Evaluation & Quality
Precision
Of the items a model marked positive, the fraction that were actually positive.
Example
Of the emails marked as phishing, precision measures how many truly were phishing.
Why it matters
It helps people judge how much trust to place in a flagged result.
You'll hear it
“Most of the alerts turned out to be real problems.”
Can you control it?
Set by the team
The people building or running the AI choose this setup. An everyday user generally needs their help to change how this part works.
Common questions
- Can precision improve while more genuine cases are missed?
- Yes. A stricter detector may flag fewer cases and leave more true problems unnoticed.
- Does a high score mean every alert is correct?
- No. Some false alarms can remain even with a strong overall result.
- Can precision change when the condition becomes rarer?
- Yes. More ordinary cases can produce more false alarms relative to real detections.