is ``Sample Ratio Mismatch (SRM) detected. P-value...
# ask-questions
is ``*Sample Ratio Mismatch (SRM) detected. P-value below 0.001*.` similar to the concept of A/B experiment allocation/selection bias ?
Hi Bhargav, Sample Ratio Mismatch (SRM) errors and A/B experiment allocation bias are related concepts but they refer to slightly different issues. With the SRM error you’re seeing, SRM occurs when the actual allocation of participants to different groups in an experiment deviates from the intended or planned allocation ratio. For example, if you plan for a 50-50 split (50% of participants in Group A and 50% in Group B) but due to a technical issue or implementation mistake, the allocation becomes 60-40, that would be a Sample Ratio Mismatch error.
We set the default P-value to 0.001, which means there is a very, very low probability that the SRM was caused by random chance. We observe this error almost exclusively when there’s been a mistake in the implementation.