It involves analysing competitors to get their anchor text ratios then averaging that and trying to "match" that with your own link building.
I'm not going to do a deep analysis here on why it's bad. I have a fairly good understanding of mathematics, statistics and systems to understand at a gut level that it's just not a reliable method.
Here's a post from Matt Diggity that explains this method. https://diggitymarketing.com/anchor-text-optimization/
The reason I don't feel this is hugely useful is because the google ranking algorithm is incredibly complex, and this method is essentially trying to simplify a complicated algorithm into a set of workable rules.
You have a complicated system with 1000's of variables that are influencing ranking, and you're looking at a tiny subset of those in an attempt to get the equivalent output.
Let me try and come up with an analogy.
Imagine you want to maximize academic success for a group of students.
So you look at the top 10% of students and you analyse their study methods. You then take the average of their study methods and apply that to your group.
It's semi-useful, but you're still missing so much. The intelligence of the students, their discipline, their motivation, their individual teachers. It's such a complicated system with so many factors.
Ranking is like this.
What's also making this even harder is you have to classify the anchor texts yourself. Exact, partial, longtail, topic, title, empty. There's so so many "types" of anchors. None of us really know how google views them.
What makes an exact bad? What makes an exact an exact. Is it search volume? Or is it just how optimized a page is for that keyword. It's an incredibly complex topic and you're essentially just making up your own groups like "topic keywords", then deciding which fit into your made up category.
Now, even IF we had exact classifications of anchor types that we knew google used, this STILL doesn't take into account the influence of other factors.
For example.
The age of a site. The topical relevance. The trust of a site. The authority. Even these things are just made up. What is authority, what is topical relevance? We don't know how google classifies any of this. We can only make a best guess, and mine is you have :-
1) Topical relevance
2) Trust
3) Power
Words like "authority" are just a combination of power, trust and topical relevance.
This is exactly why SEO is an art, not a science. It *IS* a science if you can see how the algorithm is made, but like any complex system, without knowing the rules, you have to treat it as an art.
So how does trust, power and topical relevance influence things? The average anchor text ratio analysis method doesn't take any of these into account. It treats everything like it's the same, and you have some magic formula.
A high level mathematician could show you math that proves this method doesn't work. If we took a large sample, categorized the anchors, and performed an analysis for both the a general sample across all niches, and looking individually at niches, then groups of keywords, then individual keywords we would find no patterns.
We don't know the "weight" of each link. One brand link could carry more branding/trust power than another, so because of that one site might be able to get away with more anchors.
Likewise, we don't know if more dangerous anchors like exact anchors also carry weight. Knowing what I know about engineering, I'd say they do. For example, 5 exact anchors from DR90+ sites will not be viewed the same as 5 exact anchors from small blogs, and they certainly aren't viewed the same on directories/comment links.
Google does expect certain type of links to be a certain type of anchor. Exact on comments are in 99.9999% of cases complete spam.
So you have a machine that takes 1000's of variables as input to produce an output. You don't know what those variables are or how they're classified. And you then decide to guess at 10 of those variables, average them, and try to make your variables near that average, and expect to get the same output.
This is not how to approach SEO.
SEO is an art.
Therefore, you should try to understand, from a 50,000 foot view how the system generally operates. What kind of things it likes and dislikes and how it generally approaches producing its output.
And the biggest thing we know about google is it's looking for UNNATURAL signals. So our primary concern shouldn't be micro-optimizing ratios, but trying to stay natural.
Looking at what kind of links sites with more natural links get, what kind of anchors they get, what kind of link building velocities can occur naturally, and then aiming to be as close to that as possible.
It's generally not overly complex.
Rather than try to create some fancy anchor text plan, just for the most part build lots of brand links to the homepage. Add in a few phrase/misc brand like "visit Brand", "toasters at Brand". To your inner pages, do a couple of url, a couple of phrase brand, like "this product is reviewed at Brand", and a couple of long phrase, "the super duper rigatator is reviewed here".
It involves analysing competitors to get their anchor text ratios then averaging that and trying to "match" that with your own link building.
I'm not going to do a deep analysis here on why it's bad. I have a fairly good understanding of mathematics, statistics and systems to understand at a gut level that it's just not a reliable method.
Here's a post from Matt Diggity that explains this method. https://diggitymarketing.com/anchor-text-optimization/
The reason I don't feel this is hugely useful is because the google ranking algorithm is incredibly complex, and this method is essentially trying to simplify a complicated algorithm into a set of workable rules.
You have a complicated system with 1000's of variables that are influencing ranking, and you're looking at a tiny subset of those in an attempt to get the equivalent output.
Let me try and come up with an analogy.
Imagine you want to maximize academic success for a group of students.
So you look at the top 10% of students and you analyse their study methods. You then take the average of their study methods and apply that to your group.
It's semi-useful, but you're still missing so much. The intelligence of the students, their discipline, their motivation, their individual teachers. It's such a complicated system with so many factors.
Ranking is like this.
What's also making this even harder is you have to classify the anchor texts yourself. Exact, partial, longtail, topic, title, empty. There's so so many "types" of anchors. None of us really know how google views them.
What makes an exact bad? What makes an exact an exact. Is it search volume? Or is it just how optimized a page is for that keyword. It's an incredibly complex topic and you're essentially just making up your own groups like "topic keywords", then deciding which fit into your made up category.
Now, even IF we had exact classifications of anchor types that we knew google used, this STILL doesn't take into account the influence of other factors.
For example.
The age of a site. The topical relevance. The trust of a site. The authority. Even these things are just made up. What is authority, what is topical relevance? We don't know how google classifies any of this. We can only make a best guess, and mine is you have :-
1) Topical relevance
2) Trust
3) Power
Words like "authority" are just a combination of power, trust and topical relevance.
This is exactly why SEO is an art, not a science. It *IS* a science if you can see how the algorithm is made, but like any complex system, without knowing the rules, you have to treat it as an art.
So how does trust, power and topical relevance influence things? The average anchor text ratio analysis method doesn't take any of these into account. It treats everything like it's the same, and you have some magic formula.
A high level mathematician could show you math that proves this method doesn't work. If we took a large sample, categorized the anchors, and performed an analysis for both the a general sample across all niches, and looking individually at niches, then groups of keywords, then individual keywords we would find no patterns.
We don't know the "weight" of each link. One brand link could carry more branding/trust power than another, so because of that one site might be able to get away with more anchors.
Likewise, we don't know if more dangerous anchors like exact anchors also carry weight. Knowing what I know about engineering, I'd say they do. For example, 5 exact anchors from DR90+ sites will not be viewed the same as 5 exact anchors from small blogs, and they certainly aren't viewed the same on directories/comment links.
Google does expect certain type of links to be a certain type of anchor. Exact on comments are in 99.9999% of cases complete spam.
So you have a machine that takes 1000's of variables as input to produce an output. You don't know what those variables are or how they're classified. And you then decide to guess at 10 of those variables, average them, and try to make your variables near that average, and expect to get the same output.
This is not how to approach SEO.
SEO is an art.
Therefore, you should try to understand, from a 50,000 foot view how the system generally operates. What kind of things it likes and dislikes and how it generally approaches producing its output.
And the biggest thing we know about google is it's looking for UNNATURAL signals. So our primary concern shouldn't be micro-optimizing ratios, but trying to stay natural.
Looking at what kind of links sites with more natural links get, what kind of anchors they get, what kind of link building velocities can occur naturally, and then aiming to be as close to that as possible.
It's generally not overly complex.
Rather than try to create some fancy anchor text plan, just for the most part build lots of brand links to the homepage. Add in a few phrase/misc brand like "visit Brand", "toasters at Brand". To your inner pages, do a couple of url, a couple of phrase brand, like "this product is reviewed at Brand", and a couple of long phrase, "the super duper rigatator is reviewed here".