4 min read
A/B testing when you do not have the traffic
The honest options for a team with two thousand users a week, none of which are "run the test anyway and hope".
Tag
3 pieces tagged “statistics”, newest first.
The honest options for a team with two thousand users a week, none of which are "run the test anyway and hope".
The definition is genuinely awkward, most working explanations are subtly wrong, and the practical consequences of getting it wrong are large. Here is the careful version.
Almost every number a business cares about is right-skewed, and almost every business reports it as a mean. Here is what that does, why nobody notices, and what to publish instead.