First, a quick note. This is not the kind of link we normally publish. The page is on a wiki which in this case is a sort of a multi-author "blog", So it's not authoritative. But we are making an exception, because the ideas presented are worth considering.
I've mentioned elsewhere that search volumes for short generic keywords are not a strong predicter of traffic - because for some terms as much as 40% of those searches do not result in a click. Instead, the searcher looks at the results and immediately knows that they need to revise or refine their search terms.
In this case we're discussing branded search terms, not generic search terms, and so I'd say this approach has some potential use. It will be interesting to see if the article's predictions are accurate, come April 2010.
Yes, I always check keyword matrix with Google before starting a new website. Its no hard and fast rule, but its one of the set of useful tools to use as to get an indication of how well a new website could potentially perform.
Though with some of my new websites has really got my hopes up, only to completely fail.
Your best indicator is to ask yourself, will this website be very useful to a large audience, and hasnít been done before.