Home / Search evaluation
Search evaluation guides
13 published guides. Start with How to evaluate semantic search.
A practical evaluation method: build a test set, pick metrics, compare systems and catch regressions.
Read the guideLearning path
All guides
Understand it3 guides
Start hereHow to evaluate semantic searchA practical evaluation method: build a test set, pick metrics, compare systems and catch regressions.Start hereInformation retrieval metricsRecall, precision, MRR, nDCG and MAP side by side, with the question each one answers.ExplainerRetrieval benchmarksWhat the main retrieval benchmarks test and how far their scores transfer to your data.
How it works7 guides
ExplainerRecall@kWhat recall@k measures, how to calculate it, and how to choose k for your search or RAG system.ExplainerPrecision@kHow precision@k complements recall@k and which one to optimise for search versus RAG.ExplainerMean reciprocal rankHow MRR rewards putting the first right answer high, with the formula and a worked example.ExplainernDCGHow nDCG scores graded relevance and position together, and how to compute it.ExplainerANN recallHow to measure recall of an approximate index against exact search and tune it.ExplainerOnline search metrics and analyticsThe production signals that show whether search works for real users, and how to A/B test changes.ExplainerMonitoring search quality in productionDrift, freshness and quality signals to watch after launch, with alert thresholds.