Harvard Business School Working Knowledg e Archive

You Can Bet on Idea Markets

12/1/2003
Forget the brouhaha over the Pentagon's futures market on terrorism. By gathering collective wisdom, idea markets can improve your forecasting, knowledge management, and decision making.

In July 2003, idea markets made an unexpected appearance in the headlines as the Pentagon was taken to task for its plan to use one to help pinpoint potential terrorism targets. The market would have allowed participants to trade opinions on where terrorists were most likely to strike in much the same way as one trades securities or commodities in other markets. Officials believed they could learn something worthwhile from the relative value that traders placed on different targets.

Unfortunately, these plans did not sit well with many lawmakers, who saw a market built around terrorist strikes as highly unseemly and a means to provide ideas to terrorists. And so the project was scrapped. Nevertheless, the Pentagon's plans show just how prominent idea markets are becoming in real-world applications. For a more fully developed example, consider how Hewlett-Packard has put this concept to work.

Hewlett-Packard needed a better way to forecast sales. The firm's traditional methods of gauging market conditions, analyzing the competitive landscape, and looking at sales trends were simply not up to the task in a rapidly changing environment. So HP executives decided to try an experiment. They would turn to their own sales force to predict the numbers in a new way.

But first there were some obstacles to overcome. Salespeople usually know a lot about their customers and those customers' likely purchases, but given a commission- and bonus-based compensation system, they have strong incentives to underestimate forecasts in order to exceed quarterly targets. To help the sales team really focus on making accurate predictions—and to give them an incentive for doing so—HP began using an idea market where account executives could trade in sales projections.

Idea markets have been used effectively outside the corporate world for more than fifteen years.

Charles Plott, an economist at the California Institute of Technology (Pasadena), and Kay-Yut Chen, a senior scientist in the Software Technology Lab at HP Labs, created 10 forecast markets, each of which represented one of 10 different sales scenarios—0 to 1,500 units sold, 1,501 to 1,600 units sold, and so on. Each sales representative was then given 20 shares per market to trade, buy, and sell according to their beliefs about the likely level of sales for a particular month. When the final sales figures were posted, those who owned shares in the stock whose face value (e.g., 1,501 to 1,600 units) reflected the actual sales (e.g., 1,550 units) were paid $1 per share.

As share prices in the HP market ranged from 0 to 100 cents, the final trading price represented the collective estimate of the probability a particular event will occur. Thus, if the SEP LOW 0–1,500 units market had a closing price of 5 cents per share, that was interpreted as a 5 percent probability that sales would be in the 0–1,500 range the following September. HP's idea market beat traditional forecasting methods in 15 of 16 forecasting trials.

Elections to Oscars to widgets
Idea markets have been used effectively outside the corporate world for more than fifteen years. Since 1988, the Iowa Electronic Markets have predicted presidential election outcomes more accurately than traditional polls 75 percent of the time. Similarly, the Hollywood Stock Exchange has forecast Oscar winners more effectively than even the most seasoned media critics seven years running.

Today, more and more companies are taking note of these results. Newsfutures.com now hosts a number of pharmaceutical industry-sponsored sections to forecast and aggregate beliefs about key issues, from legislation for prescription benefits to the number of new molecular entities that the FDA will approve in the current year. Of course, this idea market is run by a third party and only offers an industry-level view of how managers should make decisions.

As the Hewlett-Packard case demonstrates, however, companies can—and sometimes must—build an idea market internally to gain a more tailored prediction capability. Here are three steps that managers need to take to put an idea market into organizational practice:

Step 1: Tap into strategically important but difficult-to-measure customer behaviors

Take new product development, a process usually fraught with uncertainty. Will customers like the new idea? Will it be better than a competitor's new product? Now, imagine that a firm could solve these problems using markets, which may be twice as effective at predicting customer preferences as traditional tools like the focus groups, surveys, and conjoint analysis.

Ely Dahan at UCLA's Anderson School of Business, along with Andrew Lo and Tomaso Poggio at MIT, are building markets tailored to meet just this type of business need. Dahan's team uses an idea market where a panel of consumers trade in artificial stocks that represent each product. Consider a car company planning to introduce a new crossover model two years from now. The car will compete with five existing models and five new models. In Dahan's market, each of the eleven models is represented by a stock. Consumer traders are given a specified amount of artificial dollars (say $10,000) and equal endowments (say 100 units) of each stock. They are also provided with information about each model and a picture of its likely appearance.

The Hewlett-Packard experiment with idea markets shows that they can help solve enduring managerial problems.

Traders are then asked to buy and sell the stocks online with one another to maximize the value of their portfolios, based on their beliefs about which models will sell well. The final stock prices at the end of an hour-long trading period represent the best estimates by the group for the consumer's relative preferences for each type of vehicle. This idea market, which was originally run at MIT, tests for consumer preferences in product and service categories ranging from bicycle pumps to ski resorts. It is consistently more effective in identifying and aggregating beliefs about customer preferences than traditional measures.

Step 2: Unlock knowledge to tackle organization-wide challenges

The Hewlett-Packard experiment with idea markets shows that they can help solve enduring managerial problems. Similarly, at British Petroleum, managers built markets to improve decision making, a particularly thorny challenge in this global firm. In 1998, BP announced that by 2010 it wanted to reduce its greenhouse gas emissions by 10 percent from its 1990 levels. As a large decentralized company with more than 150 separate businesses, finding the cheapest way to reduce emissions was no easy task. Managers of the individual units were in the best position to know how to reduce emissions efficiently and economically, but costs for reductions could vary dramatically by unit.

The company turned to an Internet-based electronic market to aggregate decentralized knowledge, and trade emission rights internally to efficiently find the best ways to lower emissions. Each business unit was first given an emissions reduction target. Next, the units were allowed to invest in emissions reduction directly or to purchase reductions in emissions from other business units that were able to exceed reduction targets more cheaply. By trading emission rights online, British Petroleum was able to find the best way to allocate corporate resources and even discover the real costs of reducing emissions. For example, in March 2001 emission rights traded at close to $20 per metric ton.

As a real-time and dynamic polling system, idea markets can play a useful role in eliciting and aggregating the beliefs of individuals within an organization or a broader network. But these markets can also generate valuable signals to decision makers. When idea market prices move dramatically, decision makers should ask, Why is the price moving? What is changing in the beliefs of participants? Market makers can build message and discussion boards into systems to reveal trader beliefs and other key pieces of information.

In short, idea markets can help address the vexing knowledge management problem of just how to effectively elicit beliefs.

Step 3: Exploit markets to gain buy-in from customers and managers

Effective forecasting of winning product concepts or organizational resources isn't the only advantage to implementing idea markets. UCLA's Dahan reports that participants intuitively understand the concept of trading and winning. Many even find it fun—especially relative to filling out a survey or answering questions from a focus group facilitator.

Survey methods often require hundreds of participants, but idea markets can work with as few as twenty to thirty participants. Managers who want to implement idea markets can therefore sell them as a cheaper and more nimble way to gauge customer preferences and employee insights than traditional approaches.

Finally, markets are dynamic and allow participants to adjust their beliefs based on price and the trading actions of one another revealed in the markets. This is especially important when it comes to "status goods," where people form their preferences based on the actions of others—a factor that traditional measures miss.

Idea markets—from concept to creation
Once an academic toy, idea markets are increasingly becoming a practicable business tool. Today software for idea markets is available from several vendors, and some firms and researchers have substantial experience in designing and running these markets. Corporations in Europe and the U.S. are beginning to apply these markets, from forecasting industry trends to picking merchandise categories that will be most popular with consumers.

In time, idea markets are likely to play a vital role in evaluating uncertain investments in R&D, new technology ventures, patents, and other intangible assets. But idea markets should be only one tool and input to decision making. Idea markets are not silver bullets—their outcomes ultimately only reflect the best social consensus and aggregation of beliefs at any one time.

Reprinted with permission from "Betting on a New Market," Trends and Ideas, Volume 1, Number 1, Fall 2003.

Ajit Kambil is the global director for technology and innovation at Deloitte Research and a Distinguished Scholar in Residence at Babson College. He is the coauthor, with Eric van Heck, of Making Markets: How Firms Can Design and Profit from Online Auctions and Exchanges. He can be reached at trends@hbsp.harvard.edu.