Prediction Markets as Idea Validators: Kalshi, Polymarket and Manifold
A prediction market lets people buy and sell contracts that pay out if an event happens. If a contract paying $1 on "Product X reaches 10,000 users by December" trades at 30 cents, the market is saying, roughly, that the chance is about 30%. For founders, that raises an intriguing question: can betting markets help validate ideas?
Why markets forecast well
The argument for prediction markets is about incentives and aggregation. Traders who are right make money; traders who are wrong lose it. Over time, prices tend to reflect the combined information of many participants, including people who would never speak up in a survey or a meeting. As we noted in our survey design guide, election betting odds were a good guide to US presidential results long before scientific polling existed.
The main models
| Platform | Model | Notes |
|---|---|---|
| Iowa Electronic Markets | Small-stakes academic market run by the University of Iowa since 1988 | Known mainly for election forecasting research |
| Kalshi | US exchange regulated by the Commodity Futures Trading Commission (CFTC) as a designated contract market | Real money; contracts on economics, weather, politics and, more recently, sports |
| Polymarket | Crypto-based market settled in stablecoins | Grew rapidly around the 2024 US election; its US access has been shaped by regulation |
| Manifold | Play-money market where anyone can create questions | No real-money stakes; popular with forecasters and tech communities |
The legal position of real-money prediction markets has moved quickly. Kalshi won a federal court case in 2024 that allowed it to list contracts on control of Congress, and later expanded into sports-related contracts that some state gaming regulators have challenged. Polymarket settled with the CFTC in 2022 and blocked US users for a time, before moving to re-enter the US market through a regulated route. Because this area is changing fast, check the current status where you live before using any of them.
Internal markets inside companies
Several large companies have experimented with prediction markets for their own employees. Hewlett-Packard ran experiments in the 1990s with economists including Charles Plott to forecast printer sales, and Google launched an internal market around 2005, which researchers later analysed and found to be reasonably accurate, with some biases such as optimism about the company's own projects. Typical questions were launch dates, user numbers and quarterly targets.
For a startup, a lightweight version is easy: create play-money questions on a platform such as Manifold, or run a simple internal market among your team, advisers and early users. "Will we sign three paying pilots by the end of the quarter?" priced at 15% by the people closest to the work is a useful warning, especially if the founders privately believe 80%.
What markets can and cannot validate
- They can aggregate expert and insider expectations about measurable outcomes: launch timing, adoption milestones, competitor moves, regulatory decisions.
- They can expose over-optimism inside a team, because betting makes people commit to a number.
- They cannot tell you what customers will do if the traders have no customer information. A market of founders betting on their own idea mostly measures founder sentiment.
- They struggle when trading is thin: a handful of traders can move prices a long way.
- They need clear resolution criteria. "Will the product succeed?" cannot be settled; "Will it have 1,000 weekly active users on 1 June?" can.
Using a market alongside real tests
The most effective use is in combination. Run your interviews and smoke test, then ask your team and advisers to bet on the next milestone. If the market and the evidence disagree, find out why. Prediction markets are also a startup category of their own; if you are thinking of building in that space, read our guide to validating betting-industry ideas, where licensing is often the riskiest assumption.
