{"blocks": [{"key": "153f4389", "text": "Scenario", "type": "header-two", "depth": 0, "inlineStyleRanges": [], "entityRanges": [], "data": {}}, {"key": "c76eab71", "text": "A new ad-ranking algorithm shows a 5% overall CTR lift but a 100% lift for Indian males aged 18-24.", "type": "unstyled", "depth": 0, "inlineStyleRanges": [], "entityRanges": [], "data": {}}, {"key": "9acfac5f", "text": "Question", "type": "header-two", "depth": 0, "inlineStyleRanges": [], "entityRanges": [], "data": {}}, {"key": "83db7b46", "text": "What hypotheses could explain why the overall lift is 5% while one demographic segment shows a 100% lift? How would you validate whether this lift is statistically significant and not due to random noise or confounding? What additional metrics or slicing would you examine before rolling out the algorithm globally?", "type": "unstyled", "depth": 0, "inlineStyleRanges": [], "entityRanges": [], "data": {}}, {"key": "d5b4799b", "text": "Hints", "type": "header-two", "depth": 0, "inlineStyleRanges": [], "entityRanges": [], "data": {}}, {"key": "ba81c6f3", "text": "Discuss segmentation bias, sample size, Simpson’s paradox, experiment design, and follow-up analyses.", "type": "unstyled", "depth": 0, "inlineStyleRanges": [], "entityRanges": [], "data": {}}], "entityMap": {}}