Precision is a metric that measures how many of a model’s positive predictions were actually correct. High precision means that when the model says “yes,” it is usually right.
What it means in plain English
Precision answers: of all the cases the model flagged as positive, what fraction really were? It focuses on the cost of false alarms. A model with high precision rarely raises a false positive — when it flags something, you can trust it — though it may miss some real cases. Precision is usually considered alongside recall, which measures how many real positives the model caught.
You prioritise precision when false positives are costly.
A simple example
For a spam filter, high precision means that emails it marks as spam almost always really are spam — so genuine emails rarely get wrongly filtered out, which is exactly what users care about most.
Why it matters
Precision is a fundamental way to evaluate whether a model’s positive predictions can be trusted. Choosing whether to optimise for precision or recall depends on which error is more costly, making it a key concept in judging real-world model performance.
Related terms
- F1 Score — combines precision and recall into one number.
- Confusion Matrix — the table precision is calculated from.
- Classification — the task precision evaluates.
Frequently asked questions
What is precision in machine learning?
Precision measures how many of a model’s positive predictions were actually correct — when it says “yes,” how often it is right.
How is precision different from recall?
Precision focuses on the cost of false alarms (false positives), while recall measures how many real positives the model caught. They are usually considered together.
