← Glossary
Evaluation concepts

Precision

An evaluation metric that measures how many positive predictions were actually correct.

Precision answers: when the system says yes, how often is it right? It is especially important when false positives are expensive, such as incorrectly flagging safe content, routing a ticket to the wrong expert, or blocking a legitimate action.

Precision usually trades off against recall. Raising a classifier threshold can reduce false positives and improve precision, but it may also miss more true positives.

Related terms
recallf1-scoreconfusion-matrixclassifieragent-evaluation