Google SEO Experiments
Although I share some experiment results with my clients, I am generally very reluctant about reveiling methods behind SEO experiments. However, I've seen the 'multiple links anchor text' experiment implemented in different way and published elsewhere, so I am presenting details of my implementation observation here.
UPDATE: as web continously changes, results of these fragile SEO experiments can become invalid as soon as one of the pages under testing is changed, removed from search engine index, or linked to from other pages either non-intentionally or intentionally with the aim to ruin the experiment. Indeed, such a thing happened with the observation below, so I had to update example search query with the new one (and I may have to do it again in the future). This is the reason why I conduct SEO experiments under very controlled conditions and on my domains that I never talk about, and on pages that are linked to only by me and in a way that it can not be traced to me. That is also the reason why I don't talk about nor show those experiments. The moment other would start exploring them, most of them would become invalid.
SEO Experiment: Keyword Stuffing Links Is... Useful?
This one is not so much an experiment implementation as it was an observation of what a google webmaster central's blogger did when publishing a Google's SEO Starter Guide. On the same page (http://googlewebmastercentral.blogspot.com/2008/11/googles-seo-starter-guide.html) in second and third paragraph there are 2 links with 2 different phrases that are pointing to the same PDF document (search-engine-optimization-starter-guide.pdf). Phrases are:
PDF document only comes up for the first phrase (which is also in the pdf itself), and not for second.
If you think it comes up for the first phrase because that phrase is in the document itself, then you can check for anchor phrase from another language page and that is not in document:
PDF comes up for above query because it is linked to from here with "SEO-Reader". "SEO-Reader" is not in the PDF itself.
I find above example interesting as it was done by someone from Google supposedly spreading the SEO knowledge. A casual reader seeing two links pointing to the same page may get an impression that placing many same links with different texts will improve SEO.
As you have seen above, second link is useless. Or is it? Although second link does not benefit the target page, it may be useful for usability purposes -- it has catchy text and user may click on it, which may have been the aim of the blogger.
Also, second part of above experiment is figuring out whether that anchor text 'with no effect for target page' benefits the page itself, (e.g. is it treated as plain text or bold <b>), but if I start revealing you everything, then you won't need to hire me!
Anyhow, now I showed you how a simple SEO test is done. Many SEO factors are much more difficult for testing than this one (like for example, the second part of experiment not shown here), and rarely can they be tested by simply observing other sites (as I did above), but experiments need to be thought of and implemented before they can be tested.
NB. Note that different anchor texts on links pointing to same page represent accessibility problem for a number of visually impaired users using special browsers.
ANOTHER UPDATE. Seems Google changed altorithm again. It may give priority to non-navigational anchor and/or also consider links to page sub-sections which have '#' in the URL.
Google PageRank Log Base Experiment
Google's PageRank toolbar uses logarithm scale with an unknown base to assign each page a number from 0 to 10. Various SEO experts have proposed this number to be anywhere from the value of the natural logarithm base e to number 10, including 3, 6, and 8. Knowing accurate value of this base number would be very helpful in Search Engine Optimization. This experiment is designed to find out what is the base of this logarithm. Enter at your own risk!
SERP Experiment Results
Results of few more SEO experiments aimed to determine with high certainty weights of some major factors in Search Engine page ranking algorithms may be published here. However, for now details of experiments will remain business secret and only general results may be presented here. These results may not be valid in future as Google algorithm changes. And yes, these results relate only to Google, although same experiments can be tested in all search engines.
- Complete HTML title tag match with the query phrase is better than having additional words in the title. This means if your page is about very general topic or national/international business, don't put in title any other information but the most common description of the topic.
- Proportion and word distance of matched keywords in the title is related to the closeness looked for in the text. If title words are not closely matched, it's better to have longer non-focused article with a variety of words in them, and title should be longer too.
- Only anchor text of the first link on the page is considered by Google -- details shown above.
Next: More experiments and results may be published here soon.
Previous: Learning SEO.