Preferences (from lots of testing).
If domain is one word, split test:
example.com
Example.com
example.com/keyword
Example.com/keyword
Keyword.Example.com

I find that example.com IF it’s an ‘authority’ word/site and IF the company specializes in one product works very well.

If it is an authority site, but doesn’t meet the above criteria or isn’t an authority site at all, then example.com/keyword usually works well - but still test.

An example of the above is HP.com.
HP makes laptops, computers, printers, PDAs, etc.
Often users are afraid that clicking on an HP type ad can result in a lot of confusion of where they’re going. In this case, HP.com/printers often works better than HP.com as a display URL.

If domain is two words, split test:
WidgetExamples.com
WidgetExamples.com/Keyword
Keyword.WidgetExamples.com

(widgetexamples.com just doesn’t cut it).

If keyword is 3 or more words:
Same as above, but often Keyword.ThreeWidgetExamples.com or ThreeWidgetExamples.com/Keyword is so long it starts to wear on the user and usually ends up with less clicks, unless it’s a parody/joke type ad.

Demographic Responses
One of the reasons to test all the possible variations is that some markets respond more strongly to one over the other. For user demographics that aren’t net savvy, the /keyword or keyword.domain.com usually works better. For interent savvy markets, the reverse is often true. (Not always - just often - test)

From AdWords Hypothetical Formula.

I think the TLD discussion an interesting perspective to bring into the URL formula as well.

• How to Backup an Overture Account
• Did the Google AdWords update run you over?
• Display Spending Boosted for Local Sites
• Some Ideas for Split Testing AdWords
• AdWords shows more content network visibility