Belief and Evidence

One would hope that we believe something because there is evidence for it but we can also believe things for many other reasons including personal experience, logical reasoning or because someone else has told us and we regard them as an authority source.

With SEO, we do not know all the details of how the search engine ranks websites. Sometimes, information is available from Google itself in terms of its guideline for writing web sites. Search engines need to keep their algorithms secret otherwise they wouldn’t have a business. It is surprising that inside information does not become available by ex-employees of search engine companies or why ex-Google employees do not run their own SEO companies.

In the absence of such information, one way that we can test our ideas about what works is to use the scientific method.

The Scientific Method

The scientific method makes a reasonable attempt to find truth. It consists of several stages:

Observation – for example we notice that sites with lots of content can rank better than with very little.

Initial Guess – the hypothesis adding more content will increase my ranking.

Experiment – create two web-sites trying to control as many of the variables as possible. If the site with more content appears higher in the SERPs than the one without, then this adds weight to your idea. It does not make it true. Repeated results from other sites also add weight to the idea. If this works for long enough it could eventually become part the received  SEO theory. All of this assumes that you have done your best to control other factors which might also affect the result.

A fair experiment should ideally control all the variables except for the one we are testing. A truly objective experiment is not possible in SEO. Experiments usually take the form of comparing one site with another but in making the sites identical, we introduce duplicate content which affects the results.

Prediction – Using the results of our experiment we go on to make some further statement that we would expect to be true if the hypothesis was correct. In the true scientific method it is this that gives it its power.

Even though the scientific method can shed some light on the search engine algorithms there are reasons which limit its usefulness.

• One of the assumptions made about the physical world is called the cosmological principle, the laws of physics are  the same throughout the universe. The laws that apply in one place are thought to be applicable anywhere in the universe^*. The same cannot be said for Google. Its algorithm is constantly changing to insure that the results it provides are useful to its users. Ideas that once worked are now relegated to a graveyard of past SEO techniques such as: keyword stuffing and meta tags are now redundant.

• In SEO we rarely can perform the prediction stage and must rely  the reproducibility of results.

• Search engines contain many variables – the search engine ranking algorithm is a complex process which takes into account many factors some of which we can control and others we cannot. To quote Donald Rumsfeld:

“There are known knowns. These are things we know that we know. There are known unknowns. That is to say, there are things that we know we don’t know. But there are also unknown unknowns. There are things we don’t know we don’t know.“

The Known Knowns

These are the elements that we can control:

• Anchor text
• Template Design
• Internal/Outbound Links
• Number of Pages
• Link Velocity – Link Velocity is how quickly your site gains links, and how many appear each day. It’s best to try and achieve a upward slope, followed by a downward slope

We also have some degree of control on the

• When are Where the Links to our Various Sites Appear
• Content/Keyword Density
  
• Domain Names and File Names
• How Quickly Links are Discovered – this is out of our hands even if links are submitted at the same time. Google will pick them up in its own time. 

The Known-Unknowns

Elements of the algorithm that we know exist but do not know exactly how they work for example: the sandbox.

The Unknown-Unknowns

The unknown elements of the algorithm and changes that are made to existing elements of the algorithm which effect the result but we have yet to find out how and why.

Designing a Fair (as possible) SEO Experiment

Link Control

• The links you obtain must come from the same sites. You may have to sacrifice backlink quality to do this.
• Even things like Digg are not perfect for this, as the articles showing up on different “upcoming” pages can impact how much juice gets passed to our specific backlink page. However, by using sites like digg/propeller for the experiment, we gain a large benefit. They have SO many outbound links on any given page, they pass the same amount of juice to pretty much any page(not very much). By submitting using the same tags/category/user, this variable becomes much more controlled.
• If you’re only testing 1 or 2 sites, you can get away with directory submissions, but you lose a bit of reliability. Submitting 2 nearly identical sites to the same directory in the same day will almost assuredly get one link accepted, and one denied. So a 1-2 day gap is needed. This affects our domain age/backlink age in Google’s eyes, and throws off the results. It’s necessary in some cases though.

Content/Keyword Density

• Make a list of any and all acceptable uses of your key phrases, or anything topic-specific. Especially verbs. Note how many times you’ll use them per page, and overall content length.

Domain Names

• Testing multiple sites on one domain (via subdomains) is unacceptable.
• Use domains that will provide no benefit or penalty.
• Use gibberish letters on the same TLD when you’re testing a new variable as a keyword in the domain produce a small advantage which may skew the results. Avoid vowels to ensure accidentally containing a small word someone might search for.

Time for Links to be Found

• Submit the sites at the same time to the same places.
• If you’re submitting to social news sites, make sure you use the same tag, categories, and user names. First and foremost because different tags hold different PR(link juice). Secondly, because when the search engine crawls those categories/tags/user profile, it will see all the links at once.
• If you know of links that you can’t control in the method above, consider submitting them directly to Google, or pinging the URL if they have RSS.

Template Design

• Make sure that the links to the various locations are in the same areas. Google weights link locations differently. A footer link for example, in theory is not weighted as heavily as one in the header, or a link that (somehow) appears in the content part of every page on a site.
• Use the same link frequency internal linking can dramatically alter the PageRank of pages within a site.
• Each site has the same basic structure, with the same number of internal links to each page, from the same locations. With the same anchor text.
• Make the design as simple as possible for menus and navigation. Static HTML is the friend of experimentation.
• Use the same document layout for each site with headings and sub-headings.

Number of Pages

• Unless you are experimenting on the number of pages, use the same number of page in each site.
• One important note: Just because you create 3 pages of content does not mean that there IS 3 pages of content. If you’re using Wordpress, there many more.  Each tag and each category and each entry in the archives creates a new page that could affect your results.

_{* except in extreme circumstances where physics breaks down such as at the singularity of a black-hole, < 10^-43 of a second before the big bang.}