Setting up a proper A/B experiment involves selecting the appropriate sample size. This ensures the test is adequately powered to detect the change while minimizing statistical noise. In statistical terms, we want to detect the minimum detectable effect (MDE) with statistical confidence. Having too few samples produces statistically ambiguous results, and having too many samples can be overkill and a waste of time and resources.
The preexisting or expected conversion rate of the control group. This could be a click-through rate, retention rate, or positive rating rate.
The relative minimum difference in conversion rate between the test and control group we want to observe. This can be the expected effect, or some minimum threshold where the experiment is not worth running (ie. impact is too small to care about).
The following values are set to their optimal defaults, but feel free to play with them to see how they affect the exposures.
A/B Split Ratio (Test vs. Control)
0.5 is an evenly split 50/50 test
Significance Level (α)
The probability that a statistically significant
difference is detected even though no actual difference
exists
Power (1-ß)
The probability that the minimum detectable effect will
be detected (assuming it exists)