A/B testing
A/B 테스트
Also known as: split testing
Showing two versions at the same time and comparing results to pick the better one.
The essential part is random assignment in the same period. Comparing last week with this week mixes in seasonality and day of week.
Small samples let chance decide. Calling a winner after a few days is how results evaporate the next month.
For small products, watching five users often beats an A/B test. When the numbers are thin, observation substitutes.
- RequirementsSame period, random assignment, enough sample.
- AlternativesUser observation, surveys, staged rollout.