The Results Are In! New Test Results!!!

Hello:

Thanks for all the great comments and suggestions from the last test I performed. Some new test results are in, and they will refine the way you look at SEO and how you create your websites from this point on. Remember, the test results are very new, and we must keep checking back every week to see the new results, if they change, or if they don't. Here's the link, make sure to link it or bookmark it for future reference:

Google SEO Test Results - EDSolutions
http://www.edsolutions.ca/seo-google-test-results.html

Hope you enjoy the results as much as we enjoyed creating the test pad. Please comment and give feedback about this test and for future tests!

Hexed

 

I like it but I am easy..LOL but very interesting as I see it

 

A nice set of indicators.
I was rather surprised to see the h1 tag fall below the h2 (and the simple bolding of keywords).

It's a handy reference, although one has to be careful what they take away from this. You've chosen (for obvious reasons) to isolate various combinations, which works quite well.

I'd be interested to see however what happens in relation to the h1/h2 tags when there's more realistic keyword densities. (ie, when in real life, headings become necessary.)

As a baseline though - I like this test. Well done !!

 

Any thoughts about setting up tests for:

The number of backlinks relative to PR
The number of pages in the web relative to PR
The PR relative to SERP

Tweaking a web is all about moving up in the search engine results but it seems we’re still doing it with blinders on. Does anyone know of definitive data that ranks the importance of these variables?

 

Absolutley fantastic. I would like to see the results of the keywords being linked, either to the number 1 page or to itself. Bookmarked for sure.

 

Great! I like those results. And even more the scientifically-minded test strategy and setup. Finally results which can be easily overlooked and checked. Will be interesting whether G! reacts at some time or other - since we must take into account that this is a very systemic world...

Thanks and regards to hexed!
Leo
Coach, Coaching

 

Well I hope you don't think that I'm raining on your parade. I admire and welcome anyone who attempts to run controlled tests to try and divine what Google is doing. To this end that is exactly what we are doing with the McDar 'sleeping bags' test which is the longest thread in this forum.

Now to your test. The problem I see with this test is that it teaches nothing about the problems of geting a mildly competitive to fully commercial site placed in Googles SERPs.

1. You are using a totally unique phrase and search term. It is a very rare niche site that would ever fall into that category. So I would suggest that the result are of little practical value.

2. It has always been acknowledged that you can get a site like that ranked on "content" only. And that is what you have done.

3. Your results would seem to suggest that Google is using meta tags. I can show you multiple other tests that suggest exactly the opposite. Did you have a version of the page that used you keywords in the title and URL but not in the meta tags?

4. Did you put up any backlinks to any of these pages? It would be interesting to see what 10 PR4 backlink with no other on-page items would do to the SERP placement.

5. I don't think you can draw any conclusions from the placement of the <h> tags, bold and italics. Google probably gives them all about equal weight and the placement in the SERPs is coincidental in my view.

6. If you are going to maintain that the placement in point 5. is significant then you must also say that Google penalizes site with meta tags, because the copy of the page with nothing but content is positioned higher than the site with both meta tags, or the site with only the meta keyword tag.

I'm sorry but I think this test is almost meaningless. These pages have no competition. If you are going to do a meaningful test I think it has to be in a situation where Google has an option of placing some other pages in their SERPs. When, by the nature of the test, you only allow Google to return one small subset of virtual identical pages I don't think you have established anything.

As I say I'm sorry to be so critical of these results, but I'm concerned by all the people who have said how meaningful these results are. I'm concerned that they may be taking away totally false impressions.

 

Your comments are completely welcome and well accepted.

The test was not meant to get a mildly competitive or fully commerical site into the SERPs. The test was mainly to show the differences between different strategies that can be used in SEO and how they contrast directly to one another.

Also, the results do not suggest that meta tags are used by Google. If you read it correctly, it was stated that the meta tags were useless (hence the two pages rated the lowest). I understand that also the two highest rated pages have meta tags, but it is believed that they have no bearing on the pages. Since guessing is not good enough, we have already set up a new test to prove this theory.

There are no backlinks in this project other than the internal links used to index the project pages.

And I am sorry, I feel that this test is totally meaningful. It displays what we set out to show: the bearing each strategy has on each page that it is used in. This is an isolated and controlled test, not a full scale competitive.

To continue, I'm not sure how you can even perform tests on keywords that already have 100+ pages indexed all with different PRs, keyword densities, strategies, etc. There are just too many factors that you cannot control.

Personally, I believe that a small controlled test is much more "real" that you make it out to be. However, there always needs to be someone on the far left to even out the world. Without that, we wouldn't have much balance, now would we?

Overall, I agree with you about being mildly skeptical about preliminary results. We must wait for the new tests to come in to start really solidifying our believes and results.

Hexed

 

Your results are interesting. The real key is repeatability. If you can produce a similar stack-ranking of results over the course of several similarly controlled tests with different "unique keywords", you can then claim predictability. One or two test like this does not necessarily make it valid or predictable. Proving predictability through repeatable results gives the test validity.

There may be something to your <h2> hypothesis. That's the part that needs to be proven. I'm very interested and looking forward to seeing further test results. Good job.

 

A lot of interesting information here, what will be your next test?

 

I have always been of the opinion that the number in the <H> tags was almost inconsequential. The numbers have a cosmetic effect. <H1> is bigger than <H2> etc. So if I select <H2> rather than <H1> it may not indicate I think the text between the tags is any less important. It may be a design consideration.

Now I know that in fact you can specify and control the actual size and appearance of an <h> tag. But this is just one more reason for Google to treat them all the same. Otherwise they would just be encouraging <h> tag spamming.

So I don't think <H2> is valued more highly than <H1>. I think they probably are valued the same and their placement in the SERP was random.

 

Yes !

Repeat the test. In fact, take it a step further. Deploy several such tests (simultaneously) with different (but conjoined) keywords.

As a controlled experiment, I think it has legs. Repeat the test, and you'll have real grounds on which to make observations.

As compar has noted though, it's important to be careful what's deduced from this, but it's a solid approach in my book. The more controlled it is, the easier it is to find consistencies.

I'm looking forward to your next test, and think it's great you're sharing these results with the community.

 

Are you planning to extend the analysis to Yahoo & MSN?

Also I think creating a second set of pages with different keywords would be a great way to add weight to the results (or disprove them).

Keep up the good work.

 

The general feeling in SEO forums about the new MSN is that you just need as many links as possible, including sitewides

 

I find threads like this quite interesting. It just seems though, that there are simply TOO many variables to account for, to ever know for sure. You start with one serious unknown (google algo), plus you have a lot of other unknowns as well - including what other websites are doing for advertising, links, page design, etc to determine why you would rank differently.

Maybe I'm looking at this too scientifically. (draw back from being in grad school), but a test like this doesn't seem to have much external validity.

Still could be useful information though

 

Another praise. Thanks much for the help as you're doing quite a favor to the community.

 

What is he doing for the community? Show me one thing that we know for sure that would apply to a normal web site as a result of this "experiment".

 

Come on, Bob, if you are trying to see what google looks at when it indexes a page this is very interesting and informative. If you are trying to track changes in the algorithm, this is very interesting.

If you are trying to see how different strategies work in the competitive environment it might not be complete, but without isolating the test from what everybody else might be doing, all you are doing is shooting in the dark

 

I have to agree with Bob here. Until you have ran the test several times with different themes and keywords there is no real conclusiveness in the results. I will be interested to see if your future tests consistantly back up your findings so far however. I think it's great that you are sharing the results though. There's a vast amount of people who run experiments like this and keep the results for their own fortunes.

 

It isn't like he is claiming to be an expert in the field or charging to get the information, so there was no need to jump on his back about offering the information.