Are We Over-Engineering the Link Graph? [Best of SEW 2010 #1”
Brand versus keyword-rich anchor text: a case study on ranking factors.
Brand versus keyword-rich anchor text: a case study on ranking factors.
Editor’s note: As 2010 winds down, we’re celebrating the Best of 2010, our top 10 most popular columns of the year on Search Engine Watch, as determined by our readers. Every day over the next two weeks, we’ll repost the most popular columns of the year, starting at No. 10 and counting down to No. 1 on Dec. 31. Our countdown continues today with our No. 1 column, which originally was published on July 16. Enjoy! “Come writers and critics who prophesize with your pen, and keep your eyes wide for the chance won’t come again… for the times they are a changing.” – Bob Dylan
OK, admittedly that’s a rather grandiose opening for a case study (even for me). However, I want to give this piece a little temporal context and open with the position that it feels to me as if we’re on the cusp of significant change.
I’m not alone in thinking this. A good many of my peers and colleagues agree that there are signs, trends, announcements, changes, research, and anecdotes that point toward a potential shift. A shift in mindset certainly, and perhaps potentially a shift (or growth) in the number of ranking factors and most interestingly a “downplay” in the importance of previously held key ranking factors.
A few examples:
Earlier this week on Search Engine Watch, Eric Enge posited that search engine algorithms are seeking other factors (perhaps social media participation) as “links are no longer as valuable as a citation” — due to the rewards that could previously be attained from gaming links.
Also this week, in a guest post on Wordstream, “The Evolution of Ranking Signals” David Harry neatly stated, “Consider that when you have a signal that outweighs the others, it becomes a magnet to spammers.” Harry goes on to suggest that recent changes, such as QDF and the Caffeine infrastructure change, would allow for greater access and accountability for elements such as temporal, behavioral, and social signals.
To Our Case Study Then…
I’m involved as an early beta user of the new InfluenceFinder link building tool. InfluenceFinder has a number of extremely useful filters and signals to help classify link intelligence; however, in our work with the team, we’re seeking to collaborate on prototyping new reporting features. Features that will be useful in a new environment, with potentially shifting criteria; features that permit the discovery of additional intelligence that may not previously have been of interest.
In this case, we examined the ratio of brand versus keyword anchor text, for the top 10 sites ranking for (that) keyword.
Top 10 Sites Ranking for “Outdoor Clothing” in the UK in June 2010
Using InfluenceFinder data, and manual analysis each of the following domains’ backlink profile with link anchors, segmented as follows:
Number of Brand Keyword Links per Top Ranking Site
Clearly, the sites that rank higher in the top 10 also have high numbers of links with brand keywords as anchor text. In fact, the correlation figure shown above expresses the correlation between number of brand links and position in ranking order. If 1 is perfectly correlative, then 0.67 is certainly a strong correlative relationship and a figure of some interest, when we consider there are a couple hundred factors that reportedly contribute to rank.
Number of Target (Outdoor Clothing) Keyword Links per Top Ranking Site
In this case, it’s interesting to note that sites with the highest numbers of target-keyword-as-anchor appear in the latter half of the top 10. Could it be that keyword focused anchor text isn’t as strong a ranking signal as previously held? In this case, there’s still a considerable correlation of 33 percent between number of keyword anchors and site position, though it’s interesting that the brand correlation in the previous graph is double.
Cotswold Outdoor — Detailed Breakdown
Cotswold Outdoor was the number one ranked site in the U.K. for the term “outdoor clothing” at time of analysis, and still is at time of writing.
The backlink profile for Cotswold Outdoor is composed primarily of brand-related anchors. What stood out for me: only 2 percent of their total backlinks contain the term “outdoor clothing.” Additionally, most anchor links are composed of one, two, or three words and have a natural spread in terms of linking-site quality (as measured here by AC Rank).
Backlink Profile for Sites in Positions 2 Through 5
In this case, I’ve shown just the primary “anchor-type” chart, as opposed to “by AC Rank” range too, as in each case there was little deviation from the pattern seen with Cotswold Outdoor.com, which is one of natural, proportional decay as AC Rank increases.
In respect to second- and third-placed sites, there’s little-to-nothing to distinguish the backlink profiles of either Go Outdoors or Blacks, compared to either each other or Cotswold Outdoor in position one. A couple interesting things to comment on here:
This sparks lots of interesting further consideration. For instance, could it be that target keyword as anchor is a signal that can be over-engineered? In the case of Millets, how does Google determine relevance to keyword target, when this site has but nine links containing said target?
When we look to the bottom half of our top 10 table, things get even more interesting…
Position 6, and First Anomalous Backlink Profile in Top 10
In examining the backlink profile for our top five ranking sites, there was little to choose. All followed a similar composition and ratio of brand to target keyword. Webtogs, however, is the first site in the top 10 to display an anomalous profile.
In this case, the target keyword constitutes 14 percent of the total keyword anchors for this site, whereas for the top five the percentage contribution ranges from 0.25 percent (Millets) and 6 percent (Tiso). Also, Webtogs is the first site ranked in the top 10, to have less “brand” type anchors than any other.
Webtogs should look into status of other contextual relevance signals outside of keyword-rich backlinks and how these signals play out across other media and content types. Plus they should consider strengthening brand anchors in backlinks too, as a product of brand-building PR activity.
Position 7 and Most Anomalous Backlink Profile of Top 10
Mountain Warehouse really wants to rank highly for “outdoor clothing.” What do you think?
The number of backlinks for our target keyword — at 4,567 — constitutes 73 percent of all total backlinks. Setting aside other ranking factors, this particular case alone is intriguing and leads me to wonder if there’s indeed a ratio or composition of keyword anchor to total anchor, beyond which this signal is no longer relevant, or potentially limiting?
If tasked with taking this brand further, I’d look into brand building PR activities, unifying social touch points (I couldn’t see their Facebook or Twitter accounts referenced on their own site), and developing universal content types, particularly video.
The Bottom 3 (Positions 8-10)
As we come to the final three of our top 10 sites for “outdoor clothing,” Berghaus in position seven seems to have a similar composition of brand-to-keyword-to-other, as our top five sites. If I worked for Berghaus, I’d look to do some simple on-page optimization in the first instance, as there are indeed gaps.
SnowandRock is another site with an anomalous profile, yet still in the top 10. In this case, this site has a fairly low ratio of brand anchors, when compared to the other top ranking sites, with the exception of Webtogs and Mountain Warehouse.
Finally (and possibly most interestingly), we see that in position 10, with a total of zero target-keywords-as-anchor, we have The Outdoor Shop. So in this particular case study, it seems like it’s possible to rank in the top 10 for a keyword for which you have absolutely no anchored backlinks.
My objective here is to spark discussion and feedback (rather than present a case study as overarching truth). However, to do so, I need something from you.
Tell me… does this single case study strike a chord with you? Do you have contrary or supporting research to consider against this study? Is this the sort of backlink analysis we should put on the agenda as search continues to evolve?
Should we ask the InfluenceFinder team to make damn sure that this is a standard report? If yes, what else would you like to see to further this analysis that could be considered as a future development feature?
As an example, I chatted through this research with link-campaign specialist and co-founder of Ontolo, Garrett French; who said that “this research indicates to me, and what I’ve seen and heard over time is that each keyword space/market space (and georegion) is going to have its own definition of a ‘natural’ link graph. If InfluenceFinder can help to spot/outline this natural link graph that would be a great boon to link builders seeking direction for link campaign design.”
Although I want to refrain from drawing sweeping conclusions regarding ranking factors based on one case study, I want to stress that as a piece of research this case is a perfect example of the importance of market and competitor analysis at the outset of any link campaign. I agree with Garrett, that each keyword-market and sector is different, which is why any link building campaign starts with bespoke analysis such as this. Once we have the tools and the data, the interpretation and action is the critical part!