Last week at PubCon in Las Vegas, Matt Cutts of Google mentioned that the search engine giant will soon be treating sub-domains different in their results pages. At first it was reported that sub-domains were to be treated as folders, however in a comment thread over at Sphinn, Matt Cutts explained that this was a bit of an over statement. Here’s what he meant …

“This isn’t a correct characterization of what Google is looking at doing. What I was trying to say is that in some circumstances, Google may move closer to treating subdomains as we do with subdirectories.”

Additionally, a poster named Tedster a Sphinn spoke to Matt directly and had this comment:

“Matt said that Google will make it HARDER to get that 3rd result in a given search, and then increasingly harder for every result after that. But it will not be the exact same treatment that subdirectories get”

Exactly how this is going to play out is still up in the air. Matt promised to clarify things “at some point” after he gets back from the convention. This does bring to mind a number of questions and clarification is needed for sure. I doubt that most of it will come from Matt or Google, however, our industry is very good at analyzing an issue and making its own clarifications. I have one basic question related to this … Till now, sub-domains have basically been treated as separate web sites. How will this change if at all?

The real news in all of this is that Google is starting to make changes in how it handles sub-domains. It’s a tricky issue to solve as eradicating the sub-domain spam would surely hit quite a few legitimate users of sub-domains as well. This said, I don’t think this is the last change we will see in the handling of sub-domains. While this change will help eliminate the multiple listings in a single SERP, there are still the issues of the topical content spam so many sub-domains are filled with. I am looking for a paid linking type solution to appear which will selectively ignore sub-domains based on the quality of their content.

To repeat what I have said for quite some time now, if you can avoid using sub-domains, do it.

Apparently what has been happening is employee’s of BestBuy stores have been using a duplicate version of the Best Buy website to deny customers discounts which appear on the online store. The second version of the web site does not always display the discounts so employees can “show” the customers that the discount is not in effect. Here’s a link to the full story:

Best Buy Confirms It Has Secret Website

Now to be fair, it is still unsure if BestBuy planned this or the result a few over zealous employees who figured out a scam using an intranet version of the site. As far as I can tell the second site is not availible to surfers or spiders so it likely will not warrent a Duplicate Content Filter or Penalty, but those Best Buy employees might just get a Duplicate Content Warrent.

What make me wonder is they state right up front on their website that the technology is based on Latent Semantic Indexing or LSI. LSI is currently used by many SEO firms and private consultants and has been a staple of the professional SEO for years now. Sure they have tacked on a number of additional process to the idea, but if this patent goes through, and remains based on LSI analysis, what does this mean for the rest of the SOE community? Are we no longer going to be allowed to create automated LSI tools in house or offer the services to clients? I am sure they are banking on their additional processes to distinguish their product enough for a patent, but this whole idea is a scary one.

I will be watching this close to see what if any ramifications this may have on the industry and as soon as I can find an online version of the patent application I will post it with my analysis. Keep your eyes open on this one.

[ ARTICLE UPDATE ]

I received an email from Micheal Marshall the CTO/COO of Fortune Interactive and he clarified a few things for me. Here’s a couple of portions of his email (reprinted with permission) …

“Fortune Interactive is indeed in the process of patenting its technology. However, that patent is not an LSI patent. Some components of SEMLogic address the SEO challenges that arise from a search engine’s use of LSI or other semantic analysis methods.”

“SEMLogic is based on the concept of LSI and designed as a response to it. But SEMLogic is not equal to LSI. It is not accurate to characterize SEMLogic as an LSI method itself. It is a tool which in part deals with LSI.”

This is good news, and I am glad Mr. Marshall contacted me. I am still going to look a bit closer into the patent application when I get time. I am curious to see what type of restrictions a patent like this could possibly place on others looking to automate some if their LSI based analysis.

I am not surprised they did this as I proved that the supplemental index query was completely inaccurate and of no real value. My guess is since it was starting to circulate within the search community, they did not want people using an inaccurate tool thinking it was accurate. Will Google replace or fix the query? I hope so. I would think the best place for such a useful query would be within their Webmasters Tools. This would allow webmasters to see pages on their site which are flagged as supplemental. The one draw back to this for SEO or marketing shops is that would be you would only be able to check the supplemental results for web sites you controlled.

Kudos to “2K” for finding this in the first place, but bigger kudos go to Google for putting a lid on an inaccurate result so quickly.

If you want to try it, just use the following query format in Google’s search field..

site:www.yourdomain.com ***

The idea hit me to see if there were any significant differences in supplemental results of some of the web’s pioneers and today newer stars. While I am sure the business models and design of these sites also play a role in how many of their pages end up supplemental, I thought it would be fun to check a few anyway.

Well, the little test didn’t uncover any real consistent significant differences, but with such a small sample I didn’t really expect much. The results were basically all over the board and showed even wider variances depending on if you used the www or non-www form of the URLs. Additionally, this query from only worked on root domains, you can not for example do a search for supplemental pages on www.mydomain.com/articles/.

There were however a couple of interesting observances however. On the old school side, it is interesting to note that while EBay.com has over 1 Million pages indexed in Google, over 64% of them are supplemental. We also found that Web 2.0 is not an automatic advantage when you are trying to stay out of the supplemental index. Blogger.com has over 80% of its pages in the index, however del.icio.us has only an impressive .8% of its 2 million plus indexed pages listed as supplemental. I am betting a lot of this can be traced back to business models and in Blogger.com’s case the abundance of spam blogs being created on the site by users has to be a huge factor.

As I said earlier, this new query is not providing perfect results and the Amazon.com results above prove this point. I thought I would take a close look at what happened in the Amazon results and did a bit of digging. It turns out that the “***” portion of the query can change the status of a page. In the following example; one page is returned in both queries and when the “***” is used, it is changed from a normal listing to a supplemental listing. Look for the listing with the title beginning with the word “Image”

site:www.amazon.com 0735202753 Image

site:www.amazon.com 0735202753 Image ***

This effect can account for a portion of the inconsistency, however I doubt it can account for it all. Additionally, it is not known how this issue effects queries with no obvious reporting problems such as the one Amazon.com shows. One thing I do know, is that these results are at best estimates and should probably only be used for comparison or fun, not serious business.