Wired Pursuits

Instead of turning to paid analysts, or internal experts some companies are using the crowd to help improve the accuracy of demand forecasts as well as better manage inventory and manufacturing capacity.

Predictive markets, often also referred to as information markets, aggregate the knowledge of the crowd to make predictions regarding unknown future events. By aggregating distributed knowledge the predictions of the crowd are often more accurate than when companies rely on only a handful of experts.

How do they work?

In prediction markets, individuals buy and sell “futures” or “shares” based on their beliefs regarding the probability of the event taking place. If they are correct, they are rewarded for their efforts. Because of U.S. laws related to online gambling, rewards often take the form of play money, gift certificates, or recognition within the market.

What is interesting about prediction markets is their structure creates incentives for individuals to act on their closely held information. Because rewards are tied to correct predictions, individuals in the crowd who have access to more information, which may aid in their understanding of the market, tend to buy more shares than those who are just guessing.

For example, Google uses over 300 internal prediction markets to assess events such as customer demand for new products (“How many Gmail users will there be on January 1, 2009?”), company and product performance (“When will the first Android phone hit the market?”), and competitor performance (“How many iPhones will Apple sell in the first year?”). In addition to new sources of predictions, Google has used these prediction markets to better understand and improve the flow of information within their company.

Do predictive markets work?

As with many new uses of the crowd, we have only scant evidence regarding the effectiveness of prediction markets. Best Buy reports internal prediction markets designed to predict holiday sales of gift cards were 99.5% accurate compared to a 95% accuracy rate from external consultants. Intel also reports success with their internal “forecasting markets.”

Faced with the difficult task of predicting demand of products requiring lead times of months or even quarters, Intel found their internal markets were at least as accurate as official figures and in some cases more accurate by 20% (i.e., 20% less error).

There is, however, evidence that outside factors can impact results of these markets. Some data suggests employees may be overly optimistic regarding company performance tending to artificially inflate positive results. For publically traded companies, stock performance can influence predictions upward or downward in line with the latest market swings. And finally, there are also numerous issues related to providing the right incentives for participation as well as obtaining executive buy-in for such initiatives.