SEO News
Google

Internal Linking to Promote Keyword Clusters

davies-dave
by , Comments

internal-link-clustersIn “Keyword Clustering for Maximum Search Profitability” we discussed the idea of clustering keywords and how doing so can increase the speed and effectiveness of your link building efforts. The article was based on the principle of passing internal strength between the pages of your website.

Today we'll discuss this subject in more detail. Be warned, it'll involve a little bit of math (we are dealing with Google's algorithm after all, so there's almost always going to an element of that). That said, I'm going to do the math for you to provide what's most important, an understanding of why the formulas work and simple ways to determine what needs to be done. If you understand the why, the math essentially does itself.

What Are Internal Links?

We all know that internal links are the links within your website that enable visitors to get from one page to another, a point we won't dwell on further. The question you want to answer here is, “What do internal links mean to a search engine and how are they weighted?”

At its core, an internal link adds value to your pages in a manner similar to third party links to your site. That said, this would be a poor attempt at education if I assumed knowledge so let's take it from the top and answer first the question… how does strength pass from one page to another? (Note: many of the principles of this apply to both internal and external links to a page.) From there we'll look at how external links impact the weight flow.

When I ponder the value of a link, either internal or from a third party, I consider the world of Orwell's "Animal Farm". The first and foremost thought is my head is that each page has a vote – a chance to cast their ballot in favor of other resources. Where it gets Orwellian is in his infamous quote which I will bastardize for my use here, “All votes are equal, but some are more equal than others.”

To put this in the context of internal links, a link from the homepage of a site or another strong page will be weighted higher than a link from a weak page 12 levels deep in its hierarchy. With this sentiment we know that the old real estate adage of “location, location, location” holds as true in SEO as it does in the “real world” however it gets even more true when we consider the other elements that come into play.

A picture is worth a thousand words, so below you'll find an image of a simple site hierarchy. Due to my complete lack of design ability, hopefully this picture is worth at least 75 words or at least doesn't draw from the overall word count of this article. It will serve the purpose needed here at least.

A Simple Website Structure

Below is a seven page website (six internals and a homepage). Now let's consider how the weight will pass from one page to another. In my calculations I am not factoring in the evaporation of weight that occurs with every link (Matt Cutts discussed this at second 40 in his video here).

Because this happens with every link on your site and your competitors' as well, it can be viewed as a level playing field and negated for simplicity though it is reinforcement for limiting the number of links on a page to minimize evaporation. But back to the point at hand.

No matter what the site is, one can assume the homepage value is 100. This is because I'm only factoring in the link passing within the site, not valuing the site against others. So let's begin.

internal-link-example-1

If the homepage value is 100, the value passes as:

  • Homepage – 100
  • One – 33.3
  • Two – 33.3
  • Three – 33.3
  • Four – 16.7
  • Five – 16.7
  • Six – 33.3

This assumes that each of these pages links only to the pages in the diagram and the weight is split evenly among all links. In the real world, however, it would be more realistic (though messy in the illustration) to assume that each page links to each page higher in the hierarchy plus the homepage. So let's look at what each page will pass.

The homepage starting value is 100 meaning that it will indeed pass 33.3 value to each of the pages one level down.

Rather than linking downward however these pages will also link back to the home page giving the following values:

The Home page passes:

  • 33.3 to Page One
  • 33.3 to Page Two
  • 33.3 to page Three

Page One passes:

  • 11.1 to Home
  • 11.1 to Page Four
  • 11.1 to Page Five

Page Two passes:

  • 33.3 to Home

Page Three passes:

  • 16.7 to Home
  • 16.7 to Page Six

Page Four passes:

  • 5.6 to Home
  • 5.6 to Page One

Page Five passes:

  • 5.6 to Home
  • 5.6 to Page One

Page Six passes:

  • 8.4 to Home
  • 8.4 to Page Three

So at the end we end up with the following values:

  • Home – 180.7
  • Page One – 44.5
  • Page Two – 33.3
  • Page Three – 41.7
  • Page Four – 11.1
  • Page Five – 11.1
  • Page Six – 16.7

So clearly we see a situation where the second level in a hierarchy gains value by having a larger number of pages as sub-sections of it.

Now let's look at a more realistic (albeit advanced) example and consider the weight passing if each page links to those within its cluster as well as all the pages above it.

For example, Page Four would link to pages One and Five as they're within its cluster and also the homepage and pages Two and Three. This mimics an environment where pages One, Two, and Three are in the main navigation along with the homepage.

Further to that, we'll consider the inclusion of breadcrumb navigation adding additional links to pages within the direct hierarchy and thus part of the cluster. As discussed in the video above (and witnessed many times in practical application) the addition of these links passes more “juice” to those pages (the simple math is, two links to a page gives it twice the weight – we'll go with that for our purposes here).

So let's look at how that weight passes this time:

The Home page passes:

  • 25 to itself
  • 25 to Page One
  • 25 to Page Two
  • 25 to Page Three

Page One passes:

  • 7.1 to Home
  • 3.6 to itself
  • 3.6 to Page Two
  • 3.6 to Page Three
  • 3.6 to Page Four
  • 3.6 to Page Five

Page Two passes:

  • 10 to Home
  • 5 to Page One
  • 5 to itself
  • 5 to Page Three

Page Three passes:

  • 8.3 to Home
  • 4.2 to Page One
  • 4.2 to Page Two
  • 4.2 to itself
  • 4.2 to Page Six

Page Four passes:

  • 0.9 to Home
  • 0.9 to Page One
  • 0.5 to Page Two
  • 0.5 to Page Three
  • 0.5 to itself
  • 0.5 to Page Five

Page Five passes:

  • 0.9 to Home
  • 0.9 to Page One
  • 0.5 to Page Two
  • 0.5 to Page Three
  • 0.5 to Page Four
  • 0.5 to itself

Page Six passes:

  • 1.2 to Home
  • 0.6 to Page One
  • 0.6 to Page Two
  • 1.2 to Page Three
  • 0.6 to itself

So at the end we end up with the following values:

  • Home – 153.4
  • Page One – 40.2
  • Page Two – 39.4
  • Page Three – 40.0
  • Page Four – 4.6
  • Page Five – 4.6
  • Page Six – 4.8

So we end up with a curious situation in the math. What one may conclude is that it's better to keep a limited number of sub-pages in your site, after all – Page Six is carrying more weight than Page Four so it must be a better structure.

The only takeaway I hope you draw at this stage is that it's a good idea to take out expired products or useless pages (for your users as much as the engines). To illustrate why, I'm going to shift our site into the real world and imagine once again that we're selling bike parts (see last article for reference here).

Let's imagine the following page definitions:

  • Homepage – My bike store
  • Page One – Suspension forks page
  • Page Two – Privacy Policy
  • Page Three – Dual-suspension frames page
  • Page Four - Marzocchi 44 Rlo page
  • Page Five - Marzocchi 44 TST2 page
  • Page Six - Banshee spitfire page

To see the impact on ROI, let's imagine a scenario where I build a link to both pages Four and Six. The link built will give the same weight (let's give it the arbitrary value of 5 by our weight model above) and let's now calculate what happens.

I'm going to omit the math and simply list the final numbers. We'll assume that the starting value are those defined above so the weight and re-factoring on the engine's part will produce higher numbers (as the values from the pages low in the hierarchy add weight to the pages above it).

  • Homepage – 248.8
  • Page One – 131
  • Page Two – 125.6
  • Page Three – 129.4
  • Page Four – 25.4
  • Page Five – 20.4
  • Page Six – 26.2

So here we see that the weight given to Page Six is higher than either of the two other third-tier pages however there are two important points we need to consider before asserting that a hierarchy that puts a single path to each product is superior.

Had we divided the paths off into four from the homepage to facilitate each of the product pages the split in weight from the homepage would have yielded the following example.

internal-link-example-2

If this is the case we would have found the following to be the final values (assuming the link additions to pages four and six as above and the same breadcrumb navigation):

  • Homepage – 228.5
  • Extra Page – 87.1
  • Page One – 85.1
  • Page Two – 80.6
  • Page Three – 87.1
  • Page Four – 10.8
  • Page Five – 9.8
  • Page Six – 10.8

The weakening of the links off the homepage weakened the entire site reducing all the potential rankings on the internal pages.

While the link to page six in the first example yielded a higher page weight on that specific page when it was in a single path on the third tier of the site, the benefit was limited to that page alone. When we linked into a cluster, the benefit on the individual page was reduced slightly as the weight shifted with the increased number of internal links however the weight of a number of pages improved (including the product category page).

By clustering your targeted keywords together you'll be building links to groupings of pages that will, by design, help the rankings and page weight of each other.

A Beneficial Issue With The Calculations

It's only fair to note when there are known issues and unknowns with data. In the above calculations where I added in weight from third party links I reduced the weight of those links along with the internal weight.

The treatment of weight from external sources by the major search engines is undoubtedly different than internal weight and it's likely that the target page of the link would hold the full weight or a larger portion of it and then pass the weighting along without diminishing from itself.

What I mean by this is that the link to page four in the initial example held a weight of 5 and was divided in our math by 8 (the total number of links on the page). It's far more likely that the page would keep all or something near all 5 and then proceed to pass on weight internally without diminishing its own or diminishing its own only slightly.

Essentially what this means is that building a logical clustered hierarchy is, if anything, even more effective than outlined in the examples above.

Final Word

The data matches closely to what the search engines are trying to push webmasters toward: provide a logical and well-coded site architecture that serves your visitors well and you'll be rewarded. Imagine if Amazon tried to apply the example from the second graphic above and provide a link to each product page on their homepage. Usability would be horrible, page load speed would be a disaster, and because math works well, their rankings would be non-existent.

I don't expect all the readers of this article to draw out diagrams and do the math behind this for their own sites, it's time-consuming enough with a seven page example – let alone a 1,000-plus page site. However, as the numbers expand, the math stays the same and the benefits only amplify. And that's why clustering keywords as discussed in last month's article works.


The Original Search Marketing Event is Back!
SES AtlantaSES Denver (Oct 16) offers an intense day of learning all the critical aspects of search engine optimization (SEO) and paid search advertising (PPC). The mission of SES remains the same as it did from the start - to help you master being found on search engines. Early Bird rates available through Sept 12. Register today!

Recommend this story

comments powered by Disqus