Executive Briefings

Your Company's Future: Wanna Bet?

For some years now, betting types have been using something called a prediction market to divine the future. Players wager on a selected outcome - say, the dollar hitting a certain level against the yen 90 days from now. A bet takes the form of a contract which bears an initial price and can be bought or sold over its lifetime, then pays out a given amount over the purchase price, if the bettor turns out to be right. The mechanism has been used for everything from economic trends to disaster scenarios, Oscar winners and football scores. The basic idea is simple: the more people who participate, the more likely a prediction will be accurate. It's the old notion of the wisdom of crowds. ("Ask the audience," anyone?)

Now, a young software company believes that prediction markets will work in the everyday business world - particularly with regard to decisions on supply chain management. The company is Redwood City, Calif.-based Crowdcast (www.crowdcast.com). I spoke recently with chief executive officer and founder Mat Fogarty, and chief scientist Leslie Fine. Fogarty used to conduct corporate forecasting for Electronic Arts and Unilever. Fine spent eight years at Hewlett-Packard Labs, where she designed HP's internal forecasting mechanism known as BRAIN.

In concept, Crowdcast's software is very much like the technology that underlies such larger ventures as the Iowa Electronic Markets (http://www.biz.uiowa.edu/iem/index.cfm), Intrade (www.intrade.com) and the Hollywood Stock Exchange (www.hsx.com). They cover events from the monumental to the trivial. (How you would rate the Best Supporting Actor winner for 2009 is entirely up to you.) The question is, how well does a prediction market work on a smaller scale, one that's suited to the workings of a single company? "It's similar in that you can aggregate across the crowd," says Fogarty. "The difference is that we very much simplified the interface."

Traditional prediction markets don't ask the right questions for corporate use, Fine says. They often involve multiple choices and complex interactions. Crowdcast has stripped the system down considerably to make it work for business. Participants begin with a fixed amount of virtual "money," then are encouraged to place bets on various things that might happen to the organization within a given period of time. These might include a sales forecast, quality estimate, user reviews of a new product, regulatory decision, macroeconomic factors, even competitors' strategies. The bet might cover a specific target or, more commonly, a range of results. (The narrower one's prediction, the bigger the risk of being wrong - and the bigger the payout when you're right.) Winners are rewarded with points, prizes and sometimes actual cash. They also win bragging rights via their placement on a company leaderboard, although their true identities are concealed so that they'll feel free to bet honestly, without repercussions from other employees.

The exercise assumes that individuals at many levels of the organization have something valuable to contribute. "The beautiful thing about the tool is that it produces answers that are more accurate, because it takes more information in," says Fine.

"Secondly, it rewards accuracy and can undermine the biases that may exist in a [traditional] forecast."

All well and good - but does a corporate prediction market involve sufficient scale to yield meaningful results? Millions participate in markets that bet on political contests and Oscar races; the results are often more accurate than other means of prognostication (although bets placed on Intrade during the last U.S. presidential election yielded some disappointing results for prediction-market junkies). At HP, Fine did a lot of research into how big a pool of participants was needed to make a system work. The traditional stock market, which provides the template for all imitators, "needs millions of traders making billions of trades," she says. With Crowdcast, however, employees are usually entering a range of predictions, depending on how confident they are of being right. The result, she claims, is "a much richer expression of your point of view. We've had good success with 20 to 30 participants in a given market."

The model is especially valuable in the area of supply chain management, Fogarty claims. One obvious target is supplier stability. A group of employees might do a better job of identifying those vendors that are in danger of going under without warning, and snapping the chain. In that regard, a prediction market becomes a valuable risk-management tool. Crowdcast currently has no clients deploying the software for that purpose, Fogarty says, "but it would be a very good use of the tool."

Another candidate is the Holy Grail of supply chain management: collaborative sales and demand forecasting. In such cases, a prediction market might go outside a company's walls to include all of its relevant supply chain partners. That, of course, is the rationale behind many previous efforts, chief among them the Collaborative Planning, Forecasting and Replenishment (CPFR) process (http://www.vics.org/committees/cpfr/) that was spearheaded by Wal-Mart and others in the mid-1990s. (And has yielded variable results, depending on to whom you speak.) Under the old way of doing things, Fogarty says, individuals might have personal agendas that influence their voices, such as quarterly quotas for a sales force, or high production numbers for manufacturing. With a prediction market, the chief incentive is to be right. In addition to prizes and the ego boost that come from a winning record, the individual with the best track record gains a greater ability to influence forecasts. Meanwhile, the "many voices" approach can serve to flush out the kind of misinformation that tends to reside in various pockets of the supply chain, leading to the dreaded "bullwhip effect."

Prediction markets won't work in every part of the organization. They're a bad idea for such things as total revenue forecasts and workforce headcount (the latter for obvious reasons - who would bet on the loss of his or her job?). Merger-and-acquisition activity is another area where the idea isn't feasible. "There are some things that you want to keep confidential," says Fogarty.

Nevertheless, Crowdcast, which is funded by venture capitalists, is placing its own kind of bet: that companies will embrace a hosted software application that takes a brand new approach to the way that corporate decisions are made.

"We're in the early stages," says Fogarty. "We've got to get the technology right, and we've got to get the engagement right. We see a real future in this."

- Robert J. Bowman

Comment on This Article

For some years now, betting types have been using something called a prediction market to divine the future. Players wager on a selected outcome - say, the dollar hitting a certain level against the yen 90 days from now. A bet takes the form of a contract which bears an initial price and can be bought or sold over its lifetime, then pays out a given amount over the purchase price, if the bettor turns out to be right. The mechanism has been used for everything from economic trends to disaster scenarios, Oscar winners and football scores. The basic idea is simple: the more people who participate, the more likely a prediction will be accurate. It's the old notion of the wisdom of crowds. ("Ask the audience," anyone?)

Now, a young software company believes that prediction markets will work in the everyday business world - particularly with regard to decisions on supply chain management. The company is Redwood City, Calif.-based Crowdcast (www.crowdcast.com). I spoke recently with chief executive officer and founder Mat Fogarty, and chief scientist Leslie Fine. Fogarty used to conduct corporate forecasting for Electronic Arts and Unilever. Fine spent eight years at Hewlett-Packard Labs, where she designed HP's internal forecasting mechanism known as BRAIN.

In concept, Crowdcast's software is very much like the technology that underlies such larger ventures as the Iowa Electronic Markets (http://www.biz.uiowa.edu/iem/index.cfm), Intrade (www.intrade.com) and the Hollywood Stock Exchange (www.hsx.com). They cover events from the monumental to the trivial. (How you would rate the Best Supporting Actor winner for 2009 is entirely up to you.) The question is, how well does a prediction market work on a smaller scale, one that's suited to the workings of a single company? "It's similar in that you can aggregate across the crowd," says Fogarty. "The difference is that we very much simplified the interface."

Traditional prediction markets don't ask the right questions for corporate use, Fine says. They often involve multiple choices and complex interactions. Crowdcast has stripped the system down considerably to make it work for business. Participants begin with a fixed amount of virtual "money," then are encouraged to place bets on various things that might happen to the organization within a given period of time. These might include a sales forecast, quality estimate, user reviews of a new product, regulatory decision, macroeconomic factors, even competitors' strategies. The bet might cover a specific target or, more commonly, a range of results. (The narrower one's prediction, the bigger the risk of being wrong - and the bigger the payout when you're right.) Winners are rewarded with points, prizes and sometimes actual cash. They also win bragging rights via their placement on a company leaderboard, although their true identities are concealed so that they'll feel free to bet honestly, without repercussions from other employees.

The exercise assumes that individuals at many levels of the organization have something valuable to contribute. "The beautiful thing about the tool is that it produces answers that are more accurate, because it takes more information in," says Fine.

"Secondly, it rewards accuracy and can undermine the biases that may exist in a [traditional] forecast."

All well and good - but does a corporate prediction market involve sufficient scale to yield meaningful results? Millions participate in markets that bet on political contests and Oscar races; the results are often more accurate than other means of prognostication (although bets placed on Intrade during the last U.S. presidential election yielded some disappointing results for prediction-market junkies). At HP, Fine did a lot of research into how big a pool of participants was needed to make a system work. The traditional stock market, which provides the template for all imitators, "needs millions of traders making billions of trades," she says. With Crowdcast, however, employees are usually entering a range of predictions, depending on how confident they are of being right. The result, she claims, is "a much richer expression of your point of view. We've had good success with 20 to 30 participants in a given market."

The model is especially valuable in the area of supply chain management, Fogarty claims. One obvious target is supplier stability. A group of employees might do a better job of identifying those vendors that are in danger of going under without warning, and snapping the chain. In that regard, a prediction market becomes a valuable risk-management tool. Crowdcast currently has no clients deploying the software for that purpose, Fogarty says, "but it would be a very good use of the tool."

Another candidate is the Holy Grail of supply chain management: collaborative sales and demand forecasting. In such cases, a prediction market might go outside a company's walls to include all of its relevant supply chain partners. That, of course, is the rationale behind many previous efforts, chief among them the Collaborative Planning, Forecasting and Replenishment (CPFR) process (http://www.vics.org/committees/cpfr/) that was spearheaded by Wal-Mart and others in the mid-1990s. (And has yielded variable results, depending on to whom you speak.) Under the old way of doing things, Fogarty says, individuals might have personal agendas that influence their voices, such as quarterly quotas for a sales force, or high production numbers for manufacturing. With a prediction market, the chief incentive is to be right. In addition to prizes and the ego boost that come from a winning record, the individual with the best track record gains a greater ability to influence forecasts. Meanwhile, the "many voices" approach can serve to flush out the kind of misinformation that tends to reside in various pockets of the supply chain, leading to the dreaded "bullwhip effect."

Prediction markets won't work in every part of the organization. They're a bad idea for such things as total revenue forecasts and workforce headcount (the latter for obvious reasons - who would bet on the loss of his or her job?). Merger-and-acquisition activity is another area where the idea isn't feasible. "There are some things that you want to keep confidential," says Fogarty.

Nevertheless, Crowdcast, which is funded by venture capitalists, is placing its own kind of bet: that companies will embrace a hosted software application that takes a brand new approach to the way that corporate decisions are made.

"We're in the early stages," says Fogarty. "We've got to get the technology right, and we've got to get the engagement right. We see a real future in this."

- Robert J. Bowman

Comment on This Article