The End Of Management?

(2 of 4)

Eli Lilly, one of the largest pharmaceutical companies in the world, which routinely places multimillion-dollar bets on drug candidates that face overwhelming odds of failure, wanted to see if it could get a better idea of which compounds would succeed. So last year Lilly ran an experiment in which about 50 employees involved in drug development — chemists, biologists, project managers — traded six mock drug candidates through an internal market. "We wanted to look at the way scattered bits of information are processed in the course of drug development," says Alpheus Bingham, vice president for Lilly Research Laboratories strategy. The market brought together all the information, from toxicology reports to clinical results, and correctly predicted the three most successful drugs.

What's more, the market data revealed shades of opinion that never would have shown up if the traders were, say, responding to a poll. A willingness to pay $70 for a particular drug showed greater confidence than a bid at $60, a spread that wouldn't show if you simply asked, Will this drug succeed? "When we start trading stock, and I try buying your stock cheaper and cheaper, it forces us to a way of agreeing that never really occurs in any other kind of conversation," says Bingham. "That is the power of the market."

The current enthusiasm can be traced in part, oddly enough, to last summer's high-profile flop of a market that was supposed to help predict future terrorist attacks. A public backlash killed that Pentagon project a few months before its debut, but not before the media broadcast the notion that useful information embedded within a group of people could be drawn out and organized via a marketplace. Says George Mason's Hanson, who helped design the market: "People noticed." Another predictive market, the Iowa Electronic Markets at the University of Iowa, has been around since 1988. That bourse has accepted up to $500 from anyone wanting to wager on election results. Players buy and sell outcomes: Is Kerry a win or Bush a shoo-in? This is the same information that news organizations and pollsters chase in the run-up to election night. Yet Iowa outperforms them 75% of the time.

Inspired by such results, researchers at Microsoft started running trials of predictive markets in February, finding the system inexpensive to set up. Now they're shopping around for the market's first real use. An early candidate: predicting how long it will take software testers to adopt a new piece of technology. Todd Proebsting, who is spearheading the initiative, explains, "If the market says they're going to be behind schedule, executives can ask, What does the market know that we don't know?" Another option: predicting how many patches, or corrections, will be issued in the first six months of using a new piece of software. "The pilots worked great, but we had little to compare it to," he says. "You can reason that this would do a good job. But what you really want to show is that this works better than the alternative."