Definition
Recall is the share of truly positive cases correctly identified as positive:
recall = true positives / (true positives + false negatives)
It answers: among all actual positive cases, how many did the system find? Recall is also called sensitivity or the true-positive rate in many fields.
Increasing recall often lowers precision because the system flags more uncertain cases. The right threshold depends on the cost of misses, the cost of false alarms, and what downstream review can absorb.
Distinguish it from nearby terms
Precision measures correctness among predicted positives. Recall measures coverage of actual positives. Recall says nothing about how many false positives were produced to achieve that coverage.
Check your understanding
A retrieval system finds every relevant policy document but also returns 10,000 irrelevant files. It has high recall and poor precision, which may still leave the user unable to find the answer.