Prediction Markets – using them for innovation and ideas
May 5th, 2008If you haven’t read it, the New York Times recently published a good article on prediction markets. It covered the usual topics and players, with quotes from Robin Hanson, Emile Servan-Schreiber, and discussion about Inkling Markets and Xpree. More importantly the author had some great quotes from Jeff Severts at Best Buy and Zubin Dowlaty of InterContinental Hotels about their experiences with prediction markets. (Though not mentioned, they use ConsensusPoint and NewsFutures software, respectively.)
For anyone that’s read this blog, there is likely little new in the article that you haven’t heard before. It’s great that it was focused more toward corporations and how companies use them, and I hope that a steady drumbeat of articles like this will help generate the momentum for more widespread acceptance and use of prediction markets.
But I want to focus on one aspect of the article; using prediction markets for innovation. I believe that while you can use prediction markets to help quantify innovative ideas, using PM’s to prioritise ideas is a poor use of the tool.
Prediction markets exist to forecast a variable; this may be a metric (such as sales figures), or a probability (which candidate will win the election). For markets to work effectively, the variable should be well-defined, such as US sales in Q1, will the Minneapolis office open on schedule, etc. By their very nature, ideas tend to be quite poorly-defined. Before an idea is ever put into place, significant work takes place to develop it within a company.
When companies start using prediction markets to “harvest and prioritize ideas” I believe they are using a prediction market incorrectly. The traditional prediction market is run on a variable that’s both well-defined and committed. (You are going to continue selling a particular widget; the project to open the Minneapolis office is going to happen.) However, running prediction markets to assess ideas that may change before implementation or may not happen at all turns the entire process into a simple beauty pageant. A beauty pageant doesn’t find the most beautiful girl, it finds the girl that the particular judges on the day find beautiful.
The New York Times had an interesting article recently that discussed this issue from a different tack. “Is Justin Timberlake a Product of Cumulative Advantage” discussed a study which tracked how identical songs become (or didn’t become) popular in separate social groups. Similar principles easily translate to prediction markets on ideas. In a phrase, this is the problem:
social influence played as large a role in determining the market share [price] of successful songs [ideas] as differences in quality.
So what can you do to harvest collective intelligence to generate, prioritise and quantify ideas?
Go out there, generate some great ideas in your company. But don’t use a prediction market to prioritise them; social influence plays too big a role since the market is (at that point) too ill-defined and it hasn’t been committed to by the company. Prioritise them somewhere else.
Once the list of ideas has been fully developed vetted down to a handful of well-defined ideas, then a prediction market could be useful to quantify the opportunity. At this point, traders would know enough about the opportunity to judge the potential outcome of putting them into practice, and the outcome should be reasonably accurate.
To be fair to InterContinental Hotels (which is quoted as using prediction markets for this), they are actually a NewsFutures customer and use their Idea Pageant software, instead of the normal Prediction Market software. (The Idea Pageant seems to be an effective way of doing the non-prediction-market prioritising that I mentioned above.) However, there are a number of companies that do use standard prediction markets to prioritise new ideas.
If you’re using prediction markets for this, I would suggest you stop. As discussed in the New York Times article above, the first-movers and influencers are very influential when assessing new ideas. When an idea is new and ill-defined, people put much higher priority on information from their social network instead of their own impressions. If you took the same set of initial ideas to a completely different group of people in your company you may get a very different set of results. This means that you’re not judging what people in the group think collectively, you’re measuring what that particular group thinks.
