47 replies to this topic

[jeffroyit]

triggered by a post I found on some SEO forum, I decided to make an
experiment to test if really H1 tags score less than B tags on
google:

I made 2 pages with same content but with different tags for the target keyword:

Now if you search "khiazna" on google you can see the page containing h1 ranks 2nd below the one containing b. am I doing something wrong, or I have to start changing H1s into Bs?

Edited by searchrank, 22 February 2005 - 07:55 PM.

[skip_]

I can attest to that. I originally had a page with bold text, which I later changed to H1 (as part of optimization effort ) The page actually dropped in ranking... until I changed it back to the bold.

Now, this is a pure speculation, but here is what I thought:

Header tags are originally meant to define hierarchy on the page (e.g. H1 being the title, h2 defining categories, h3 defining sub-categories etc.). H1 on its own doesn't really define hierarchy and kind of looses its purpose, in which case I can see why Google would be putting more importance to bold text (since it is bold text that's used for making certain parts of text stand out). I hope everyone follows my thought...

If you want to extend your experiment to test this hypothesis, try adding some structure to your first page with H1, H2, and H3 tags, while repeating the same layout on the second page using bold text instead for headings.

...and please don't forget to share your findings!

[Randy]

A slight caution...

Be very careful when testing that you're not mixing up cause and effect.

Testing any theory is very difficult since you're not doing it in anything remotely resembling a vaccuum. And in order for any test results to be valid they have to be repeatable. So you need to weed out as many other factors as humanly possible.

One thing that I noticed right off the bat is that Google lists the www version for the page ranking first and a non-www version for the one in second place. Knowing how much effect that can have with other normal pages it would make me immediately suspect the validity of the results.

Of course that's without really looking at your test closely, so I certainly won't dispute your results. Just a reminder to make sure your tests are treating each page equally and to get it as close to a double-blind as possible.

[skip_]

good point Randy - I didn't notice the "www" part. Naturally, in order to make any conclusive judgements the two pages must be identical with the exception of the test subject (hx/bold).

[Renagade Master]

How about testing:
1) <h1><b> Test </b></h1>
2) <b><h1> Test </h1></b>

[Jill]

What would that tell ya, RM?

[DanThies]

That you have a lot of spare time...

[jeffroyit]

about test conditions:

- conditions are all the same: pages are similar, only tags differ.
- pages are linked from only 1 other page (and now this forum also)

www part is very strange because normally both should have the link with www also. when I started linkimg main domain that holds the pages, I used without www then started with www so probably google indexed both but from some unknown reason only one has www version indexed.

I added 2 more pages to the test for big and for h2, let's see what happens.

anyway, it is clear for me h1 is underranked against b, I will start changing

[Jill]

What about the fact that you have duplicate content (if both pages are the same)? That could completely skew your results, I would think.

[Clintorius]

When you do scientific test you must control all parameters.

That means you must create your own universe with several domains not linked to the outer world. The content must be unique for each page (G. only show one version of duplicate content). So you must generate some random content with only the keywords and their locations common to the test pages.

If you work with ordinary live pages you must provide documentation that states all other sites has not gone through any changes or optimizations that might influence your own ranking.

In addition you must avoid testing in periods where Google optimize their algos.

Life is not easy.

[Renagade Master]

What would that tell ya, RM?

It was a tounge in cheek post really.... a comment on the fine tuning of their optimisation. It made sense to me

But if your going to test the difference between <b> and <h1> why not add these test cases to, maybe after seeing the results the question will make sense.

Either way I doubt I'll change my pages to which ever method proves to be best.

[mal4mac]

Good to see some science going on, nice start jeffro. Hope you continue these experiment and refine them after considering the comments in this thread. When you get a result we all agree is significant then I'd be glad to repeat the experiment, as I hope others will. You don't get nuclear fusion until all the labs get nuclear fusion :-)

Some pointers on your experiment that challenge your hypothesis:

1) Maybe Google marks down pages that only have khazana in the heading and not in the content.

2) Maybe low/identical content will produce random/spurious/bad results (most SEO advisors recommend you have more than 150 words and that pages should differ significantly).

3) I suggest using nonsense words and incomplete sentence is a bad idea, Google may be into semantic checking.

[SanDiegoMedia]

Why not use both header and bold tags on the page??

[mcanerin]

Seems to me that a good way to test this would be to use a licensed google search appliance and run all the controlled tests you want.

Ian

[Rythm]

Now if you search "khiazna" on google you can see the page containing h1 ranks 2nd below the one containing b. am I doing something wrong, or I have to start changing H1s into Bs?

H1s are always more important than Bs.......tried and tested!!