Prediction markets are known to forecast the outcome of elections more accurately than polls. Now Google is using an internal prediction market to improve business decisions and learn how employees exchange information.
Basically operating like a stock market, Google’s prediction system was designed by its own software engineers and lets employees bet on probable outcomes. Will a project be finished on time? How many users will Gmail have? Who will win the World Series?
With about 1,500 employees making 80,000 “trades” over the past two and a half years, Google officials say it’s the largest corporate experiment with prediction markets they are aware of.
Google employees bet “Goobles” instead of real currency, but at the end of each quarter the most successful bettors can win $1,000 cash prizes or T-shirts.
More importantly, said Google economic analyst Bo Cowgill, the trading system lets the Google hierarchy discover its employees’ uncensored opinions.
“If you let people bet on things anonymously, they will tell you what they really believe because they have money at stake,” Cowgill said. “This is a conversation that’s happening without politics. Nobody knows who each other is, and nobody has any incentive to kiss up.”
Cowgill described the prediction market Tuesday at the O’Reilly ETech conference on emerging technology in San Diego. Cowgill and economists from Dartmouth College and the University of Pennsylvania also published a paper on Google’s experiment in January. From a practical standpoint, the Google prediction market gives managers insights that might impact business decisions and which they might not have obtained in any other way. In one example, “the market was predicting that [a project] was behind,” Cowgill said. “A manager says what’s going on here, . . . starts investigating and finds some glitches.”
Analysis of the markets found a distinct pro-Google bias, however. Cowgill separated outcomes into the categories “good for Google” and “bad for Google,” excepting such unrelated prediction markets as “will Harriet Miers be confirmed?” — and found that people betting on “bad for Google” were able to make a killing because traders bet too heavily on the good outcomes.
People were even more optimistic about specific events related to Google on days when the company’s stock price went up, even if the increase had nothing to do with the events being wagered on.
In addition to discovering biases, Cowgill was able to figure out which types of relationships affect the way someone bets, and, in turn, chart how information flows through the organization.
Although users are anonymous to each other, Cowgill and fellow researchers can access tons of details on specific users, including where they sit.
“We have GPS coordinates of every desk in every building,” Cowgill said.
Google looked at all sorts of relationships — people who work for the same boss, who went to the same school, who work on the same projects, who review one another’s software code, who are on the same e-mail threads, who speak the same language.
One factor stood out above all others: physical proximity. Googlers who sit near each other bet alike.
“The distance measured in feet was a good predictor of whether people will be trading alike,” Cowgill said. This reinforcement of the value of human contact “is an ironic finding for us to have at Google, where we’re selling products that help you overcome distance and collaborate without having to be nearby each other.”
The association between physical proximity and trading patterns was strong no matter how Cowgill and colleagues crunched the numbers. People who sit near each other trade alike even if they don’t work on the same projects. People who work on the same projects but don’t sit near each other do not trade alike.
When Googlers move to a new office or desk — as they do about once every 90 days — they begin trading like nearby employees within about six weeks. Information exchanges obviously are happening within a few feet of employees’ desks.
Google traditionally liked to seat people as close together as possible, but had been moving away from that ideal a little bit, Cowgill said. The prediction markets confirmed that placing employees near one another was the right way to go.
As a philosophical matter, Google likes to “pack people in tight . . . so they can share information,” Cowgill said. “As a company gets larger, people don’t always adhere to the founding tenets, or if they’re in a big building they’ll spread out because it’s more comfortable. This was something where we could say ‘there’s value and we can measure that, and we can compare it to what it’s like when people e-mail.'”