Predictify is a survey engine cleverly masquerading as a prediction market. On the site, users can answer questions that test their predictive abilities in various fields. If they prove to be right, they can win actual real money. Users even get a small payout for answering a question if they end up being wrong.
People and companies wanting to do market research can submit questions to the Predictify audience--for a fee--and the answers that prove to be most accurate split the bulk of "pot" that is attached to the question.
The money is ingenious misdirection. The point of rewarding accuracy is not to actually pay people for being right. The money is there, rather, to ensure that people who answer questions try to be right. Data from a Predictify survey is broken down in many ways for its users who pay for results, and it's that demographic breakdown that Predictify is really selling, not absolute predictions.
For example, suppose a company wants to know how to price a product. It can ask a question, "What do you think the price of this product will be when it hits the market?" The answers will be correlated with demographics, revealing what different groups (gender, age, ZIP code, etc.) think the item is worth. The "winners" who select the right price aren't predicting the price so much as determining it, and the people who select the price point are basing it not on the aggregate wisdom of the crowd but on the pricing level their target demographic has zeroed in on.
It's one of the cleverest Web 2.0 mind games I've seen in a while, and it just might work. Like many prediction markets, though, the community will run out of gas unless there's a strong incentive to keep people engaged. Money is just part of it on this system. Predictify also ranks users, and pays out more to the more accurate ones.
Users also get spiffed for referring their friends--they earn 10 percent of their friends' winnings for the first six months of their activity on the site. (It's not a multilevel scheme--friends of friends don't pay.)
People who pay for the wisdom of the Predictify crowd (a dollar an answer, right or wrong) own the data and can hide their identity. There are also free questions on the site. Anyone can ask them, but Predictify staff has to approve them before they're put online. Answers to these questions are public, and the "winners" don't make money from them. But correct answers do increase users' reputations on the service, which ultimately does have value, as explained above.
As an intellectual exercise, it's complicated, but the two boy-genius founders (Parker Barrile and Michael Agnich, both now in the Stanford Graduate School of Business after graduating from Stanford, respectively, in math and computer science) have done a good job of hiding the social and mathematical convolutions in the service. They've made it easy and somewhat fun to play, and certainly when there's the potential of earning some money by answering simple questions that only take a few seconds each, people are going to try a least a little to be right.
The site is in open beta right now and is very young--I counted only about 320 registered users. Growth will likely come from paying marketing departments, who will recruit users from their own sites. The founders hinted at such a deal being set up with an American car company.