Field test

Significant

0 XP

Sound
Haptics
Replay calibration

Behind the field test

How Significant was designed

Why this game exists

Product teams still have to decide what an experiment supports. Significant turns early peeking, tiny samples, and broken traffic splits into short practice cases. The game scores the evidence shown before the call. Hidden simulator values never turn an unsupported choice into the right answer.

Key trade-offs

Ten seeded scenario types cover clean results and common experiment failures. Ship means the visible evidence supports release. Kill means it supports stopping the tested version. Keep Running means more valid evidence can still change the decision and continued exposure is acceptable. Progress stays in your browser without an account.

The engagement system and its limits

The game uses streaks, variable rewards, and a daily puzzle. It does not use fake scarcity, guilt copy, or notifications. Sound and motion follow your preferences. Analytics record coarse interaction events. Exception Room start events include a random run seed for reproducibility. We do not send scenario text, clipboard contents, or your local progress profile.

How it was built

Next.js and TypeScript on Vercel, built with AI coding agents. The statistics are real: binomial simulation per arm per day, two-proportion z-tests, and invariant tests across hundreds of seeds.

Back to the game