False positive

A good customer flagged as fraud.

A genuine person tries to sign up or pay, and the system wrongly treats them as a fraudster, so they are blocked or sent for extra checks.

A legitimate event that crosses a fraud model's decision threshold and gets classified as fraudulent. It is the fraud-detection counterpart of a false rejection in a biometric system, because both turn away a genuine person. The decision threshold is the score above which the system treats an event as fraud.

There are two ways a fraud system can be wrong:

• False positive, where a legitimate customer is flagged as fraud.
• False negative, where a real fraud case is missed and treated as legitimate.

The two trade against each other through the threshold:

1. A stricter threshold catches more fraud, so false negatives fall.
2. It also flags more honest customers, so false positives rise.
3. A looser threshold does the opposite.

False positives sit next to terms that are easy to mix up:

• False rejection rate is the biometric version, where a genuine person fails a face match.
• Manual review is where flagged cases often go for a human decision, which can rescue a false positive.
• Conversion rate is the share of people who finish a flow, which false positives reduce.