What if your boss, rather than dismissing you off hand when you suggested a different way of doing something, let you bet on the fact that your idea was better. And paid you if you were right.
That is the premise of prediction markets – a tool companies are increasingly using to make better decisions by allowing employees to trade in a mock stock market based on information they have about the business.
They work a little like a futures market. Say a company wants to release a new product but is unsure whether a launch date can be achieved. It creates a market for a prediction that the product will launch in a given month. Employees are then given a fixed amount of a mock currency they then use to trade contracts in that future.
Those who think it is likely the product will launch in that month will buy the future, and the price goes up. Those who don’t, sell, sending the price down.
At the end of the trading period, the value of the contracts that have been bought and sold are tallied, and those traders who have been most successful are given cash prizes or bonuses.
The company benefits since the final ‘price’ reflects a broadly based forecast of the likelihood that a given launch date can be achieved.
Several large American firms – including Google, GE, Hewlett-Packard, and Best Buy – have begun using prediction markets to help make assessments about all kinds of projections, including how customer numbers will grow, what demand will be for a product, and when it will launch.
Companies can also set up more complex trading systems which allow employees not only to trade in a given prediction but also to make their own estimates of when they think a product will launch, say, and trade contracts in that future.
Two recent reports – by McKinsey, the management consultancy, and Forrester, the analyst – have suggested that such markets are likely to grow in importance, particularly as the technology which facilitates them becomes cheaper and more widespread.
The advantages, exponents say, are many: the trading system gives employees an incentive to share information they have that may be valuable: although the currency is artificial, the cash rewards are real.
A market also aggregates knowledge more efficiently, as employees can – in effect – give feedback to their boss by trading more or less aggressively on information they have. The social and cultural issues which may have prevented an employee from sharing information are also, in theory, swept aside, because all trades are anonymous.
Companies that have used prediction markets say they provide more accurate information about aspects of their business than could be learnt by more traditional methods, such as polling employees or consulting outside experts.
In one instance, a malfunction in the furnace of a chip manufacturer led employees at the factory to update their positions on future yield more quickly than could employees elsewhere.
Google, the search giant, has been using such markets since 2005 to estimate – among other things – traffic volumes, and when its international offices will open.
About 1,500 employees will trade at any one time, and each has 10,000 ‘Goobles’ – a mock currency – to play with per quarter. The Goobles are issued weekly, “otherwise people would lock up their positions early and there’d be liquidity issues,” Bo Cowgill, an economic analyst at Google, said.
The markets run for anywhere between two weeks and three months. When they close, successful traders are given cash prizes and specially branded T-shirts which, Mr Cowgill said, end up being a greater incentive than the money.
At GE, the healthcare to aviation and media conglomerate, about 40 to 50 predictions – typically about the types of new technologies the company should invest in – are traded at any one time by up to 10,000 of its 330,000 employees.
“We use them as another point in the decision-making process, alongside asking experts and other business leaders,” said Christina LaComb, a computer scientist in the R&D lab at GE.
The window displaying the live trading system sits on the desktop of the employee’s PC. The employee sees the current asking price for any contract and its trading history, and there are simple buttons for buy and sell.
One problem is the inherent biases of such markets, which can make the information they provide less useful.
“The first is a long-shot bias – people tend to overesimate the likelihood of a long-shot paying off, so they tend to overpay to make that bet; conversely they tend to underpay for a sure thing,” said Jeff Severts, a VP of services at the US electronics giant Best Buy, which has used prediction markets for two years.
“The second bias is a home-team bias: employees have been overpaying for ’stocks’ that are based on outcomes that would be good for Best Buy, but we believe we can mitigate the biases through continued refinement of the market.”
The US Commodity Future Trading Commission is also examining prediction markets from a regulatory point of view, and companies which use them have steered away from running markets which predict financial results.
“If we ended up with an application that did an amazingly good job or predicting our results, there would be a concern that anyone who saw it would be an insider,” said Mr Cowgill.
Todd Henderson, an assistant professor at University of Chicago Law School who has written about prediction markets, said that, assuming prediction markets were approved by regulators, the case for using them was compelling, and that the “real puzzle” was why every firm didn’t.
“Companies are the members of society most comfortable with markets as processors of information, and yet when it comes to decision-making, they display these socialist ‘command and control’-style tactics to make them,” he said. “Twenty years from now, prediction markets will be ubiquitous.”
Companies pay anywhere from $25,000 to upwards of a $1 million to run a market, according to David Perry, president of Consensus Point, a business which implements web-based internal trading systems.
Consensus Point is the leader in prediction markets, the social predictive analytics solution.