메인 내용으로 이동

Sample Ratio Mismatch

Let's say we split the experiment (A/B testing) into two groups, 50% and 50%. We will use some routers in the middle to decide if the users fall into category A or B. If the final result turned out that Category A accepted 49.999999% and Category B accepted 50.000001%, that could happen by chance. However, if the result happened to be 80% and 20%, that means the test is either misconfigured say it simply happened by chance. This is Sample Ratio Mismatch, and it may skew the experiment result that we have to abandon this test result.