Has the Anchor Text Signal Been Dampened?

Revising Anchor Text Strategy for Link Building

[martinibuster]

Excellent article over at SearchEngineWatch titled,
Are We Over-Engineering the Link Graph? [searchenginewatch.com]

The article analyzes the backlink profiles of several sites ranking for a specific term in the UK. The number ranked company had a surprise in it's backlink profile. The exact match anchor text ratio was very low, only 2% of the backlinks contained the targeted phrase.

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."

The position 7 site contained a higher ratio of exactly matched keyword phrase:

Mountain Warehouse really wants to rank highly for "outdoor clothing." ...The number of backlinks for our target keyword -- at 4,567 -- constitutes 73 percent of all total backlinks.

The author then asks:

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?

Has the anchor text signal been revised?

[Robert Charlton]

Has the anchor text signal been revised?

By itself, the test information in the article wouldn't make me think so. I've never felt that multiple repetitions of an exact phrase in any way looked "natural" to Google, and I'm often surprised that Google doesn't nuke sites that get thousands of exact matches for a short phrase.

By itself, the number of exact target phrase matches is kind of a simplistic metric.

Also, there's a factor which the article only mentions in passing... viz, Link anchor containing a "brand keyword"... which for me is what most probably accounts for the ranking.

The word "outdoor" in the company name, Cotswold Outdoor, coupled with some sufficiently good inbound links for the exact phrase... plus the right kind of page title and onpage text matches... should make the page do well for [outdoor clothing], and it does.

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."

What stood out for me when I clicked on the serp link was that the title on the page returned wouldn't have ranked in the described competitive situation....

<title>Cotswold Outdoor - Home</title>

In fact, it isn't the page that's ranking. While a page with Flash content is served up to users, the default page (which you can see in the Google cache) is the page that ranks.

It has the exact phrase match both on the page and in the title, along with other helpful vocabulary on the page....

<title>Outdoor Clothing, Climbing and Camping Equipment | Cotswold Outdoor</title>

The text returned on the default page is in fact very well crafted to rank for [outdoor clothing] ...not entirely to be dismissed in analyzing why something ranks when it's also got sufficient inbounds in several combinations to confirm it.

It would be interesting to know what other phrases containing the words "outdoor" and "clothing" were included in various inbound anchors of all the sites. I think that the combination, along with page titles and onpage text, need to be considered all together.

And there are many other link quality factors left out of the discussion that are involved as well.

[martinibuster]

Good catch, Bob. You're right. That site is showing one page for users with Flash and another page for those without (presumably Googlebot falls into that camp). Visit the page without JS enabled and the CSS fails to load, resulting in the left hand navigation looking like a site map.

What's the deal with the #7 ranked site with 73% of the backlinks matching the keyword phrase? I did a Yahoo Search of their backlinks which showed 1,480 backlinks [search.yahoo.com]. No doubt that's an incomplete result, but for the purposes of this discussion it's probably close enough. I did the same backlink search but with the words "outdoor clothing" [search.yahoo.com] specified to appear on the page, which includes anchor text since that is text- but also words occurring on the page, meaning the results should overcount. I only received 498 results, which is less than 73%. Furthermore, not all of those pages where that word occurs are anchors to MountainWarehouse.com.

I am not saying that the data the SearchEngineWatch author derived from the LinkFinder tool she is beta testing is wrong. Just noting that it doesn't match what is being shown on Yahoo.

Something that bugs me about that article is that the author speaks about percentages but not actual numbers. Percentages can be misleading. 2% of something can appear to be less than 73% of something else. Unless the 2% is of an exponentially larger amount than the 73%. What would you rather have, 73% of a dollar or or 2% of a thousand dollars?

Take a look at CotswoldOutdoor.com's backlinks [search.yahoo.com]. Yahoo shows 16,000 results. If you do a backlink search for backlinks on pages with the word "Outdoor Clothing" [search.yahoo.com] on it Yahoo returns 2,900 pages. Although not all of those pages are anchor text to CotswoldOutdoor.com, it does show there are almost six times as many sites linking back to CotswoldOutdoor.com with those words on the page than for the seventh ranked page. Could it be that "words on the page" may be an overlooked ranking factor in that article? Over 18% of the pages linking back to CotswoldOutdoor.com have those words somewhere on the page.

[RhinoFish]

"Percentages can be misleading."

MB, I guarantee you'll love this wiki article about exactly that (Simpson's Paradox):
[en.wikipedia.org...]

I'm begging you to read the whole article - one of the ways percentages can fool us is explained there, will blow your mind.

[RhinoFish]

in their analysis, did they simply count link anchor text occurrences without regard to the authority of the individual links?

with quality surpassing quantity several years ago now, i imagine count studies don't mean much.

i just skimmed the report, i bet i just missed their count weighting factors.

[martinibuster]

...with quality surpassing quantity several years ago now, i imagine count studies don't mean much.

That's a very important point.