The New York Times
Printer Friendly Format Sponsored By

April 9, 2008

Betting to Improve the Odds


CORPORATIONS live and die by ideas, and many enterprises have used Web-based technologies, like blogs, wikis and social networks, to gather thoughts and hasten their way into new services, products and cost-saving steps.

Now executives say they are harnessing a new Web tool, called prediction markets, to transform the idea pipelines inside their companies. Companies like the InterContinental Hotels Group, General Electric and Hewlett-Packard are using prediction markets to try to improve forecasting, reduce risk and accelerate innovation by tapping into the collective wisdom of the work force.

Like blogs and wikis, prediction markets can spur communication and collaboration within a company. Yet they add rigorous measurement to business forecasts, like estimating the sales of a new product or the chances that a project will be finished on time.

Corporate prediction markets work like this: Employees, and potentially outsiders, make their wagers over the Internet using virtual currency, betting anonymously. They bet on what they think will actually happen, not what they hope will happen or what the boss wants. The payoff for the most accurate players is typically a modest prize, cash or an iPod.

The early results are encouraging. “The potential is that prediction markets may be the thing that enables a big company to act more like a small, nimble company again,” said Jeffrey Severts, a vice president who oversees prediction markets at Best Buy, the electronics retailer.

The store chain has experimented with prediction markets on everything from demand for digital set-top boxes to store-opening dates. For example, Mr. Severts said that in the fall of 2006, the prices in a prediction market on whether a new store in Shanghai would open on time — in December 2006 — dropped sharply from $80 a share into the $40 to $50 range. Players made yes-no bets, and the virtual dollar drop reflected increasing doubt that the store would open on time.

Indeed, Best Buy’s first store in China opened late, in January 2007, but the warning signs from the prediction market helped prevent further slippage.

Mr. Severts noted that prices in a current prediction market — betting whether new offerings from its Geek Squad service will be introduced on time in June — are in the $90 range, an encouraging sign.

Best Buy plans to move beyond pilot projects in prediction markets to involve more workers throughout the company, starting next month. “It helps on two fronts, the speed and accuracy of information, so that management can move faster to deal with problems or exploit opportunities,” Mr. Severts said.

For years, public prediction markets have been used for politics, like the Iowa Electronic Markets and Intrade, where buyers and sellers bet on which candidate will win a particular race. And there are prediction markets where people place bets on news events (Hubdub, among others), video game sales (simExchange) or movie box-office receipts (Hollywood Stock Exchange).

These markets have often been more accurate than professional pollsters or market researchers. The idea is that the collected knowledge of many people, each with a different perspective, will almost surely be more accurate than an individual or small group or even experts. The concept has been championed by academic economists and was popularized by James Surowiecki’s 2004 book “The Wisdom of Crowds.”

Robin D. Hanson, an economist at George Mason University, proposes a “futarchy,” a form of government enhanced by prediction markets. Voters would decide broad goals of national welfare, but betting in speculative markets would determine the policy steps to achieve those goals.

Few in the corporate world go that far. An important issue is whether prediction markets are mainly an innovative way to gather information from employees or a font of reliable answers. “It’s still an open question whether the wisdom of crowds is really wise,” said John Kao, a consultant and the author of “Innovation Nation.”

So far, most of the companies using prediction markets are doing so in limited ways, in one or two departments, testing the concept to see how it goes. But in the last few years, corporate experimentation has moved beyond high-tech businesses into other industries, including retailing, consumer packaged foods, hotels, health care, steelmaking and telecommunications.

Today, analysts say, there are dozens of major corporations testing these markets. The companies include Google, Cisco Systems, GE Healthcare, General Mills, ArcelorMittal, the world’s largest steelmaker, and Swisscom, a large telecommunications company.

At the same time, a network of software and service suppliers is developing to cater to corporate prediction markets. The vendors include H.P. and smaller specialist firms like Consensus Point and NewsFutures. The field is attracting start-ups as well.

In 2006, for example, Adam Siegel and Nate Kontny left Accenture, the consulting firm, to found Inkling Markets. Mat Fogarty had been director for financial planning at Electronic Arts, where he experimented with prediction markets, before founding Xpree last year.

“Prediction markets are starting to move into the mainstream, and they will really change the way companies are run in the future,” said Emile Servan-Schreiber, the chief executive of NewsFutures.

At InterContinental Hotels, Zubin Dowlaty, vice president for emerging technologies, decided to create an online market last fall to “harvest and prioritize ideas” from within the hotel’s 1,000-person technology staff. “We wanted to tap the creative class that may not be able to voice their ideas,” Mr. Dowlaty said.

With InterContinental’s prediction market, players were asked to submit ideas anonymously, with a description and the benefit to customers and company. The bettors were given virtual tokens, each receiving 10 green ones to be placed on the best ideas and three red for bad ideas.

There were no limits on the number of times bettors could change their wagers as new ideas came to market, and the market was open for four weeks. The five top ideas (most green tokens), five bottom ideas (most red) and the top five bettors (most accurate, according to market consensus) were listed regularly.

The winners got $500, while second- and third-place finishers received $250 each. The winners, Mr. Dowlaty said, were engineers, analysts and contractors, not managers.

More than 200 people participated, submitting 85 ideas. One person proposed bringing back quarter-operated vibrating beds. “That one got beat down really fast,” Mr. Dowlaty said.

The winning ideas were suggestions to improve searching the company’s Web site to find and book hotel rooms. Two projects have been started as a result of the market, Mr. Dowlaty said.

Next, he said, prediction markets may be opened up to InterContinental’s customers, probably beginning with members of its Priority Club loyalty program. They could bet in markets for improving service and offerings, with points redeemed. “It’s the next frontier and the natural progression for this,” Mr. Dowlaty said.

Setting up corporate prediction markets can be tricky. Public markets for presidential candidates will attract thousands of bettors, but a company may want to run a market only for people with expertise in a certain product or project. At Hewlett-Packard, researchers have been working on techniques and software to make even small prediction markets efficient.

“We want to reduce the wisdom of crowds to the wisdom of 12 or 13 people,” said Bernardo A. Huberman, director of the social computing lab at Hewlett-Packard. Among the techniques, he said, are preliminary tests to assess the “behavioral risk characteristics” of participants to shade predictions from people who are inherently risk seekers or risk averse.

Starting a year ago, a group in the purchasing unit at H.P. began prediction markets on the price of computer memory chips three and six months ahead. The prediction markets, Dr. Huberman said, were up to 70 percent more accurate than the company’s traditional forecasting models. The more accurate predictions, he said, can be used to finesse purchasing, marketing and product pricing decisions.

The H.P. research project has become a service offering called Brain, for Behaviorally Robust Aggregation of Information in Networks. The service is now used in pilot projects by H.P. clients that include Swisscom, which is trying it to predict demand for new services like Internet television on cellphones.

At GE Healthcare, Steven Linthicum, manager for advanced prototyping and innovation, has recently managed prediction markets projects that have generated ideas for software products for the hospital and health care market. Based on that work, patents have been filed and projects are under way.

“These markets bring not only ideas, but what your organization thinks of the ideas,” Mr. Linthicum said. “That’s what leadership needs to know.”

G.E. is also evaluating how broadly prediction markets could be used in the health care division, a $17 billion-a-year unit, and elsewhere in the company.

“We’ll know a lot more at the end of the year how much this becomes part of the decision-making process,” Mr. Linthicum said. “We’re at the crossroads right now.”