To calculate recall, divide the number of correctly detected positives by the total number of positives in the evaluation data. Undetected positives count against recall.
Increasing sensitivity may find more positives while also producing more false alerts. Comparing recall and precision helps teams choose settings that fit their review capacity and the consequences of missed cases.
ELI5
For example, if 20 invoices contain errors and a tool correctly flags 15 of them, it finds 75 percent of the actual errors.
To see if those alerts are useful, also check how many of its flagged invoices were fine.
Is recall the same as sensitivity?
In binary classification, recall for the positive class is also called sensitivity or the true positive rate.
What does low recall indicate?
It indicates that the system missed a large share of the positive cases in the evaluation set.

