I just got this in John Alexander's newsletter and thought it was interesting:

For example, instead of just gearing up for a bunch of generalized
sports interests, you might do specific research and come
up with a very focused strategy based on people's true behavior.

Hmmm...for example, if you were selling a product to an audience of
sports lovers and with a little research, how many people are aware
that there is an audience of "Hockey fans" that appear to really
like "poetry" related to hockey?

(later)

Okay let's go back to the beginning and start again.

Here's what most people do wrong:
They hold their product or service in their hand and think...
How can I promote or advertise this item or service on the web?

They scan for related keywords and struggle to come up with
something in high demand that relates to what they happen to
hold in their hand (their product or service).

Is this not true? How many people will build a site selling a
product or service and then think about researching keywords
and optimization as a second step. I see it all the time. This
does not mean that we should not search for related keywords
and there are benefits of optimizing by taking a standard
approach.

But here's just one thing you can do as an alternative:

Don't start by building a Web site - start by researching peoples
behavior. Don't hold a product in your hand and then sweat over
how to attract an audience to it. Try the exact opposite approach.
Instead of choosing a product and then hunting to see if there is
a market for it, try hunting for a unique market by examining
people's behavior, then feed a "much needed" product into that
marketplace which you have proven there is a demand for.

Never start by randomly guessing at what people MIGHT be doing.

Compare it to the real world. Most people who are investing in
a business selling a new product are certainly going to research
the need for the product. Why would they spend a year developing
some new product and then just "hope" that people will want it.
It's far better to KNOW what people want and then offer them
something, right?