Question 1

Why should my team run an internal prediction market?

Accepted Answer

Google ran internal prediction markets for nearly 20 years across two platforms. The consistent finding was that the aggregate of employees' private estimates outperformed any individual expert, any official forecast, and any executive's gut feel. Ford found the same thing — their internal market beat the official sales forecast by 25% on mean squared error. The mechanism is simple: when people estimate privately and independently, rather than reacting to what the loudest person in the room just said, the real signal surfaces.

Question 2

Why are prediction markets better than polls or surveys?

Accepted Answer

A Slack poll tells you what people are willing to say publicly, after seeing what everyone else already said. A prediction market asks each person to commit to a number before they know anyone else's answer, and then holds them accountable to it over time. The accountability part is what most tools skip entirely. Tetlock's research on thousands of forecasters found that the people who improved the fastest weren't necessarily the smartest or most experienced — they were the ones who tracked their forecasts against outcomes and updated their thinking accordingly. Most teams have no idea who their best forecasters actually are, because nothing is ever written down in a way that can be checked later.

Question 3

Do prediction markets work with a small team?

Accepted Answer

Yes, and arguably better than the trading-style version. Markets that use fake currency to buy and sell shares need lots of participants to function — with 5 to 10 people there's no liquidity, prices get stuck, and a single person can move the whole market. Probablynot works differently: everyone submits an independent probability estimate, and the average is the crowd forecast. Eight honest estimates, averaged, are surprisingly well-calibrated. This is also why Cultivate Labs, one of the longest-running enterprise forecasting platforms, dropped the trading mechanism entirely in 2022. They found the currency mechanic confused most participants, made it easy to game the leaderboard, and obscured whether someone was forecasting their true belief or just trying to profit from other people's bets. A simple probability estimate, it turns out, gets you most of the accuracy with none of the overhead.

Probablynot.bet

Find out how good your team is at forecasting the future.

Free internal prediction markets for your team - ask a question, share the link, and start keeping score.

Frequently asked questions