In a three-week experiment, GE Research turned its 85 employees into day traders, letting them watch market movements on their screens to decide whether to buy or sell any of 62 stocks. Only the stocks were product ideas in which the company had the option to develop. At stake was $50,000 in research funding that GE would allocate to the highest-valued project.
When the markets closed, GE ended up with a prioritized list of ideas that the collective wisdom of the market thought would best help the company. Topping the list was an algorithm in the area of machine intelligence, an idea pitched directly by a researcher, not through the normal hierarchy of lab managers and senior management.
Dell looked to an even broader market for new product ideas, using Salesforce.com's online voting service called Ideas and launching Dell IdeaStorm, where anyone can submit and vote on new features and options for Dell products. Perhaps best known of these ideas is a Linux-based laptop Dell introduced in May 2007. Starbucks uses the same voting platform, at MyStarbucksIdea.com, and took an online suggestion posted Oct. 7 by BillMac to offer a free cup of coffee Nov. 4 to anyone in the United States who voted.
The use of these collective decision-making technologies, both sophisticated prediction markets and simple voting tools, is spreading, and they're increasingly being paired together as a component of corporate innovation programs, helping companies sort through reams of ideas--from new products to customer service to productivity improvements--to find that handful of blockbusters.
Few things matter more to a company. Think of the impact a single product, whether the iPod or New Coke, can have on a company's fortunes. IT needs to make itself part of that process, and one way is by providing tools to help their companies make better decisions. Like blogs, wikis, and other social software, these tools tap into a free exchange of ideas. Unlike other social software, they lead to a definitive outcome and measurable results.
Still, prediction markets aren't money in the bank. That GE Research experiment that pushed the algorithm idea to No. 1? That was back in 2005. And while the group has run nine markets in all, including one for GE Healthcare that led to its filing for patents, it's still evaluating a product from Consensus Point, and it's not an everyday part of its innovation process. The algorithm itself was basic research, not a product.
Dell remains a believer in the community's intelligence after more than a year of using the voting technology and thinks that of the 200 or so ideas it has implemented out of the process, 4% are "potential game changers," says Bob Pearson, Dell's VP of communities and conversations.
Markets have proved their worth in predictions. Hewlett-Packard has used faux markets to predict the cost of DRAM and found them to beat official HP forecasts six out of eight times. (It's now marketing its own prediction product.) One major software company says prediction markets accurately predict project completion times as often as 95% of the time. In The Wisdom Of Crowds (Anchor, 2005), James Surowiecki laid out four conditions under which the crowd tends to be more accurate than experts: a diverse population; a decentralized population, so no one dictates an answer; an independent population, so voters focus on what they know, not what others think; and a summary of opinions into one verdict.
The more complicated approach is to create markets, making each idea a stock, which players buy and sell to accumulate as much virtual cash as they can. Players are likely to give the decision more thought than when simply voting, since they're trying to win, not just throw out an opinion. And since participants choose how much to invest, the price reflects intensity of expectations, providing a better projection of a given outcome (see table, "Vendors, And Some Questions To Ask", below).