Tag Archives: Search engine optimization

Search Intent: Is the arms race for links finally over ?


Despite mounting evidence, businesses of all sizes cling to an outdated understanding of search engines, in particular there is a strongly held belief about how Google decides which sites to present in search results.

Over the past few years, Google has been placing more weight (!) on machine learning than on simpler measures such as Page rank, yet a fuller appreciation of how this affects positions in search results and subsequent traffic flows is not being explained by the industry. Even when we know what we are doing has dubious benefit at best. Our industry continues to mislead business by promising to ‘build links’, seek out ‘high authorty domains’ and generally waste clients budget by engaging in out-dated practices that are highly unlikely to benefit the client in the long term. I am not referring to spammy automation, we have all moved on from those days, I am referring to continuing to ‘build’ links at all.

Are links really that important in 2013?

The most prevalent belief is that sites are ‘rewarded’ for simply having more links from high authority sites than their competitors. This belief is based in turn on a belief that links correlate to rank, and that rank can be somehow ‘cemented’ by building links.  Consequently, agencies respond to competitive RFPs by agreeing to build links rather than challenging the clients firmly held beliefs.

This post will examine one page of results for a query, and I’ll attempt to show how far Google has come from simply matching keywords within the query to on-page content, then using links as a measure of trust to determine rank…

For this discussion, I will use the 3 term search  ‘list pax man’ using Google.co.uk

SERPs ordering for ‘list pax man’ query

Notice that Google suggests ‘pax man’ may be a mistake, I am presented ‘paxman’ as a single keyword suggestion.

Notice also that Google includes a link to Wikipedia, for a list of ‘Pac-Man’ video games – even though I did not type Pac-Man.

The first result is the most relevant to my query:

pax is a unix command, man refers to the manual and list is a common option

Both second and third results relate to Jeremy Paxman.

The second result contains the word ‘list’ once on the page (Amazon wish list).

The third result does not contain the word ‘list’ at all.

The fourth result shows how hard Google is working to provide all the possible options for my initial query.

Now let’s look at the number of inbound links for each result… if links were an important factor in the results set, then it would follow that the result suggested in #1 would have a handsome quantity of inbound links.

The stats (from MajesticSEO) for each URL in the SERPs are:

#1: 21 referring domains, 36 backlinks

#2: 1 referring domain, 4 backlinks

#3: 723 referring domains, 5,789 backlinks

#4: 25 referring domains, 156 backlinks

OK, so what ? Perhaps the ordering of the SERPs still has only something to do with links, but the quantity and volume are less important than the relevance of the page content…

Let’s make it more obvious what we are searching for, by repeating the ‘pax’ keyword in our query:

SERPs for the query ‘list pax pax man’

Now Google understands. We have achieved ‘entity disambiguation’ – a search engine nirvana.

We don’t mean ‘Pac-Man’ and we don’t mean ‘Paxman’. The repetition of the keyword ‘pax’ has ensured a full page of results about the Unix command ‘pax’.

Notice that the previous ‘best result’ for the Unix command has gone from #1

A URL from the apple development site takes its place.

In fact, the top results are now dominated by well frequented sites such as Apple, sourceforge, IBM and Wikipedia. The domain ‘manpagez.com’ is not a competitor in this new well-defined search space – which is odd, as it used to be the best possible match, just one query ago!

Looking at links to pages:

#1: 6,994 referring domains, 91,196 backlinks

#2: 64 referring domain, 156 backlinks

#3: no links reported

#4: 28 referring domains, 154 backlinks

If Google had a more ‘authoritiative’ match for the original query – if we assume links convey power and authority, then why wasn’t the result from Apple presented in that first search test ?

Look again at result #3 above from IBM. There are NO back links reported to this page by MajesticSEOs historic index.

The results above reflect the ‘importance’ of the results in relation to the query based on machine learning – not simply accumulations of links.

In order to be presented in SERPs, considerable effort must be made in ensuring that the clients site serves the need of the audience.

People looking for man pages on the pax command are the sort of people that visit developer.apple.com. Once Google has clear signals that the query is for the Unix command and not ‘Pac-Man’ or ‘Paxman’, then the SERPs are populated with the domains that serve this audience best. Unfortunately for manpagez.com, they do not attract and retain attention sufficiently to be returned for the second search.

I encourage you to try this for yourself and further to demonstrate to clients the dissociation of links with position in SERPs for queries. There is simply little to be gained by adding links to a site – the most important challenge is to give visitors what they want on the site and be clear about the content offered.

Growing an audience that uses the site regularly is more important than just adding links. Work on building advocates and the links will follow to support that effort. If you work on building links, then you are only increasing risk for the future.

Let me know what you think matters most by commenting below 😉


Enhanced by Zemanta

SEO Colour Experiment Update

The Google images experiment has been running for 4 days, and so far, just two of the ten images published in the previous post (laying out the details of the experiment) have been indexed; namely the purple image and the gray image.

Getting images indexed in Google
Getting images indexed in Google - purple preferred

Getting Google to own up to indexing the gray image requires advanced search options, it is revealed for searches of black and white and gray images:

Indexed images revealed using black and white filter
Indexed images revealed using black and white filter
Indexed images revealed using gray filter
Indexed images revealed using gray filter


There are some early takeaways from the experiment so far:

  1. Google chose to index just two of the available images in the post, a purple image and a gray image. This might suggest that galleries of images are not a good idea for SEO, as some of those images might not make it into the index. This could have been caused by the naming convention I used in the test – most images had very similar filenames – so perhaps the indexing of just two of a page full of test images is to be expected ? Galleries may cause Google to make a choice about which image to index and relate to the content…
  2. Looking back at the number of results in Google for different coloured images, it is interesting to notice that there were only 55k purple results and 61k gray results… perhaps this is why those two images were chosen and indexed by Google, to boost the number of available search results for those colours (purple and gray)? As with keyword research for universal search results, picking under-competed image types could provide quick wins…

What else is interesting so far ?

Searching using the keyword ‘experiment’ and then filtering for ‘large’ images and selecting ‘purple’ images shows our test image in #1 as might have been predicted ! After all, it was a large purple image with ‘experiment’ in both the filename and alt text.

Results for search 'experiment'
Results for search 'experiment' with large/purple filters set

The other images in the search results had the following filenames:

#2 http://i259.photobucket.com/albums/hh289/LynnViehl/Experiment2.jpg

#3 http://www.eu-atp.org/wordpress/wp-content/uploads/2009/07/experimento-1024×768.jpg

#4 http://4.bp.blogspot.com/-_34x_4u4wgM/TzBW2aQ0SzI/AAAAAAAAAp4/yr-2nm5TeXI/s320/the%2Bgolden%2Blight%2Bexperiment%2Bpurple%2B%2526%2Bhay%2B%2528preview%2529.jpg

#5 http://www.tablix.org/~avian/blog/images2/2012/02/two_spectrograms_recorded_at_the_crew_munich-t.jpg

So the obvious difference is that the test image had the keyword ‘purple’ as well as well as the word ‘experiment’ in the filename.

What about the gray image ?

It appears in position #4 in SERPs when using the same filters:

Keyword experiment and filters 'large' and 'gray'
Keyword experiment and filters 'large' and 'gray'

The competitive images for this search have the following filenames:

#1 http://jn.physiology.org/content/103/6/2938/F10.medium.gif

#2 http://www.sns.ias.edu/%7Ejnb/JohnphotosHtml/images/John%20Bahcall%2C%20Homestake%20mine%2C%20Cl%20Solar%20Neutrino%20Experiment%2C%20SD%2C%20%7E%201964.jpg

#3 http://arkansasagnews.uark.edu/Eason_rdax_250x345.jpg

#4 http://www.seoeditors.com/wp-content/uploads/2012/05/gray-seo-colour-experiment.jpg

#5 https://share.sandia.gov/news/resources/releases/2006/images/mat-celina_nr.jpg

So, on closer inspection the images at the top of the SERPs are all from authority sites… Journal of Physiology, two .edu sites and a .gov

The test image from this blog for the keyword ‘experiment’ performs quite well given the competition here !

By searching specifically for ‘gray experiment’, the image performs better and is returned in #1:

Search specifically for gray experiment
Search specifically for gray experiment

Positions #2 and #3 discuss the YouTube ‘gray’ channel theme.

At this stage, not many more insights are possible so I will return to this experiment in a few days time when more of the original set of images may have been indexed.



Enhanced by Zemanta