Sphinn Home » Google SEO
Avatar Went Hot: December 27, 2007 - 8:59 am
Posted By: SlightlyShadySEO 237 days ago
Topic Type: News Story (Jump to http://www.slightlyshadyseo.com)
Category: Google SEO
In todays world, search engines maintain a high level of secrecy. Having such a lack of control also makes it incredibly hard to design a proper experiment. However, we have more control than we give ourselves credit for, and a properly designed experiment can easily identify what is giving you your edge, or what is taking it away.
7 Comments     

Comments

from sza 236 days ago #
Votes: 1 | Vote:
+ -

SlightlyShadySEO, I'd never have the patience to run such experiments, but here's another idea (partly based on your suggestions).

Important variables (that would skew results) can be filtered out from the experiment if you use a non-existing word to rank for. When you target the query "splmnthjwdng", the competition obviously won't really interfere with your findings.

It's even better to create the whole page from such gibberish so meaningful copy around "splmnthjwdng" don't influence rankings in some semantic way, either.

By creating several test pages (on separate random-consonants domains) you could run a controlled ranking competition between those different test pages for the same non-existant word.

The only problem is: you won't be able to submit such meaningless pages to directories, social news sites etc.

Therefore you need to create a cover:
  1. a legit-looking, well-designed home page where the (meaningful) copy is in images, so it's practically empty to a bot
  2. a (non-obtrusive but not hidden) link to the sub-page that actually takes part in the experiment
All your test sites should be created in exactly the same pattern (same structure, same link location etc. as you've detailed in your post) and you'll have your race horses ready to go, none of which has an unintended advantage over the others.

And the results of the experiment can continuously be monitored by checking the SERP for your chosen "splmnthjwdng".

from emanuelh 236 days ago #
Votes: 0 | Vote:
+ -

What you suggest is apparently the only way for some experimentation. However, I think that you should add to the discussion your view of the problem of not being able to know directly the actual relevance scores of your several experimental pages, so you'll have to experiment with only one parameter at a time and infer relevance score change from observing change in the rankings.

Moreover, suppose you want to gain some insight into linking and at some point in time during the experiment a page that links to one of your test pages gets, for instance, an unexpected relatively high TR link. In the very low relevance score arena you operate it will push that page to the top, but you may attribute it to something you did within the experiment variables.  So you also have to isolate all pages that contribute links to the test pages by metatags and text content of pure gibberish too or by empty metatags and text content.

from SlightlyShadySEO 236 days ago #
Votes: 0 | Vote:
+ -

The whole way you formulate the content deals with relevance scores, trying to keep them static acrossed a series of sites. As for your second point, it's quite legitimate, although your second point is why only one variable can be tested at a time...unless I don't follow you properly(which is entirely possible)

from emanuelh 236 days ago #
Votes: 1 | Vote:
+ -

SlightlyShadySEO - the point is that since you cannot measure relevance scores directly you don't know whether variable A lowering the score of page 1 (formerly at # 1 and now # 2) has caused its fall in ranking or whether it's variable B increasing the score of page 2 (formerly at # 2 and now # 1). So you can't test two variables at once.

from theGypsy 236 days ago #
Votes: 0 | Vote:
+ -

Well, I shall give ya a Sphinn for the effort Shades, but man, get in touch next time you do testing. The list looks like SEO 2006 for me. There are so many other variables that I would be adding to the mix given the nature of modern search.

Being a patent junky and technical search junkie, much has changed that we could possibly be testing for. I certainly don't have the time to do so alone, so get in touch and I can throw some other factors to fill out what U started there.... (great start BTW). I am always up for seom fun (time permitting)

Dave

from SlightlyShadySEO 236 days ago #
Votes: 0 | Vote:
+ -

Thanks Dave. I know the list seems a bit simplistic in some ways. However, many of the variables that we worry about in today's SEO world are little more than conjecture backed up by uncontrolled experiments. These basics allow you to actually TEST those, their affects, and their ability to influence. The more solid truths you find, the further in depth you can test, and the more reliable your results become.
But yes, I'd love to chat. Always good to have a new mind at work :-) I'll drop a message with some contact information in the contact form of your blog.

from hugoguzman 236 days ago #
Votes: 0 | Vote:
+ -

Hey Shady. Another great post. Sign me up as well. I would love to get involved in testing stuff out. I did a rather basic test recently on the effect of allinanchor on rankings (in Google) but I've yet to get confirmation from Matt Cutts (or anyone else) on whether or not that operator still works.

In any case, I'm always up for finding Google's rules on my own or in groups ; )


Log in to comment or register here.

Top 10 Most Sphunn
In What's New

  1. How Do You Measure Your SEO Campaig... (21)
  2. 43 Killer Viral Tips From Top Marke... (20)
  3. Yahoo Grouping Search Results (19)
  4. Domain Name Gaming: Tricking Engine... (16)
  5. Top 5 Legitimate SEO Techniques to ... (15)
  6. Blumenthal & Mihm Jump Over Buildin... (14)
  7. 5 FireFox Extensions to Enhance Lin... (14)
  8. 24 Ways for SEO Detractors to Prove... (14)
  9. How many businesses really ‘get’ th... (13)
  10. Examining Backlink Profiles in Doma... (13)

Latest Comments

Search Marketing Expo

Save the date for:
• SMX China (Nanjing) - Sept. 23-24
• SMX Stockholm - Sept. 23-24: See who's speaking or register now.
• SMX East (New York City) - Oct. 6-8: See the agenda or register today and save!
• SMX London - Nov. 4-5: Pre-agenda rate now available. Click here.