Great SEO isn’t a practice of outwardly predicting the current state of a non-determinable algorithm – it’s the practice of imagining a perfect search engine based within the restraints of realism (meaning: probability of being achieved), and aiming there. This post isn’t about what Google is doing now or what they have done, it’s a hypothesis about what a perfect search engine should do, and based within those constraints, what that means for us, SEOs. This isn’t a Bill Slawski introspection into a patent that’s passed – it’s a Ross Hudgens introspection into a search engine characteristic that should, could, and might even be – just as my post How Google Might Turn Down The Knob on Exact Match Domains does the same.
Link Growth and Mathematical Certainties
In this imagining, we have two types of link profiles – one that is organically driven, and one that is driven by black hat tactics – best described as “manipulative acquisition” in all forms. Manipulative SEO link profiles, besides many other characteristics that Google could determine instantly (page-level characteristics), also frequently have long-term failiabilities (domain-level characteristics) that are determinable on a long enough timeline. That is, manipulative links are normally obtained in one of two states – a) at the apex of a website’s domain value or b) when the value of that domain has already began to recede. This occurs because black hat links signal are acquired when a webmaster “gives up”, or is first approached to sell a link. When he sells a link, this eventually attracts the attention of competitors – they sell more links, and webmaster, whether aware or not, begins to lose his readers/linkers trust, and less links come in. As such, this generally signals the end of link growth – while it’s possible many of these domains will maintain their value, it’s unlikely that they will ever continue an ascent into high value stratas.
This is of course not an absolute statement – sometimes manipluative links are obtained at the bottom of a website’s link growth or mid-cycle – but as a generality, it can be accurately assumed that links of the manipulative variety most often come at the apex of a domain’s value – or when it has already begun declining.
If we look at the converse situation, an organic, non-manipulative link profile, the effect is quite the opposite. As compared to a manipulative profile, external domain value post link placement is generally steady and upward sloped, because the placement of these links never impairs on user experience – it is organically placed, uses anchor text less obtrusively, and includes links that are relevant and informational to their reader base.
If we look at the two link profiles, there is a clear dichotomy that can be expressed mathematically – and as such, also picked up algorithmically.
For the sake of clarity (and also to not provide inaccuracies), it is a near certainty that the domain value growth of link hosts are likely to decline across both profiles over time. SEOMoz reported an extremely high churn rate of the web (after just a single month, 25% of the web’s URLs were “unverifable”, according to SEOMoz), and it is that basis that makes it clear that both profiles value of hosted links on external domains will see a naturalized decline over time, despite their manipulative or non-manipulative intentions. However, the difference in value reduction will be sharper for the black hat profile as it will for the white – because of the aforementioned linking characteristics. So, for illustrative purposes, I created two hypothetical graphs – and analysis otherwise – based on an up versus down dichotomy.
Connecting the dots, it seems apparent what a search engine should do in a situation. If Manipluative Link Profile X has an overall aggregate domain value of 80 based on this “muddy hill” (one step backward, two steps forward, one step backward, two steps forward) characteristic, as opposed to a smooth hill profile (one step forward, one step forward, one step forward) with an absolute link value of 70, it would seem to be in their best interests to flip Profile X with Profile Y based on the algorithmic probability that the “muddy hill” link profile had created an non-organically proven ranking.
Whether or not this is a current, domain-level link valuation judgement is unknown – but whether or not it makes sense is not.
Muddy Hills and Constant Links
How a search engine would completely flesh this out is uncertain. Some things I have thought about it recursive link activation – such as activating the full power of a link on a domain only after the link confirmed to the generalities of the domain value had in the past – such as high page traffic, clicks, or continued inbound links to the page and/or domain.
If a link added did not match or close in on the average characteristics of the domain that had placed it, perhaps it was not indicative of that domain’s value and as such, not the complete power that was possible. This thought process is an iteration of the reasonable surfer model – if a domain link is placed somewhere on a high value site but does not match the characteristics of those pages that originally made it a high value site – such as a link on a user’s profile page – it seems probabilistic, and correct, for a search engine to not value these links in the same way – even if they still appear shallow on in a site architecture or according to other commonly cited characteristics of the “reasonable surfer” – or good SEO in general.
If a link, post-placement, accords to the same mathematical patterns that proved out the domain’s original value, then it seems appropriate for it to be imbued with the maximum strength a link from a strong site such as SEOMoz might allow. If it does not, it seems more appropriate for a search engine to divide this value more appropriately, as it does not match the accord that had got the domain it’s value originally – or, sometimes, even come close to it.
Moving forward, good SEOs should think about links more than absolute placement, or even, absolute location on a page. Given free reign on external placement on a domain, it would be in their best interest to think deeper about things like normalized domain and page-level characteristics – and not only on more vanilla values like SEOMoz’s Page Authority, PageRank, and above the fold placement.