Links Vs. Web References As Relevance Signals

For a long time, links have been the largest factor in search rankings. While (in my opinion) this is still true today, other signals are gradually growing in importance. One of the most interesting of these is “web references” or “citations”. A web reference refers to a mention of a brand, product or web site, which is not referenced in a link (or perhaps referenced as a no-followed link).

An analogy will help illustrate the potential of this concept. Many have referred to links as a voting process, where each link is a vote for the site receiving the link. Consider a medieval country where the nobles vote for the next king when the current one dies. There is an election process, but only the upper class gets a vote. This is an elitist process, as the great majority of people have no vote.

This is what the current link-centric ranking algorithms are. You need to have a web site to participate. No web site, no vote. What would happen if everyone who uses the web had a vote? This can happen as a result of the growing use of social media web sites such as Twitter, Facebook, Delicious, StumbleUpon and similar sites. Mentions of brands, products, or web sites on these social media sites can be considered as votes as well. For example, if Coca Cola gets mentioned more often than Pepsi (or vice versa), that could be used as a signal to elevate the rankings of the Coca Cola web site.

You can take this a bit further too. Let’s say that over time that Coca Cola has received 1.2 times as many mentions as Pepsi. Then say Pepsi starts growing its presence, and for three consecutive months Pepsi gets more mentions than Coke. This type of change might represent a short term marketing campaign by Pepsi, but if it sustains itself it could represent a shift in the importance of one company over the other. This may cause the search engines to start ranking Pepsi over Coke—even before the total mentions over all time of Pepsi exceed those for Coke.

Some current examples

The biggest issue in these signals is signal quality. To help deal with this, expect the search engines to start using the signals in select ways where the quality of the data is quite high. Here are some examples.

Web references are already known to be a ranking factor in local search. In David Mihm’s study of Local Search Ranking Factors the panel of experts surveyed put citations (or web references) as the second most important ranking signal, only behind that of location. These citations can as simple as a phone number, a street address, business name, product name or some combination of these.

These may show up on the web in yellow page sites (which often don’t link to the listed businesses), and in raw data provided by data aggregators such as InfoUSA and Acxiom. One of the experts, Ed Reese commented of the importance of “citation sources related to your industry (like associations and social networking groups).”

Another way that web references can provide a stable signal is when surges in activity occur. Consider a brand that is getting a dozen or so mentions a day across the web that suddenly gets thousands of mentions across a few days. This is a strong indicator of a hot news item. This type of signal could be used to discover news and rank it by comparing the signal strength of one item to another.

For my final example, I will speculate a bit. Early in 2009 Google pushed out an algorithm change that eventually got the label of the “Vince change” (named after the Googler who did the work). In February people started reporting that some Google update had taken place, as Aaron Wall did in this post on Google Branding. A common belief was that Google began to heavily favor brands over non-brands. Matt Cutts later came out and said that this was not so much about brands, as it was about an increase in weight on trust and authority.

While it is premature to state that brand mentions in social media sites were a significant factor in the Vince change, you can see how a long history of regular mentions across the web would factor into a trust algorithm. For example, if a brand gets mentioned hundreds or thousands of times per day over a long period of time, that could be treated as a positive trust ranking signal. You may be able to do a similar analysis with just links, but this is an example where the social media signals show ongoing stability, and this provides the search engines with more data—which is a good thing.

There are two big problems with search engines using social media signals today.

The first is that social media sites are a bit elitist too at the moment. While they are available to everybody, not everybody is using them. More ways to use their data will emerge as use expands, simply because the data is getting better.

Second, the “wisdom of the mob” is inherently quirky. Consider the surge in mentions for Stephen Colbert during the campaign to get him to rank number one for the query “greatest living American”. This was a fad that ultimately had little to do with Stephen Colbert. It faded over time, of course, but the search engines would prefer not to be fooled by similar surges in the future.

Expect ongoing efforts by the search engines to find good ways to use these types of signals. As a publisher of a web site, it means you should begin to get active in social media environments. While this will unfold over a long period of time, major new brands are being created within social media environments today. It is OK to begin with some experiments to determine how you want to engage, but the time to start those experiments is now.