seoMoz Search Engine Ranking Factors: Underscores beat Hyphens in URLs

[deepsand]

The 2011 seoMoz Search Engine Ranking Factors report is out, and contains some interesting data re. the ongoing debates over the merits of hyphens and underscores in URLs.

Page Level Keyword Agnostic Features — Correlated Data (http://www.seomoz.org/article/search-ranking-factors#metrics-8) shows the following:

Underscore in URL - 0.02
Hyphen in URL - negative 0.04


This means that, while the presence of an underscore in a URL gave it a small upward boost in the SERPs, the presence of a hyphen gave it doubly large downward push!

While these correlation coefficients are admittedly small, they certainly stand to counter the claims that hyphens are superior to underscores, and that underscores are in some way problematic.

[AboutWeb]

I always thought that hyphens are meant to separate words and create SEO friendly URLs, and underscore is not recommended. I am confused now.

[SteveGerencser]

In the beginning, before there were SEs, we always used underscores for spaces.. Somewhere along the way the rumor got started that hyphens were equal to underscores.. Shortly after that we heard that hyphens were better than underscores.. And now no one uses underscores because they are under the impression that it matters.. Apparently it does.. Just not the way people (MC) said..

And here it is.. The source rumor that caused all the issues..

http://www.mattcutts.com/blog/dashes-vs-underscores/

[deepsand]

I can still recall the first time that I read that post, thinking to myself "wtf?"

And, despite the fact that it is easily demonstrated that Google parses Word1Word2Word3, Word1-Word2-Word3, and Word1_Word2_Word3 into Word1 Word2 Word3, the myths continue to live on.

[deepsand]

I always thought that hyphens are meant to separate words and create SEO friendly URLs, and underscore is not recommended. I am confused now.
See Hyphens in URLs - Are Hyphens in URLs Good or Bad for SEO: Post No. 19 (http://www.webproworld.com/webmaster-forum/threads/98772-Hyphens-in-URLs-Are-Hyphens-in-URLs-Good-or-Bad-for-SEO?p=566062&viewfull=1#post566062) for a test that you can quickly and easily do in 3 simple searchs.

[dburdon]

These are very low levels of correlation. I tend to use both, with a slight preference for hyphens.

[MrGamm]

I don't know... sounds strange to me... but perhaps from the perspective of the English written language...

http://en.wikipedia.org/wiki/Hyphen#Varied_meanings


A man-eating shark is a shark that eats humans.
A man eating shark is a man who is eating shark meat.

So maybe if I was searching for an article on "Japanese Shark Recipe" one would be more valid than the other? Or perhaps if I was searching for the word "eating" the one without the hyphen would be more relevant?

I know that is poor example, but a hyphen is actually a single word, where words without a hyphen are actually two words, not one.

The rest of the report is interesting as well.

[deepsand]

The English language is lacking for consistency re. the usage of hyphens.

Both man-eater and man eater are accepted as being correct, and mean the same.

The phrase "man eating shark" actually allows of two interpretations, depending on the context.

[deepsand]

These are very low levels of correlation.
That they are.

But, the salient point is that underscores are in no way inferior to hyphens. And, are parsed identically.

[AboutWeb]

See Hyphens in URLs - Are Hyphens in URLs Good or Bad for SEO: Post No. 19 (http://www.webproworld.com/webmaster-forum/threads/98772-Hyphens-in-URLs-Are-Hyphens-in-URLs-Good-or-Bad-for-SEO?p=566062&viewfull=1#post566062) for a test that you can quickly and easily do in 3 simple searchs.
OK I tried that on google. The term about_web gives 15,500,000 results, aboutweb same 15 mil., about-web 2,030,000,000 results and about web 1,980,000,000.
Now, this tells me that underscore did not separate two words, only one thing got weird, the additional results when I used the dash.

Later Edit: I think when I included the dash, the results changed, websites that had keyword+dashes in title/url ranked better.

[MrGamm]

But, the salient point is that underscores are in no way inferior to hyphens. And, are parsed identically.

Its not the best example. I know that. So lets try looking at it from a different perspective.

Maybe a word or phrase which is out of context, absent from menus, is disconnected to any document content, is a signal of keyword stuffing and causes a hit to the rankings?

Thus, underscores create a single word, and it not perceived as many single words, or many spammy keywords stuffed throughout the document.

It could be anything. Their results are probably to one degree or another specific to what they are doing with their documents. Right?

[deepsand]

OK I tried that on google. The term about_web .
You didn't do the test exactly as I prescribed.

I provided a specific query string for which I knew that there were pages indexed that were essential to the test.

[deepsand]

Thus, underscores create a single word, and it not perceived as many single words, ... .
Both try the test that I prescribed; and, re-read the statistical findings.

Underscores do not "create a single word." That is the myth!

[AboutWeb]

And, despite the fact that it is easily demonstrated that Google parses Word1Word2Word3, Word1-Word2-Word3, and Word1_Word2_Word3 into Word1 Word2 Word3, the myths continue to live on.
What did I did wrong in my test ? aboutweb, about-web, about_web, about web.

[deepsand]

What did I did wrong in my test ? aboutweb, about-web, about_web, about web.
You picked query strings for which it is unknown what matches should be had. I.e., there is no control data.

You need to first have a page which you know is indexed for Word1 Word2 Word3 before you test the query strings Word1Word2Word3, Word1-Word2-Word3, and Word1_Word2_Word3 to see if they get a hit on Word1 Word2 Word3.

That is why I said to use the 3 query strings I posted elsewhere, as I know that the target page in question exists and is indexed.

[deepsand]

Domain Level Keyword Agnostic Features — Correlated Data (http://www.seomoz.org/article/search-ranking-factors#metrics-2) shows the following:

# of Hyphens in Domain Name - negative 0.01

[MrGamm]

Both try the test that I prescribed; and, re-read the statistical findings.

Underscores do not "create a single word." That is the myth!

The test would have to be a page in the index which had no links pointing at it, and the words joined in the content could not be in the document itself. Otherwise, a filter could simply be applied regardless as to what the search query returned, to bold matching words.

[deepsand]

The test would have to be a page in the index which had no links pointing at it, ...
Immaterial to determining how query strings are parsed and matched against Google's indices..


... and the words joined in the content could not be in the document itself.
In which case you have no controlled test at all. The potential target strings must first be indexed if their presence or absence in the SERPs is to be observed.

The test that I set forth demonstrates that all 3 forms of said query string are matched both the spaced and concatenated forms of the 3 individual words contained in said query strings.

[MrGamm]

What I am saying is that what might seem like a single word based on the bold formatting in the search results, might not be a word used to return the result in the ranking.


Immaterial to determining how query strings are parsed and matched against Google's indices..

Simply not true.

[deepsand]

What I am saying is that what might seem like a single word based on the bold formatting in the search results, might not be a word used to return the result in the ranking.
Huh? Google only bolds terms found in the query string.


Simply not true.
Well, then, explain precisely how links in any way affect how the parser analyzes query strings.

And, explain how links affects how the search algorithm finds matching values in the indices.

I.e., in both instances, demonstrate that link data are input variables into both the parsing and the search functions.

[MrGamm]

Huh? Google only bolds terms found in the query string.

Exactly. Last I looked they might go so far as to bold the exact words used in a query rather than synonyms or pluralized words as well.


Well, then, explain precisely how links in any way affect how the parser analyzes query strings.

What parser? Like a sanitizer for the POST input or do you mean to say how the Big Table indexes keep reference to documents.

When I point a link at a website, such as CLICK HERE, that text makes that page relevant to those words.

http://www.google.com/search?q=click+here


And, explain how links affects how the search algorithm finds matching values in the indices.Google uses some proprietary database technology called Big Table. I imagine it is just a relational database.


I.e., in both instances, demonstrate that link data are input variables into both the parsing and the search functions.In what instances?

[deepsand]

What has any of this to do with the issue at hand?

Either substantiate your claims re. links affecting the parsing of query strings, etal., or withdraw.