A fraud team's classifier flags very few transactions, and fraud is rare in the data. The business says missing a fraudulent transaction is far worse than a false alarm. Which metric should headline the report?
Show answer and explanation
Answer: A
Sensitivity, also called recall, is the fraction of actual positive cases the model identifies. It is the metric that directly tracks missed fraud, which the business has named as the costliest error. Because the positive class is rare, accuracy would look high even for a model that misses most fraud, so it should not headline this report.
Why the other options are wrong
- B. Accuracy is dominated by the abundant legitimate class under heavy imbalance, so a model that misses most fraud can still show a high score and hide the problem.
- C. Precision measures how trustworthy the alerts are, which matters for review workload, but it does not show how many fraudulent transactions slip through unflagged.
- D. Specificity describes how well legitimate transactions are cleared, which controls false alarms. It says nothing about missed fraud, the error the business cares about most.