Even Split Increases Power of A/B Tests

Your Experiment Reaches Statistical Significance Faster with an Even Split

Do experiments finish faster when I allocate more traffic to the treatment bucket? Should I minimize risk by assigning fewer users to the treatment bucket? These are common questions asked by data scientists. This blog post shares our study on setting up the experiment sample ratio to reach statistical significance faster.

Your experiment has a higher power with an even split

The general recommendation …

a-b-testing data science experimentation power tests