To conduct an [A/B test](https://www.evanmiller.org/ab-testing/) and [calculate the min. sample size](AB%20Test%20Sample%20Sizes.md), it is required to answer the following questions first: 1. What is the subject of measurement and the population? - A user, a click[-through], a purchase, a search, etc. 2. What are the two different variations you want to A/B test? 3. What is the definition of success across both variations? - Phrase this definition using "How much ..." or "How many ...". 4. What is the minimum detectable (numeric) effect between the two variations you want to establish? - This numeric "difference" is the min. effect size the test can reliably detect. - It can be formulated either in absolute terms for both variations or as a relative (%age) change with respect to the $H_0$ baseline of A. If this reminds you of [the PPDAC Cycle](The%20PPDAC%20Cycle.md), that isn't coincidental - you should use the cycle to conduct th test.