Wisdom of the crowd? Why prediction platforms could be big for business
How can the C-suite be sure it’s making a good decision? It’s a difficult question. Corporate history is filled with examples of cognitive dissonance, groupthink and other destructive management biases. Having a platform through which staff could offer their own forecast on an event may give bosses more confidence to pursue a certain course of action.
Prediction market platforms are being touted as a solution for executives looking for a second opinion. One private service, Percypt, is due to launch in a few days’ time. Its aim is to allow managers to ask a private crowd of staff members for their thoughts on the likelihood of an event which they expect to occur.
For a bank, this could be the chance that a certain client will default. Such services could be used in auditing, compiling market diaries or even board reporting – giving the C-suite more ammunition when convincing the board that it will hit its quarterly sales target, for example.
But how would corporate prediction platforms work, and how accurate could they be?
Principles of prediction
First, it is important to stave off the herd bias, where estimates are anchored by an ad hoc estimate already known by the crowd. According to Karl Mattingly, chief executive of SlowVoice, the firm which is developing Percypt, each crowd member’s forecast should remain anonymous.
Second, for estimations to be accurate, it is necessary to reward forecasters who are accurate. “Staff could convert points into days of leave, or rewards could be financial,” says Mattingly.
Third, crowd members must also have skin in the game. Betting markets, it is said, are more accurate than surveys because players risk losing their money if they are wrong. Gamblers will only bet if they are confident in the outcome.
Almanis, which has been functioning as a proof of concept for Percypt, uses a points based system. It is a free, public platform which focuses on geopolitical events, and tries to avoid the pitfalls of other prediction sites which are based on stock markets. It allocates each player 1,000 points to be put behind any of the site’s forecasts (such as: “Will a UK MP with revealed connection to the Panama Papers resign before the end of September 2016?”) based on how confident they are. This way, better forecasters are rewarded and gain more currency over time, while weaker forecasters are weeded out.
Fourth, Percypt will use episodic surveys, rather than having open polls. Questions must contain a fixed time period, and qualitative issues should be avoided. “Questions like ‘Is management doing a good job’ should be avoided,” says Mattingly.
Accuracy
The jury is still out on the accuracy of this system. Almanis incorrectly predicted that the ECB would announce an increase in its monthly asset purchase targets in December.
However, it claims that its forecasts are 94.7 per cent correct at the time the outcome is settled, and 80 per cent accurate 20 days beforehand. The platform has only been going since 2015, so the true wisdom of its crowd has yet to be determined.
Percypt functions more tightly. It is designed to provide a snapshot of opinion, and SlowVoice anticipates that surveys will often be timed to provide additional justification for a decision managers are already confident will happen. Unlike Almanis, Percypt's crowd cannot participate in question curation. Neither can they see the results of the forecast being constantly updated as predictions are made, which would prevent people from simply following the herd.
Whether or not the crowd is wise remains to be seen.