Todd “Oilman” Friesen recently did a post emphasizing the importance of location, location, location. In it, the Oilman takes Kevin Rose to task for dividing up the powerful Digg empire into 3 sites. It’s a mistake that lots of webmasters make.
There are many reasons for this. Not to long ago, a big trick was to use sub domains (e.g. subdomain.yourdomain.com) because for a while search engines treated a link from a sub domain as if it came from a different site. Pretty simple way to get additional inbound links. I still get questions about this trick from people. But the search engines closed the door on that one a long time ago.
Another problem is that webmasters look at large media properties, such as CNet Networks. They see a large array of web sites, all heavily cross linked. They convince themselves that the existence of the large number of domains that is what makes the whole network successful. However, you can rest assured that when a major search engine sees such a large network of interlinked sites that they pretty quickly figure out that they have one owner.
I have seen companies insist on developing new sites, or putting things on sub domains for these very reasons. Don’t do it. It’s a waste of your time.
One of the biggest reasons is that you end up having to market numerous web properties, instead of one. This means separate messaging, different promotional campaigns, and subdivided results. Another big reason is captured by the old saying “a rising tide lifts all boats”. A link to the home page of a single master site benefits all the on site content.
I know of one good exception to the rule of “keep all your content on a single site”. If you acquire a third party site that has a boatload of links to it already, you may wish to keep it separate. This is the unique case where the acquired web site has its own audience (traffic) and brand that you wish to keep in place. This is a major reason that CNET Networks looks the way it does.
If that’s your strategy, go for it. Otherwise, keep it all on one site.