Sunday, June 8, 2008

The TLD Debate Revisited

April 19, 2008 – 7:11 pm
Do Search Engines Give Preference to Certain TLDs (Top Level Domains) Than Others?

This question has been circulating around SEO circles for some time now. SEOs are always looking for “white hat” methods to improve search engine rank placement (SERP), and so naturally one of the questions we ask concerns preferential treatment of TLDs. The general consensus among SEO professionals is that the search engines (SEs for short) do not grant any special treatment to various TLDs. That is, they do not give certain TLDs a “head start” by granting more intrinsic trust or authority to them.

Nevertheless, I’m not wholly convinced of that, and I have learned the scientific value of never simply taking anyone’s word for things. In the absence of hard data, I find that many claims of members of the “SEO Intelligentsia” are simply unfounded and do not hold up under rigorous, methodical experimentation.

So, I decided to conduct my own experiment. I purchased six different domains covering what I considered at the time to be the top four TLDs: .com, .org, .net and .us. In order to control for all variables except the TLD choice, I took the following steps:

* I purchased all the domains at the same time and registered them all with the same hosting company
* I chose the same pattern for all the domains, ensuring that each of the keywords contained therein received roughly the same amounts of search volume
* I parked all the domains at the same parking host
* I did the same amount of linkbuilding on all of them, which was to submit each to only one popular social bookmarking site
* I ensured that the parking pages of each site contained roughly the same amount of outbound links

And that was it. I did no further optimization on any of the domains. Then, I left them alone and forgot about them. The domains selected were:

1. gamusic[dot]org
2. gamusic[dot]us
3. gapoetry[dot]org
4. gapoetry[dot]us
5. gapoetry[dot]net
6. gawriting[dot]com

I feel that I should say that, going in, my assumption was that, if there were any SE TLD preference at all, it would be given to the .us. My rationale for this was that many government agencies were beginning to use the .us in conjunction with the .gov in the URLs. And, since I had conclusively demonstrated in other experiments that .gov and .edu TLDs possess more intrinsic authority and trust that other TLDs, I reasoned that this might be a case of trust by association.

So Which TLD(s) Performed Better?

Well, initially I felt that my original hypothesis - that the .us TLD would be granted some SE “brownie points” - seemed to hold true. I noticed about two weeks after parking the domains that the .us was the only one which recorded at PageRank (PR) of 1; all the others were either zero, or still greyed out.

I redently removed all the domains from their parking lots and moved them back to my original host server. The domains had been parked for about five months now, and I decided I no longer liked the parking company. When I went to remove them, I was startled at what I found.

The .net domain outscored the others - and not by one or two PR points - but by three or four! was now recording a PR of 4, while the other domains reported no change at all!

Of course I thought this rather strange. I began to ask myself what factor(s) were responsible for the substantial difference in PR. Since the only two variables not controlled for were back links and traffic, I started with these. To my amazement, I found that had the same one or two backlinks it did when I started. Even more interesting was that one of the .orgs had one more backlink that the .net, but still reported no change in PR!

Next I looked at traffic reports. I was again surprised. None of the sites received even a modicum of real traffic. In fact, the Alexa and Compete people counts could not even be collected because the sites received so few visitors. Thus, traffic could not have been a factor.

Then, I considered IP address neighborhoods. Nope, this could not have been a determining factor, either; all domains were parked at the same DNS server locations.

What about content? Nope. All domains had similar parking page templates (templates which were assigned by the parking host and which could only be very minimally customized). All parked pages consisted of roughly the same amount of moderately-keyword-optimized text.

Additionally, all domains had virtually the same amount of outbound links. Furthermore, all the outbound links went to the same repository of sponsored ads, so the dramatic PR difference certainly could not have been attributed to outbound link destination.

Indexing? Well, of the sites that were actually indexed by the SEs, all were indexed at roughly the same time. And domain age could not have been a factor, as all domains were conceived and registered on precisely the same day and at precisely the same time.

So, Which TLD Got Preference?

Scientifically speaking, the experiment would have to be repeated a statistically significant number of times in order to conclusively state that the .net TLD received any preferential treatment by the SEs. However, the results are interesting and surprising, nonetheless. In the best traditions of science, the results show that which is not-intuitive: that .nets may be granted some degree of intrinsic trust and/or authority which the others are not. It may not be due to any conscious coding on the part of the algorithm programmers, of course. However, there may be some inherent code in the algorithms which cause them, for whatever reason, to favor the .nets.

A Few Additional Notes

Additionally, according to Wikipedia, the .net TLD is the third most popular TLD behind .com and .de. And according to domain data we collected at the New Fine Arts Lab, the number of sites with the top four TLDs is as follows:

1. .com = 13.22 billion sites
2. .org = 1.41 billion
3. .us = 1.53 million
4. .net = 1.78 billion
5. .de = 1.37 billion

(source: Google)
Also, we collected some data on domain name appraisal values. We found that .coms generally go for the highest average prices when sold. Right behind them are the .nets. Might there be some correlation between the “hidden hand of the market” (to use Adam Smith’s phrase) and the trust/authority scores of a domain?

No comments: