Many aspects are involved when you evaluate a link. As we just
outlined, the most commonly understood ones are authority, relevance,
trust, and the role of anchor text. However, other factors also come into
play.1. Additional Link Evaluation Criteria
In the following subsections, we discuss some of the more
important factors search engines consider when evaluating a link’s
value.
1.1. Source independence
A link from your own site back to your own site is, of course,
not an independent editorial vote for your site. Put another way, the
search engines assume that you will vouch for your own site.
Think about your site as having an accumulated total link juice
based on all the links it has received from third-party websites, and
your internal linking structure as the way you allocate that juice to
pages on your site. Your internal linking structure is incredibly
important, but it does little if anything to build the total link
juice of your site.
In contrast, links from a truly independent source carry much
more weight. Extending this notion a bit, it may be that you have
multiple websites. Perhaps they have common data in the Whois records
(such as the IP address or contact information). Search engines can
use this type of signal to treat cross-links between those sites that
are more like internal links than inbound links earned by
merit.
Even if you have different Whois records for the websites but
they all cross-link to each other, the search engines can detect this
pattern easily. Keep in mind that a website with no independent
third-party links into it has no link power to vote for other
sites.
If the search engine sees a cluster of sites that heavily
cross-link and many of the sites in the cluster have no or few
incoming links to them, the links from those sites may well be
ignored.
Conceptually, you can think of such a cluster of sites as a
single site. Cross-linking to them can be algorithmically treated as a
single site, with links between them not adding to the total link
juice score for each other. The cluster would be evaluated based on
the inbound links to the cluster.
Of course, there are many different ways such things could be
implemented, but one thing that would not have SEO value is to build a
large number of sites, just to cross-link them with each other.
1.2. Linking domains
Getting a link editorially given to your site from a third-party
website is always a good thing. But if more links are better, why not
get links from every page of these sites if you can? In theory, this
is a good idea, but search engines do not count multiple links from a
domain cumulatively.
In other words, 100 links from one domain are not as good as one
link from 100 domains, if you assume that all other factors are equal.
The basic reason for this is that the multiple links on one site most
likely represent one editorial vote. In other words, it is highly
likely that one person made the decision. Furthermore, a sitewide link
is more likely to have been paid for.
So, multiple links from one domain are still useful to you, but
the value per added link in evaluating the importance of your website
diminishes as the quantity of links goes up. One hundred links from
one domain might carry the total weight of one link from 10 domains.
One thousand links from one domain might not add any additional weight
at all.
Our experimentation suggests that anchor text is not treated
quite the same way, particularly if the links are going to different
pages on your site. In other words, those multiple links may
communicate link value in a way that is closer to linear. More links
might not mean more importance, but the editorial vote regarding the
topic of a page through the anchor text remains interesting data for
the search engines, even as the link quantity increases.
Think about the number of unique linking domains as a metric for
your site, or for any site you are evaluating. If you have a choice
between getting a new link from a site that already links to you as
opposed to getting a new link from a domain that currently does not
link to you, go with the latter choice nearly every time.
1.3. Source diversity
Getting links from a range of sources is also a significant
factor. We already discussed two aspects of this: getting links from
domains you do not own, and getting links from many different domains.
However, there are many other aspects of this.
For example, perhaps all your links come from blogs that cover
your space. This ends up being a bit unbalanced. You can easily think
of other types of places where you could get links: directories,
social media sites, university sites, media websites, social
bookmarking sites, and so on.
You can think about implementing link-building campaigns in many
of these different sectors as diversification.
There are several good reasons for doing this.
One reason is that the search engines value this type of
diversification. If all your links come from a single class of sites,
the reason is more likely to be manipulation, and search engines do
not like that. If you have links coming in from multiple types of
sources, that looks more like you have something of value.
Another reason is that search engines are constantly tuning and
tweaking their algorithms. If you had all your links from blogs and
the search engines made a change that significantly reduced the value
of blog links, that could really hurt your rankings. You would
essentially be hostage to that one strategy, and that’s not a good
idea either.
1.4. Temporal factors
Search engines also keep detailed data on when they discover the
existence of a new link, or the disappearance of a link. They can
perform quite a bit of interesting analysis with this type of data.
Here are some examples:
When did the link first appear?
This is particularly interesting when considered in
relationship to the appearance of other links. Did it happen
immediately after you received that link from the New
York Times?
When did the link disappear?
Some of this is routine, such as links that appear in blog
posts that start on the home page of a blog and then get
relegated to archive pages over time. However, perhaps it is
after you rolled out a new major section on your site, which
could be an entirely different type of signal.
How long has the link existed?
You can potentially count a link for less if it has been
around for a long time. Whether you choose to count it for more
or less could depend on the authority/trust of the site
providing the link, or other factors.
How quickly were the links added?
Did you go from one link per week to 100 per day, or vice
versa? Such drastic changes in the rate of link acquisition
could also be a significant signal. Whether it is a bad signal
or not depends. For example, if your site is featured in major
news coverage it could be good. If you start buying links by the
thousands it could be bad. Part of the challenge for the search
engines is to determine how to interpret the signal.
1.5. Context/relevance
Although anchor text is a major signal regarding the relevance
of a web page, search engines look at a much deeper context than that.
They can look at other signals of relevance. Here are some examples of
those:
Nearby links
Do the closest links on the page point to closely related,
high-quality sites? That would be a positive signal to the
engines, as your site could be seen as high-quality by
association. Alternatively, if the two links before yours are
for Viagra and a casino site, and the link after yours points to
a porn site, that’s not a good signal.
Page placement
Is your link in the main body of the content? Or is it off
in a block of links at the bottom of the right rail of the web
page? Better page placement can be a ranking factor. This is
also referred to as prominence, which has
application in on-page keyword location as well.
Nearby text
Does the text immediately preceding and following your
link seem related to the anchor text of the link and the content
of the page on your site that it links to? If so, that could be
an additional positive signal. This is also referred to as
proximity.
Closest section header
Search engines can also look more deeply at the context of
the section of the page where your link resides. This can be the
nearest header tag, or the nearest text highlighted in bold,
particularly if it is implemented like a header (two to four
boldface words in a paragraph by themselves).
Overall page context
The relevance and context of the linking page are also
factors in rankings. If your anchor text, surrounding text, and
the nearest header are all related, that’s good. But if the
overall context of the linking page is also closely related,
that’s better still.
Overall site context
Last is the notion of the context of the entire site that
links to you (or perhaps even just the section of the site that
links to you). For example, if hundreds of pages are relevant to
your topic and you receive your link from a relevant page, with
relevant headers, nearby text, and anchor text, these all add to
the impact, more than if there happens to be only one relevant
page on the site.
1.6. Source TLDs
Indications are that there is no preferential treatment for
certain top-level domains (TLDs), such as .edu, .gov, and .mil. It is
a popular myth that these TLDs are a positive ranking signal, but it
does not make sense for search engines to look at it so
simply.
Matt Cutts, the head of the Google webspam team, commented on
this in an interview with Stephan Spencer (http://www.stephanspencer.com/search-engines/matt-cutts-interview):
There is nothing in the algorithm itself, though, that says:
oh, .edu—give that link more weight.
And:
You can have a useless .edu link just like you can have a
great .com link.
There are many forums, blogs, and other pages on .edu domains
that spammers easily manipulate to gain links to their sites. For this
reason, search engines cannot simply imbue a special level of trust or
authority to a site because it is an .edu domain. To prove this to
yourself, simply search for buy viagra site:edu
to see how spammers have infiltrated .edu pages.
However, it is true that .edu domains are often authoritative.
But this is a result of the link analysis that defines a given college
or university as a highly trusted site on one or more topics. The
result is that there can be (and there are) domains that are
authoritative on one or more topics on some sections of their site,
and yet can have another section of their site that spammers are
actively abusing.
Search engines deal with this problem by varying their
assessment of a domain’s authority across the domain. The publisher’s
http://yourdomain.com/usedcars section may be
considered authoritative on the topic of used cars, but
http://yourdomain.com/newcars might not be
authoritative on the topic of new cars.
One technique that link brokers (companies
that sell links) use is the notion of presell
pages. These are pages on an authoritative domain for which
the link broker has obtained the right to place and sell ad copy and
links to advertisers. The link broker pays the domain owner a sum of
money, or a percentage of the resulting revenue, to get control over
these pages.
For example, the link broker may negotiate a deal with a major
university enabling it to place one or more pages on the university’s
website. The links from this page do have some of the inherent value
that resides in the domain. However, the presell pages probably don’t
have many (if any) links from other pages on the university site or
from other websites. As with other forms of purchasing links, presell
pages are considered spam by search engines, and pursuing these types
of links is a high-risk tactic.
Ultimately, every site and every page on every site gets
evaluated for the links they have, on a topic-by-topic basis. Further,
each section of a site and each page also get evaluated on this basis.
A certain link profile gives a page more authority on a given topic,
making that page likely to rank higher on queries for that topic, and
providing that page with more valuable links that it could then give
to other websites related to that topic.
Link and document analysis combine and overlap, resulting in
hundreds of factors that can be individually measured and filtered
through the search engine algorithms (the set of instructions that
tells the engines what importance to assign to each factor). The
algorithms then determine scoring for the documents and (ideally) list
results in decreasing order of relevance and importance
(rankings).
2. Determining a Link’s Value
Putting together a link campaign generally starts with researching
sites that would potentially link to the publisher’s site and then
determining the relative value of each potential linker. Although there
are many metrics for evaluating a link, as we just discussed, as an
individual link builder many of those data items are hard to determine
(e.g., when a link was first added to a site).
It is worth taking a moment to outline an approach that you can
use today, with not too much in the way of specialized tools. Here are
factors you can look at:
The PageRank of the home page of the site providing the link.
Note that Google does not publish a site’s PageRank, just the
PageRank for individual pages. It is common among SEO practitioners
to use the home page of a site as a proxy for the site’s overall
PageRank, since a site’s home page typically garners the most links.
You can also use the Domain mozRank from SEOmoz’s Linkscape tool to
get a third-party approximation of domain PageRank.
The perceived authority of the site. Although there is a
relationship between authority and PageRank, they do not have a 1:1
relationship. Authority relates to how the sites in a given market
space are linked to by other significant sites in the same market
space, whereas PageRank measures aggregate raw link value without
regard to the market space.
So, higher-authority sites will tend to
have higher PageRank, but this is not absolutely the case.
The PageRank of the linking page.
The perceived authority of the linking page.
The number of outbound links on the linking page. This is
important because the linking page can vote its passable PageRank
for the pages to which it links, but each page it links to consumes
a portion of that PageRank, leaving less to be passed on to other
pages. This can be expressed mathematically as follows:
- For a page with passable PageRank n
and with r outbound links:
- Passed PageRank =
n/r
This is a rough formula, but the bottom line is that the more
outbound links a page has, the less valuable a link from that page
will be.
The relevance of the linking page and the site.
Organizing this data in a spreadsheet, or at least being
consciously aware of these factors when putting together a link-building
campaign, is a must. For many businesses, there will be many thousands
of prospects in a link campaign. With a little forethought you can
prioritize these campaigns to bring faster results.
