Tserum

world's first continuously-updating expert discovery engine

A/B Testing

Statistical Analysis

Community

Visit WebsiteSee on Product Hunt

I started building Tserum because I was fascinated by a simple problem: how can we measure taste? Traditional surveys are great at collecting opinions, but they have a fundamental issue: people often don't have a strong incentive to think carefully or, more importantly, reveal their true beliefs. Prediction markets try to solve this by adding incentives, but they come with their own challenges such participation barriers and domination by sharks. Moreover, prediction markets can't answer subjective questions that a company could be interested in (i.e. "Which ad campaign is better?" or "which color scheme will get more users on my website?"). I became interested in Bayesian Truth Serum, a mechanism proposed by an MIT reseracher in 2004 designed around a different idea—and one that works on quesitons with no ground truth. Each Bayesian Truth Serum question asks just two things: what an individual believes (yes/no question or multiple choice), and what percentage of everyone else will agree with them. Afterwards, each respondent is given their own score. This setup is mathematically-PROVEN to be robust to cheating. The optimal strategy is to be honest, which I found fascinating. The challenge was turning that idea into something people would actually want to use. Feel free to read more about the scoring system (and understand why honesty is truly the best policy) on the "How it works" page on the website, or by scrolling down on the landing page. The first versions of Tserum were much closer to a research tool: lots of categories, percentile scores, and analysis. But after testing, I realized the experience needed to feel much more immediate. People don't want to wait for a complex analysis—they want to answer something interesting and instantly understand what happened. That realization led to a few major changes: - Closed questions so users get immediate feedback. - SCIV an Elo-style rating that turns judgment into a simple, understandable number. - A stronger social loop where every question becomes something you can send to a friend and compare perspectives. - A focus on progress and discovery (gamification). The long-term vision is simple: create a continuously updating map of human judgment and discover the people whose opinions are genuinely valuable. Would love feedback on ANYTHING.