Experiment Bucketing: Deterministic Hash Assignment

Behind every mobile A/B test sits a simple but easy-to-get-wrong mechanism: how do you decide, consistently, which variant a given user sees? This simulation renders 900 experiment "buckets" as a 3D grid and lets you ramp a treatment allocation up and down while comparing two real implementation strategies — deterministic hashing of a stable user ID (the correct approach used by production feature-flag systems) against naive re-randomization on every change (a real bug pattern that silently reshuffles already-running users). Watch how a hash-based rollout only ever grows the treatment group without disturbing users already assigned, while the naive approach flips large swaths of the grid on every ramp change, and use the "new experiment" control to see why two concurrent tests need independent salts to stay statistically uncorrelated.