Every Verdika memo declares how much confidence its recommendation deserves, with a mathematical cap based on evidence quality. And every decision can record how it actually turned out. This page shows the aggregate, straight from the database: no cherry-picked rows, no makeup. If declared confidence and the real hit rate ever diverge, it shows here first.
The method is documented: how a decision memo is structured and why confidence cannot be faked. The data on this page comes from the public endpoint /calibration/public, which anyone can query.
Analyze your decision, free