hard20 pts · first answer countsab-testingvariance-reduction

Using pre-experiment data

Your revenue metric is very noisy, so tests take months. What does CUPED (using pre-experiment covariates) do?

A.It removes outliers from the treatment group onlyB.It adjusts the metric using each user pre-experiment value of a correlated metric, reducing variance without biasing the treatment effectC.It doubles the sample size by copying usersD.It replaces the t-test with a non-parametric test

← Unexpected group sizes Difference-in-differences assumption →