Precision
Of all positive predictions, what proportion was actually positive. High precision means few false positives - when the model says "yes," it's usually right.
In-Depth Explanation
Precision measures prediction quality for the positive class. It answers: "When the model predicts positive, how often is it correct?"
Precision formula: Precision = True Positives / (True Positives + False Positives)
Precision focus:
- Minimises false positives
- Important when positive predictions trigger action
- Trade-off with recall typically
High precision needed when:
- False positives are costly/harmful
- Resources to act on positives are limited
- User trust depends on prediction quality
Examples:
- Email spam filter (false positives lose real email)
- Fraud alerts (false positives annoy customers)
- Medical screening (false positives cause anxiety/procedures)
Business Context
Optimise for precision when false positives are expensive. If each "yes" triggers costly action, you want those predictions to be right.
How Clever Ops Uses This
We tune AI models for appropriate precision levels based on Australian business contexts and the cost of different error types.
Example Use Case
"A fraud detection system with 90% precision means 9 out of 10 fraud alerts are actual fraud - manageable for human review."