Big changes in Google content scoring

[Spielmann]

Thanks a lot. The info is really useful.

 

[C.S.A.]

ipopbb:

What steps are you taking to make sure you're not seeing sandbox testing effects? Google has publicly stated that they provide different results to different subsets of users as part of their rank testing.

I would, at minimum, search from different locations, without user accounts, and using fresh and stale ones, over the course of at least a week or two. Getting statistical confirmation on data like this is important.

Do you have several different experiments running at the same time to confirm the results?

Are your pages otherwise filled with non-duplicate content that is at least machine-recognized as semantically correct? I assume any content you use mirrors your client content, but I would be interested to know if you have similar tests using different lengths of pages.

If I get motivated, I may build a script to construct similar test pages. That would make tweaking things and deploying more tests a lot easier.

M.A.D:

You really misunderstand how the engine is tweaked. Yes, off-site SEO is very important, the OP didn't say otherwise. He doesn't say content isn't important either. There's no google engineer who says "title tags suck". If they see a highly ranked page that is clearly inappropriate, they feed that into the model, and the SERPs are adjusted appropriately. One of the factors used in that model is the tags which contain the keyword.

 

[noph34r]

Content scoring is based on algebra rather then tags a search engine might or might not like.
If I were you I would do some more research on "Term Vectoring".
More on it can be found here;
http://www.hray.com/5264/math.htm
A later evolution that is based on "Term Vectoring" is a little thing called "PageRank".

Also there is a conference, sponsored by the National Institute of Standards and Technology, called TREC (Text Retrieval Conference).
They release summaries each year, which are worth the read if you like content scoring.
The first few parts are free to read on google books. Here's a link to the first;
http://books.google.com/books?id=1-...Retrieval+Conference&cd=1#v=onepage&q&f=false

The whole idea of tags doesn't add up for the following reason.
1.The content isn't scored but the tags are scored. Search engines are about content.
And providing the best content with the query.
Ditching content for the use of tags goes against there very nature.
2. The whole idea of someone sitting in a office deciding that the following tags are crap.
Doesn't go down well with me.
Especially in the way that you state,

Which means that not only does someone in a office say those tags i don't like.
He also comes in ones a year and changes them randomly. "This year i'll like those better", "Title tags are so last year".

A more logic explanation is that google acquires documents via a http connecting.
They send a GET request to your server, when document is send to them, response headers are included.
In the response header the following is included;

LAST-MODIFIED: Date and time at which the server believes the resource was last modified.

After the spider re-crawls your document sees the changes you made, the document gets a higher value because it was updated with new information.

Cheers

I agree and you cite great information. It's so useless ro make a rank for tags. The result one achieve with a site is not replicable with other ones.

I find absolutely not true that (meta:description).content could be so low, in the SERPs results we can read
1) Page <title>
2) (meta:description).content
It's a vital element.

 

[theviralguy]

thanks, some great information, what is a "samp" tag please

 

[mylander]

Thanks for the heads up. I've also noticed that keyword density in titles greatly increases your chance of ranking if it's good.

"just keyword" vs. "just keyword with catchy title"
^^^^^^=always wins for me. As far as on-page

 

[blackhataffiliate]

Thanks for the heads up. I've also noticed that keyword density in titles greatly increases your chance of ranking if it's good.

"just keyword" vs. "just keyword with catchy title"
^^^^^^=always wins for me. As far as on-page

So you are saying you have better luck with the main keyword in title only

ex - free acai berry

instead of

free acai berry :: the free acai berry site

or

free acai berry scam

 

[ipopbb]

The whole idea of tags doesn't add up for the following reason.

I would be foolish to disagree with you. But I can't disregard my own successes in SEO. Google sorts the tags and the sorting changes over time and my last-modified dates on the documents do not. The data might be a "Fuzzy Indicator" (a modified boolean that accounts for how much on or off by allowing values between 0 and 1) I agree there is probably not a person picking tags... There quite possibly could be a Baysian categorizer saying how probable a page is trying to artificially influence it placement. Statistical categorizers would eventually train down to the tag. The rankings of the tags coming from google have improved my sites placements. Beyond that I don't know... I'm sharing my data in hopes of learning more from it.

I read the links you posted. Thank you for those. They have given me some ideas and I need to read the text retrieval one again and wrap my mind around it some more.

M.A.D. Please also take a look at my other thread. "More data from the SEO Lab"... I have charts and graphs of other SEO factors that could use a knowledgeable skeptic's perspective.

I am with you that extraordinary claims require extraordinary proof. The only extraordinary proof I have is that my revenue attributed to SEO has seen 90-130% growth year over year for the past 11 Months. I'm averaging 5 page 1 links for each of about 40,000ish high competition keywords... (I manage SEO for 15 sites PR 6-8, 9 of them are online retail.) Unfortunately, due to the nature of my work I'd be foolish to publish or even share those sites. The only way I can deliver extraordinary proof is to hope others try and have similar results.

Do you have ideas for how I should modify my experiment or are you saying the experiment is not worth continuing (I do disagree with the later at this point, but obviously I have my reasons.)?

Thanks

 

[ipopbb]

ipopbb:
What steps are you taking to make sure you're not seeing sandbox testing effects? Google has publicly stated that they provide different results to different subsets of users as part of their rank testing.

Right now, all I have is time. And access to 2 other aged analogs to my experiment which produce the same ordered list. These experiments have no SEO performed upon them. No inbound links from external sources and 0 changes or modifications since their launch. We do know that young analogs to this experiment produce slightly different lists for their first 6-10 months then fall inline with the aged analogs. So there is a 6-10 month age factor that impacts the order. Only about 10 or fewer tags are effected by sites being new. I think the algorithm accounts for certain tags being less trustworthy on younger sites. I don't know for sure but it sounds like a Baysian outcome if site age is a factor which is what I think google is doing with the content scoring. I wish I knew what they were categorizing... but I instincts are smacking with Baysian outcomes from this experiment.

I would, at minimum, search from different locations, without user accounts, and using fresh and stale ones, over the course of at least a week or two. Getting statistical confirmation on data like this is important.

That's a good idea. I have been using a cookie free session each time but I wonder how accounts or user agents or browser capabilities might change the results. Time is my enemy on which experiments get done or not. You have me thinking on this one though!

Do you have several different experiments running at the same time to confirm the results?

I have one I control and access to the results of 2 others that were entrusted to me. The outcomes are consistent across 3 experiments 1 year + created by 3 different people on 3 different servers.

Are your pages otherwise filled with non-duplicate content that is at least machine-recognized as semantically correct? I assume any content you use mirrors your client content, but I would be interested to know if you have similar tests using different lengths of pages.

No content. Here is the actual page source for the SAMP HTML tag.

<html>
<head>
</head>
<body>
<samp>GlobalyUniqueSecretKeywordThatOnlyExistsInMyExperiment</samp>
</body>
</html>

If I get motivated, I may build a script to construct similar test pages. That would make tweaking things and deploying more tests a lot easier.

Please do, and if you age it to 10 months I am up for sharing secret words.

M.A.D:

You really misunderstand how the engine is tweaked. Yes, off-site SEO is very important, the OP didn't say otherwise. He doesn't say content isn't important either. There's no google engineer who says "title tags suck". If they see a highly ranked page that is clearly inappropriate, they feed that into the model, and the SERPs are adjusted appropriately. One of the factors used in that model is the tags which contain the keyword.

Thanks for the support. M.A.D. is right in that it is false until proven true. Unfortunately it is difficult to prove publicly without shooting myself in the foot which would be dumb. I'm open to suggestions but the kind of proof purists want I can't share.

Ultimately I'd like to create a scoring system that tests google then explains why the results are placed as they are. I'm still quite a ways off on that goal.

 

[ipopbb]

I agree and you cite great information. It's so useless ro make a rank for tags. The result one achieve with a site is not replicable with other ones.

I find absolutely not true that (meta:description).content could be so low, in the SERPs results we can read
1) Page <title>
2) (meta:description).content
It's a vital element.

It's just empirically measured data. M.A.D. doesn't say my data is bad. He disagrees with its interpretation which is fair. You are using your beliefs to say the data is bad, which isn't science method. That's just being in denial about the data.

I disagree. I replicate results across 15 websites page rank (6-8). There is little chance for others to replicate my results because content score is only one of the factors I use in my overall SEO strategy. It's a big one in my strategy because it is one where I control or influence most if not all the known variables.

 

[chockomonkey]

Thanks ipopbb for this information and discussion. I hope to one day know enough about SEO to create my own lab for testing.

Cheers mate.

 

[Deleted member 107938]

It blows my mind that url is so low on this list. Has it been low for a long time?

 

[ipopbb]

Thanks ipopbb for this information and discussion. I hope to one day know enough about SEO to create my own lab for testing.

Cheers mate.

Thanks!

Funny thing is... I created the lab BECAUSE of how much I don't know about SEO.

LOL.

 

[ipopbb]

It blows my mind that url is so low on this list. Has it been low for a long time?

URL has never been in the top 20.

I have separate tests for
Domain name
Host name
Directory Path
Document Name
Query Parameters
Anchor Text

URL represents best slot out of all... In this particular data URL was a Document Name match which is usually the case.(I just rocked your world didn't I? )

I really started to succeed in SEO when I abandoned the beliefs I built up by reading the same regurgitated SEO advice that appears everywhere on the web. Everyone can't do the same thing and succeed. That's not how ranking works and with Google the rules are constantly changing about every three months.

The question is what does work and why, I don't really know, but I have interesting data to share and I do better than most at SEO! That's all I have going, which really isn't very much for anyone outside of my shoes. My advice: The best thing you can do is to start finding things out for yourself. Then you'll know for sure. (And don't ever experiment with your actual sites... Experiment with throw away sites always... Its just not worth the risk. SEO is a multi-million dollar channel for me. Mistakes are not allowed. Pissing off Google... not allowed. But blindly publishing content without understanding the forces that be... not allowed.)

 

[GreyWolf]

Wow this is some great info.

The average person has no way of testing these things, that's why its so hard to reverse engineer the algorithms. It's amazing that you've A) been able to design a system to do testing, and B) have access to a large enough site to conduct the tests on. Hopefully you'll continue keeping this info up to date.

Thanks so much for your time in conducting your experiments, and even more thanks for sharing them here.

I'm going to post a link to this thread in my basic seo thread. Basic SEO practices for newbies

Hopefully people can learn to use proper html on their pages, and using your test results get some real help for their on-site seo.


Great Share :beerchug:

 

[GreyWolf]

I agree and you cite great information. It's so useless ro make a rank for tags. The result one achieve with a site is not replicable with other ones.

I find absolutely not true that (meta:description).content could be so low, in the SERPs results we can read
1) Page <title>
2) (meta:description).content
It's a vital element.

The fact that the description meta tag can be read in the SERP has no bearing to it's value in your position in the index.
Google flat out says in official statements that they do not consider the content of the description meta tag at all for indexing. The purpose of the meta:description is solely used to provide the summary displayed for your listing.

What this means is that you should not try to optimize the meta:description for the SE, you should optimize it for the human reader. The description is you opportunity to create a sales blurb to convice the google user to click on your listing instead of any of the other listings they see on the SERP.

 

[ipopbb]

You probably going thats what i've been saying the whole time.

I think you are right that we aren't entirely in disagreement. I think you are also correct that my communication of what I am seeing in my experiment is lacking.

I'll try again this time hopefully with better clarity even though I know my observations depart from concepts you have provided documentation for and claim Google is using.

Google is sorting my experiment pages which test a secret keyword in different semantic nodes of HTML. There is a definitive order that is not random coming from Google that slowly changes over time.

I do not change my experiment pages yet the sort order of them coming from google slowly changes over time. If Google is Tag agnostic I would rarely expect change in the order of unchanging documents.

Upon swapping better ranking semantic tags, which is what I can alter in my source, as deemed by my experiment I am seeing substantial and continued gains for the 15 aged sites I manage. The reasons for this aren't entirely understood but I am inclined to think the content changes I made based on the experiment are somehow related, but I can't prove that at this point even though my observations are very suggestive.

I like the idea of tag agnostic content scoring but I believe that when Google starts sorting documents by relevance that it ultimately is telling you something about the "learned reputation" of the semantic usage, all the way down to the specific tag being used. I postulate this because my secret word being tested is imaginary and has no context or definition and isn't used anywhere else on the internet yet google still finds a way to sort the test pages consistently and somehow slowly change the sort logic over time. The only difference in many experiment cases are the HTML tag names used. If google truly believed all tags are equal then my results wouldn't be reproducible over time and across the two other similar experiments I have access to, which use different secret words and W3C HTML Tags.

The tags do not appear in alphabetical order and changes would not occur the way they do.

The order would change wildly and frequently if cross system hashcodes were in play. Hashcodes values do not remain consistent from server to server and runtime to runtime. I would expect randomness from hashcode order. This does not happen.

The order is not random.

The test is reproducible.

I can only conclude that how they determine relevance and how they define web spam ultimately yields a base weight effect of the semantic tags being used. It might not be intended or purposeful, but I think the order is an effect of their methods. I make no claims to intended or unintended by Google. I don't care about Google's intentions or claims. I care about their actuals.

I understand that my observations & opinions are inconsistent with claims made by Google and it's patents and that for reason of practicality there is only so far I can go in proving my case. I don't think Google or it's patents are being completely open and transparent about how they operate in actuality.

Given Google's claims, Google's patent's claims, and field observations. I am more inclined to consider field observations over the others. This is because in "my experiences with" and "my opinions of" Google it has a history and reputation of lies of omission and waging plausible deniability campaigns regarding their algorithms and rules. I would not put it past them to file decoy patents or make false statements regarding their methods.

Matt Cutts is Google's "Cancer Man" of X-files fame in my opinion. The official smoke and mirrors man who gets the cover up jobs to keep the world in the dark and to keep everyone guessing.

If it works for me and IF it turns out to work for others, I'm not going to let Google or academics tell me it doesn't work. That isn't scientific method either.

Tag order is one part of my system. My system works well enough that I actually project my SEO revenue growth into the future and continue to meet or exceed it. My SEO projections are actually a part of the company books and the results aren't random.

My system is built upon looking at my keyword space macroscopically and microscopically and building systems based on the observations and trends I'm seeing in the data.

My theories are often wrong, but my results are surprisingly stable, effective, and recurring. For whatever explanation there may ultimately be. Acting on empirical observations in SEO is a formula for success.

I know I am still not align with the Math and the claims. The only other way I can put it is that the math and the claims don't explain the observations.

Either the observations are bad or there is something happening at the semantic level of the source over at Google. I'm still inclined to believe the later as silly as it sounds because I have not encountered an even marginally plausible theory that produces these results that is to the contrary. The rub is that it isn't a theory if it can't be tested and there are a whole lot of practical barriers to what can be tested in terms of SEO insights.

So limited to experiments that can be performed. How can I test to the contrary. I'm not opposed. I just don't know how best to do that.

 

[freefox]

How do you compile all results, by "hand" ??!?

 

[anil2006]

MaTT said there are over 200 factors for ranking.
But value of those vary from time to time. I feel they do some changes daily. What more i sould say when one of my friend told they do minor changes at every 5 minutes... OMG

 

[Bugalugs]

A huge thank you to all the above contributors. Such a valuable thread with some actual meat in it. Real reading matter. More please!

Bugs

 

[boobooboy]

03 acronym

acronym ranking no 3 makes sense because these actually end up helping users that visit your website and Google is all about user experience.

Still trying to figure out how to use samp

BIG THANKS