View Full Version : Subdomains and Search Engines

11-17-2005, 10:45 PM
Subdomains are a recurring quandary. Hardly a week goes by in the forums that someone somewhere doesn't ask if they can maybe improve their SE rankings by creating a few subdomains. Since last issue's article, To WWW or Not, generated so much mail (some of which I'm still struggling to answer), I thought perhaps we could address at least a few similar issues in this issue.

For those still scratching their heads, let's define some terms.

The DNS hierarchy consists of the root-level domain at the top, followed by top-level domains (TLD) like .com, .net, .org and a host of others. Then comes the domain, sometimes called a second-level domain, like example.com (try going to example.com, sometime, and you'll discover why writers should use this for all, uh, examples). Finally, a subdomain, sometimes called a child domain or a third-level domain, can be defined as part of the hierarchy. That leaves us with subdomain.example.com.

Astute readers will quickly realize that subdomain.example.com looks a whole lot like www.example.com. In our last issue, I explained that www existed because, back in the old days, one computer couldn't always handle all our Internet traffic, so we'd put different parts on different machines. So, www.example.com would handle our Web traffic, mail.example.com would handle our email, and maybe ftp.example.com would be our FTP server.

Subdomains can be used in much the same way.

In fact, one of the earliest uses of subdomains was in Universities, with each department often running their own server, giving us med.stanford.edu or math.stanford.edu. Using multiple servers is STILL one of the best justifications for subdomains, something you see done a lot at microsoft.com, as just one example.

And that should answer one of the more common forum questions I see. A lot of hosting companies are now offering 10 free subdomains with each account, but when you try to configure one you discover it has to go inside your main domain's folder. That essentially makes blog.example.com an alias for example.com/blog/. This is a limitation of the hosting company and NOT a limitation on subdomains. It's also a good example, I suspect, of something being worth what you paid for it. Free subdomains will probably always be limited to some extent.

So, what do search engines thing about subdomains?

Somewhere around 1999, a few of us discovered we could almost instantly capture a number one ranking on the old Excite search engine by the simple expedient of creating a subdomain using our keyword. For example, spaghetti.example.com was almost guaranteed to be number one for spaghetti. A very large part of the reason for this was that Excite always gave the home page of a domain a *significant* boost over interior pages. Ergo, spaghetti.example.com/ was always better than example.com/spaghetti /.

That's no longer true, of course, and I'm sure I don't need to tell anyone that subdomains were BADLY abused for a year or more. Like white text on white background, it was just too easy to last for long. In today's world, I've seen no indications in any of the major search engines that using a subdomain, in and of itself, will improve rankings.

However, that "in and of itself" is an important qualifier.

Inbound links and, especially, anchor text have grown dramatically in importance, and using a subdomain can still indirectly impact your site in that arena. You don't get much of a boost for having a keyword in your subdomain (some will say you get none), but you DO get a boost for having the keyword in any anchor text pointing to your site. Since a lot of people still create "naked links" like http://spagetti.example.com/, where the link and the anchor text is the same, having a keyword in your URL can gradually build some serious advantages.

Another possible use of subdomains is to defeat clustering. Back in the old days, it was entirely possible for one site to dominate the first page in a search; all you needed was ten pages that ranked higher than anyone else's. Searchers wouldn't even see your competition (and it was a huge ego boost, too). Clustering defeated that and today's search engines only show, at most, two pages from any one site. However, subdomains are treated as entirely separate entities, so it's entirely possible to have one.example.com show up with two pages, two.example.com show up with another two pages, etc., once again dominating the first page and burying your competition.

In my opinion, trying to defeat the clustering algorithms with subdomains is potentially a dangerous road to travel. If Google wants to give their visitors a wider array of choices, far be it for me to get in the way of a 900 pound gorilla. I think any trick that subverts the *intentions* of a search engine can only lead to short-term gain.

In June of 2002, I started a design for what would have been a large and diverse web site for a local community college. The diversity prompted me to use subdomains, especially since keywords in the URL then counted more heavily than today. I had already started to use thematical subdomains when I discovered a warning on Google's webmaster page specifically warning that such use, if done only to influence ranking, would not be seen favorably. Less than two weeks later, when I wanted to reference that warning on a forum, I found it had already been removed. I've never been able to prove it was there, nor have I seen any other suggestions from Google with similar foreboding regarding subdomains, but I've always taken that short-lived warning as an indication of an attitude that "may" lie just below the surface waters at Google.

Aside from SE backlash, which may exist only in my mind, there are some other disadvantages of subdomains.

One important disadvantage is that you need a whole lot more inbound links to rank well. Eight sites with 1,000 links each isn't as strong as one site with 8,000 inbound links. Similarly, eight sites with 100 pages each isn't usually as strong as one size with 800 pages.

Of prime concern, I think, is the very real possibility of running into a cross-linking penalty, especially on Yahoo. If you have eight subdomains and none of them link to each other, you don't have a problem. If you have eight subdomains and every one of them links to seven other subdomains on every page, you almost certainly WILL have a problem. Somewhere in the middle of that spectrum is where you need to be -- and, frankly, unless you have a good feel for where that middle is, I wouldn't recommend taking too many chances. Right now, if you get booted from Yahoo's index for excessive cross-linking, there's nothing for it but to throw away those domains and start over from scratch. The possible reward, in my opinion, simply isn't worth the possible risk.

So, does that mean I would never recommend using subdomains?

Never is a long, long time. Let me close with two examples, one where I feel subdomains would be dangerous and one where I think they could work well.

The site I was building for a community college in 2000 was about creating and running a successful e-commerce site. I initially wanted to have subdomains like hosting.example.com, design.example.com and promotion.example.com. In today's search engine world, I'm convinced a site like this would be a recipe for disaster.

On the other hand, one of the threads currently active at SEF talks about splitting a celebrity site into subdomains. The result might be related sites at britney.example.com, leno.example.com, and maybe robertredford.example.com. Unlike my e-commerce site, I suspect this might work well.

What the difference?

Any visitor interested in e-commerce is going to want to know about hosting AND design AND promotion. On the other hand, a visitor interested in Britney may have no interest at all in Leno. A designer that takes the user's needs into consideration is going to very heavily inter-link the e-commerce sites, but will inter-link the celebrity sites very loosely.

Another way of saying the same thing is that if you want to build separate sites they should all BE separate sites. They can all use the same design, use the same server and IP address, but any compulsion to tie them too closely together from a linking standpoint (as would have been necessary for my e-commerce site) is a good indication you only have one site that has been arbitrary chopped into pieces for better rankings.

And you can bet dollars to donuts the search engines will see it, too.

By Ron Carnell