Precision measures how many returned results are actually correct, so low precision means false positives. Recall measures how many of the truly relevant items were captured, so low recall means things were missed.
The two trade off against each other, and the F1 score combines them into a single figure.
Alternative Names:
Precision, Recall, Accuracy Metrics
Why it Matters?
These are the metrics that make accuracy claims meaningful, and a vendor citing a single accuracy percentage is usually obscuring one of them. The right target depends on the task: privilege review demands very high recall because one missed document creates waiver exposure, while a first-pass issue tagging tool can tolerate lower recall if a human reviews the output. Asking which metric a tool optimizes for is the most useful procurement question available.
Frequently Confused with
Related terms
Frequently asked questions
Which matters more, precision or recall?
Why is a single accuracy number insufficient?
