• Register
• Log in

SEO Theory - SEO Theory and Analysis Blog

Algorithm analysis, Web community relationship analysis, SEO practices and techniques, industry news, etc.

• Home
• About SEO Theory
• SEO Glossary

What every good SEO should know about link analysis

Posted by Michael Martinez on September 24, 2007 in Intermediate SEO, Link Theory

The only valid test for determining whether a link passes value is to determine whether the link passes anchor text. There is more than one kind of value that can be associated with a link but most people are only interested in either internal PageRank or anchor text. You cannot in any way determine whether a link passes PageRank but if the link has unique anchor text you can at least search for that and see what comes up.

Of course, the search engine may not show you everything it knows. You may see the destination page, the linking page, both, or neither. And what you see may not accurately reflect the truth as stored in the search engine’s index. In other words, there is no accurate test for determining whether a link passes value.

If you place ten outbound links with unique anchor text on a page, wait until you are sure the page has been crawled and indexed (by checking its cache in search engine indexes), you may find that only five of those unique anchor text expressions appear in search results. Worse, you may find the linking page only shows up in 2 or 3 of those search results.

Does that mean the page doesn’t pass value through its links? Does it mean that some of the links have been penalized?

Don’t overcomplicate these questions by assuming there may be extenuating circumstances (such as Javascript or “rel=’nofollow’”). In these examples, there are no qualifying points. You just see that only some of your perfectly natural, normal links seem to be passing value. You cannot explain why that is. You should not even attempt to explain why.

In search engine optimization and link analysis, why doesn’t matter. What matters is that you document what you see and evaluate it. You have to test your linking observations carefully. You have to prevent yourself from making assumptions. That’s not easy.

Linking data may not be complete. It almost certainly is not complete. You can agonize over a link that doesn’t seem to pass value, give up in frustration, and come back six months later to find that the link now passes all your tests. Is the search engine playing tricks on you or did you just make too many assumptions and ruin your analysis?

A page may pass value through all of its links.

A page may pass value through all of its links except those links that point to pages that are not in the search index.

A page may pass value through all of its links except those links that point to pages that are not permitted to receive value.

A page only receives value from links that appear in the search engine’s index and that have the ability to pass value.

A page may not receive value from links that are in the index even though there is no reason for them not to pass value. That is, neither the linking page nor the destination page are being filtered or penalized.

Link value is vague and ambiguous, but if a search engine is constanty updating its index then it has to implement “break” points where it stops processing data and just moves on. It may pick up where it left off later. Each search engine handles its own processing priorities. A search engine may even concurrently use several different versions of its algorithm at different data centers.

You may not see consistent data in the search results, so your analysis has to allow for the inconsistencies. You cannot assume that everything is exactly as reported by the queries you run. For example, you may see query results where (according to the cache) the keywords only appear in links pointing to the page but no other pages are apparent. If the search engine is telling you that link anchor text is making a page relevant to a query but it won’t show you the linking page, what can you know about the linking page?

Nothing.

You’re in no position to be making assumptions and don’t even pretend you can make “educated guesses” because you don’t have any information with which to make those guesses. If you try to explain why other than to point to whatever the search engine tells you, you’re just making up nonsense.

Search engines don’t really lie. They don’t have to lie. All they have to do is withhold a few pertinent facts. The exclusion of data is sufficient to throw most search optimizers off the scent and the ones who do figure out something on the basis of ignorance are usually just doing it on pure dumb blind luck. You don’t have to be anyone famous or have had ten years’ experience to be lucky, but don’t confuse luck for skill or knowledge or a deeper understanding of the things you cannot see in the search results.

Link analysis needs to tell you a few things. Don’t try to make it tell you everything. Link analysis is not link research. You have to do the research first in order to do the analysis. You can try to retrofit research to your “best guess analysis” (some people confuse that SEO myth with “a hypothetical suppostion”) but if you go looking for research that backs up a conclusion you’ll always find it.

Retrofitting research to conclusions is easy: you just ignore everything else that doesn’t fit the pretty little picture you have concocted for yourself and your rah-rah group.

But link research is not what it may seem like. You don’t go looking for backlinks when you’re conducting link research. In fact, you don’t have to see the links to see their effects. Link research is in some ways like astrophysics. You can detect the links by examining search results where on-page factors are obviously not the only data affecting the results.

Link analysis has to tell you what a link is doing, may be doing, and cannot do. Link research has to tell you what is possible, what is not possible, and where you look next.

What effect can a link have?

It can pass anchor text. It can help validate a page (pass internal PageRank or something like that). It can give a search engine a reason to elevate one page over another. It can show a search engine another page that should be included in the search results.

But you also have to ask what effect a page has on the links it carries. Does the page help the links or hurt them? Does the page make the links more important or less important? You don’t need to tally up points on a score card to get a feel for which links are the most important links on a page. You don’t have to reverse engineer algorithms to see that one link is more prominent than another.

Search engine algorithms are trying to reverse engineer your own evaluative process. Search engines are trying to replicate through their algorithms the criteria your subconscious and conscious minds use to sort data. You have the more complicated, sophisticated algorithms in your head. You have to look at links as if they were points on a grid. Do you want those points to be connected to each other or not?

There are many ways to prevent pages from linking to each other. Time is only one of several natural barriers between pages. If a search engine crawls a page today but doesn’t crawl the pages it links to for a month, the relationship between those pages won’t appear in the search results for at least a month, perhaps longer.

So knowing what we know — that we’ll never be sure of what we are seeing — we have to reach for analytical tools that will help us make “best guesses” about what is happening with links. When you have insufficient data to work with at the granular level you must turn to aggregating data so that you can look for trends and patterns.

Trends and patterns won’t tell you why a search engine ranks a page more highly than another, but they may give you insight into why some pages are crawled more often, why some pages help boost others in search results, why some pages are better for certain topics, and why internal links can be at least as powerful if not more powerful than external links, and why internal links fail to achieve what is expected of them.

Trends and patterns cannot be gamed or spammed. They are the footprints that mobs and marching companies leave in the dirt behind them. Trends and patterns are the foundations upon which we build the science of SEO theory.

Remember the SEO Method: Experiment. Evaluate. Adjust.

But in link analysis you don’t experiment by creating content or links. You experiment by creating queries. You evaluate by comparing queries with on-page and off-page data. You adjust by creating new queries.

When you can create queries that produce results you expect, you have found at least one trend, one pattern. Now you’re analyzing links. Now you’re ready to take the next step.

Read other articles in this series:

1. Fundamental Principles for Link Analysis
2. What every good SEO should know about link analysis
3. Linking analysis: Who is linking to you and why
4. Link Analysis: Crunchy numbers and soft tallies