Run faster experiments with CUPED

CUPED uses regressions on pre-experiment data to significantly decrease your experiment's time to convergence, and it's available out of the box with Eppo. Get equal signal from less data via the same statistical method used at Airbnb, Netflix, and Uber.

Eppo homescreen screenshot

A basic CUPED explainer: how to run faster experiments

Bolt icon

Originally developed at Microsoft in 2013

CUPED stands for Controlled-experiment Using Pre-Experiment Data and was initially introduced in a 2013 paper.
Diagnose icon

Regressions based on pre-experiment data

CUPED reduces the variance of experiment metrics by running regressions on pre-experimental data.
Chart icon

Less variance means smaller sample sizes

Reducing the variance of your experiment metrics means you can reduce your sample size and still get the same insights.
Database icon

Spend less time waiting on your experiments

Implementing CUPED for your experiments helps you make better decisions, faster.

Use any feature flagging tool or in-house setup

Unleash logoOptimizely logoLaunchDarkly logo

Integrate natively with most modern cloud data warehouses

Snowflake logoGoogle BigQuery logoAWS redshift logo

Run better, faster experiments with CUPED in Eppo