NDCG for Search Relevance

What to do

1. Use literal NDCG when you have a ranked list and graded relevance labels: web search results, retrieval candidates, recommendation lists, RAG retrieval quality, or ranked support suggestions.
2. Use an NDCG-like ranked-evidence metric only when the system is actually ranking evidence, sources, files, actions, or plans before deciding what to do.
3. Choose the cutoff deliberately, such as NDCG@5 or NDCG@10, based on how many results users actually inspect.
4. Watch for label quality, position bias in click data, query mix drift, and improvements that help common queries while hurting rare critical queries.

Evidence to preserve

• Preserve the inputs, versions, configurations, raw outcomes, and results for NDCG, search relevance, confidence engineer, ndcg search relevance needed to reproduce work on NDCG for Search Relevance.
• Report results for NDCG, search relevance, confidence engineer, ndcg search relevance by relevant slice, separate blocker failures from averages, state uncertainty and blind spots, and connect the result to a release decision.