Paul Saffo's talk at the Long Now Foundation (MP3 here) is a very good overview of foresight research heuristics/rules of thumbs/methods. Some notes:

"Hunt of Bin Laden, experts agree, Al Qaeda leader is dead or alive" is a great forecast because it accurately captures the uncertainty of the moment. The biggest mistake is to be more certain than what the fact suggest, especially today, at this very uncertain moment in time (where indicators are going in different directions). As Peter Schwarz says: "The difference between a good forecast and reality is...a good forecast has to be believable and internally consistent"
The job is not about predicting but rather mapping the “cone of uncertainty” on a subject. And, uncertainty means opportunity. It's a cone shape for commonsensical reasons and because uncertainty expands as you project further into the future. The important thing is to find edges: Where might they happen? There you should look for wild-cards to define the boundaries and science-fiction can be a good candidate for that matter (as well as bad press about the future).
Change is not linear and very slow and most big technological changes take 20 years to develop ("new technologies take 20 years to have an overnight success"). This means that you need good backsight, BUT because evolution is slow, you still have time even if you miss an early indicator.
Look then for early indicators ("prodromes" or "prodroma": an early symptom or leading indicator) as claimed by William Gibson's observation that the "future is already here, it's just not unevenly distributed". Look for indicators and things that don’t fit.
We tend to over-estimate the speed of short-term adoption and under-estimate the diffusion of the technology ("Never mistake a clear view for a short distance"). In addition, things aren’t accelerating and every society has always complained that things were getting faster, even in the 16th century ("every generation thinks things are accelerating").
Look at failures and cherish them (Preferably other people’s). Silicon Valley has been built on the ashes of failure. Look also for people who failed in a company and went starting their own.
Prove yourself wrong: look for indicators that proves what you say BUT also weak signals that prove it wrong
"Be indifferent. Don’t confuse the desired with the likely"
Know when not to make a forecast
The problem for forecasters is not of being wrong, it about persuading people to act on forecasts.