Markets are predictive. Securities are valued by the market with an eye to future growth potential. Discounted cash flow analysis puts a risk rate on potential future earnings over time, discounting it to arrive at a present value to ascertain the investment potential of companies.
[This doesn't mean they are always right - it just means they reflect what the collective considers to be the future. However, markets make actions visible, and thus suffer from information cascades, which can cause bubbles.]
In the Wisdom of Crowds, James Surowiecki demonstrated that, under certain key conditions [including independence, which means the actions can't easily influence each other, which helps prevent information cascades] groups of people working individually for their own interests are very good at predicting how many beans are in jars and how much bulls weigh and lots of other things that you can assign values to.
Prediction markets are a way to leverage this ability: they are speculative markets built to make predictions - the Hollywood Stock Exchange is one of the most famous and successful.
IQ has written a whitepaper which looks how we might apply prediction markets to advertising pre-testing.
[His Long Tail of Branding paper won the WPP Atticus Award, so you know Sorrell digs it.]
The paper has a great round up of the current cristicims of advertising pre-testing and offers an alternative which would be effective in certain contexts, depending on the objective of the research.
We are much better at understanding others than we are at understanding oursevles.