Your question is Subset vs Whole Experiment Result. Take a moment with it on the right.
Talk me through your thinking if you like. When you're confident, submit your answer and I'll grade it like a real screen (7/10 or better passes).
How do you handle scenarios where an experiment shows a positive result in a subset but not the whole?
Explain how you would determine whether the subgroup result is credible, distinguish treatment-effect heterogeneity from noise, and decide whether to ship. Address metric definitions, subgroup pre-registration, statistical testing, multiple comparisons, and guardrails. State what additional information you would request before making a recommendation.
