You work on an online product where a new experience is being tested in an A/B experiment. The team believes there is a small lift, but the metric is noisy and the test may be underpowered.
How would you use CUPED, or another variance-reduction technique, to improve the experiment? Explain how you would define the covariate, estimate the lift, and decide whether the variance reduction is valid.
You work on an online product where a new experience is being tested in an A/B experiment. The team believes there is a small lift, but the metric is noisy and the test may be underpowered.
How would you use CUPED, or another variance-reduction technique, to improve the experiment? Explain how you would define the covariate, estimate the lift, and decide whether the variance reduction is valid.