SEO

Link Building: Hyperlink Cliques and Clusters

Cliqueishness is part of normal human behavior – an extension of clannishness. Though when it comes to websites, you don’t expect that sort of thing, except maybe in a blog network. Why not? Well, isn’t the Internet about promoting global views? Being part of a clique of websites linking to each other could harm your SERPs rankings. (I’m not insisting this is true, but it could be – which is what I’m examining here.)

How do the SEs (Search Engines) know you’re part of a clique? Hypothesis: probably using graph theoretic methods, which are a sub-discipline of mathematics. Look at the “Hyperlink Cliques” graph diagram. Each node in the graph represents a website (or weblog). Each line segment represents a hyperlink, in the direction of the arrow, from any page on the source site to any page on the target site. A double arrow means there’s a reciprocal link, but not necessarily page to page. The terminology here, cliques and clusters, is mine.

link cliques Link Building: Hyperlink Cliques and Clusters

Can you spot the link cliques? Why are the five sites A-E in the bigger circle part of a link clique? Why are the other two clusters not cliques?

Purple Cluster

Look at the purple cluster. Every site there has a link to every other site. This might signify a blog network, but only if the topics are sufficiently different. If they cover the same general topic, they’re a topical clique. This could be perceived as link collusion between the bloggers and could affect your rankings. (Suggestion: vary your blogroll.)

Blue Cluster

Is this a clique? No. Site F is the authority domain (relatively speaking) because all of the other sites, G-J, are pointing to it. But there isn’t a hyperlink between each and every pair of sites. There is, however, a short “path” between each pair. That is, through 1-4 clicks on hyperlinks, you could travel from any one site (F-J) to another in that cluster. For example, to get from F to I, you’d travel F-H-I, in that order.

Peach Cluster

Mmmm. Peach cluster. Ahhhhhh, tasty. Ahem. This cluster of sites isn’t a clique. In fact, it’s not even a cluster since you cannot travel from, for example, node M to N without going through K then I (which is part of Blue cluster). Thus, the linking between Peach sites is much more natural (excluding the social behavior of similar blogrolls) and likely “tastier” to spiders. Of course, this is only an example, and cliques do form naturally over time. So just because you’re a part of a topical clique isn’t necessarily a bad thing if it came about organically, over time. Reciprocal links just shouldn’t come too quickly.

Now, there are other clusters in the graph. Can you see them? Growing niches start with small cliques that expand, drawing in satellite cliques and clusters into their connectivity graph, but which are not part of the primary clique. This is normal linking behavior.

This is just an overview. I’ll be exploring the graph theory aspect of link building again in the near future, including having a look at different topographical types of blog networks.

You Might Also Like

Comments are closed.

13 thoughts on “Link Building: Hyperlink Cliques and Clusters

  1. My sites ranking just keeps going up. Pretty much all my links are from my Clique. I guess the Clique thing could happen if your click was very closed. But if your Clique has several sites that have a large amount of authority then I don’t see it happening.

  2. @ogletree: Actually, I’m only suggesting the possibility of a problem if all sites in a clique have the same main topic (topical clique).

    Aside from that, i think it’s okay. I had a network clique in the past and the sites gained rank superfast. And then again, algorithms change :)

  3. Fascinating theory. However, programming the search engine algorithms may create false positives when a new website comes along with only a small sample of links to it, and from it. I think the best conclusion we can come to is that too much interlinking within your own networks is bad, the search engines will know.

    The purple cluster provides a perfect symmetrically pattern example of what not to do when link building.

  4. @MMi: Yes, agreed about purple cluster, especially topical cliques, which are true cliques. But network cliques seem to be alright in some situations. Although it may depend on whether or not you suddenly publish 10 brand new websites and all of them are “fully-connected” (i.e., a clique). They have no authority yet, so cliquishness in this situation is probably a bad idea.

  5. If I read AIRWeb 2007 call for papers right is this what you call a clique called “nepotistic linking” in the research community to that topic (the other side, called Search Engines hehe).

    I understood clique right away, but for the word “nepotistic” did I have to use a dictionary to find out what it means ;)

    aehm, yes and search engines don’t like it, whatever you call it.

  6. @Carsten: Yeah, nepotistic sounds about right (though that’s traditionally reserved for when there’s an actual family relationship involved, no?)

  7. Yes, but it is used to describe a certain behavior pattern that members of a family usually show (helping each other out) and which turns into something bad, if applied to business (in the majority of cases) .

    “It’s all family, Tony.” ;)

  8. Interestingly enough, a “clique” in graph theory is a fully connected subset of a graph. You were spot on with your terminology.