Spearman's rank correlation coefficient measures how consistently two ranked measures move together. In search evaluation, it links offline ranking metrics with online business metrics such as CTR and CVR. section 12.4.2
A practical measurement workflow uses it to:
- Collect SERPs from the current search algorithm and calculate offline metrics such as NDCG and MRR. section 12.4.2
- Measure CTR and CVR across different time periods, then identify which offline metric correlates most strongly with each online metric. section 12.4.2
- Use the selected offline metric in a simple regression model to provide an initial estimate of the minimum detectable effect for an A/B test. section 12.4.2
Search-metric calibration
A stock-photo search team compares human- and implicitly labeled NDCG and MRR with observed CTR and CVR. It uses the strongest relationship to guide future experiment planning, while deferring behavioral simulation as a more complex refinement. section 12.4.2
