MediaPredict: A Market For Predicting Media Hits


This is an interesting idea. has more information about the company.

GUT instinct plays too large a part in the publishing industry's decisions, says Mark Gompertz, vice-president of Touchstone, an imprint of Simon & Schuster, a publisher based in New York. So Touchstone is trying a new approach: a service concocted by Media Predict, a start-up based in New York. It uses the internet to obtain editorial feedback from a large number of volunteers in order to help executives decide which manuscripts should become books.

Media Predict hopes to do this using a virtual stockmarket for unpublished books, unsigned music acts and proposed television shows. It opened last week at Artists or their agents post samples of their work (a book chapter, say, or a television pilot). Traders armed with a free wad of virtual cash buy shares in the material they feel has the greatest potential. The idea is that as traders buy and sell shares in competing content, the cream will float to the top-where entertainment-industry bosses can skim it off. In September Touchstone plans to choose one or more of the top 50 book manuscripts on offer for publication.

Can the wisdom of crowds, or more accurately, a trading market like this, predict who will get record contracts and how much movies will gross? The answer is maybe.

I've done a lot of thinking about wisdom of crowds over the years, since the days of TBE, and I think in my analysis I've left out a key component – the distribution function of the underlying issue. Is it a bell curve? Is it a power law? The former may benefit from wisdom of crowds, if all other necessary components exist (aggregation of independent decisions, shared context, etc). The latter can never benefit from a wisdom of crowds process because votes on opposite sides don't cancel each other out.

For something like Mediapredict, the trading market may accurately predict yes/no decisions like whether or not an author gets a contract. But I think it will fail at issues like predicting how much money that same book will gross. Why? Because the distance between a poor selling book and an average selling book is very different than the distance between an average selling book and a best selling book. The median and the mean for book success are far apart.

Mediapost may have some high profile successes, but eventually they will have to face some high profile failures. Once the founders realize the limits of their process for predicting success, the business model may not be as viable as it seems today.

  • K. Williams

    I don’t think the logic here makes sense. Assuming that the event being predicted is independent of the crowd that’s doing the predicting — which is the case with book sales or movie grosses, etc. — there’s no reason that the outcomes need to follow a bell curve for the crowd to do a good job of predicting. In fact, we have quite a bit of concrete evidence in this regard: the crowd of bettors at the Hollywood Stock Exchange actually does a good job of predicting the grosses of Hollywood films, better, in fact, than any Hollywood studio’s internal forecast.

  • I’ve heard in weather forecasting that prediction models are tested by feeding past data and checking to see of the model predicts the past outcomes. There must be a way to do that with this system. I’ll make a prediction: if there is a way (like in weather forecasting) to test the model’s accuracy the promoters will never participate.

  • Rob

    I must misunderstand the business model of mediapredict, because I don’t see Hollywood Stock Exchange as a relevant comparison. I thought their goal was to sell voting information to companies to help them make go or no-go decisions on projects, and I don’t think that will work for power law results. If they are just going to sell companies access to daily market prices of the media stocks people trade, that is a different idea entirely.