Experi

Production-grade A/B testing with CUPED + Bayesian Inference on Databricks

Most A/B tests take too long to reach significance and produce results nobody trusts six months later.

Experi fixes both problems. CUPED variance reduction strips pre-experiment noise so your tests reach significance faster. Bayesian inference gives you P(Treatment > Control), a probability your team actually understands, instead of a p-value nobody can explain. Every experiment is logged to MLflow and persisted to Delta Lake so any result can be reproduced and audited.

How to use it: Adjust the sliders and click Run Experiment to see the posterior distributions. Upload your own CSV with columns treatment, pre_conversion_rate, and converted to analyze real data. Use the Sample Size Calculator to plan before you run.