You're planning an experiment on a product surface and need to decide how much traffic to allocate and how long to run it. You want the test to be sensitive enough to detect a meaningful lift without running longer than necessary.
How do you determine the right sample size and test duration for an A/B test?