Recall
Of all the actual positive cases in the data, the fraction the model correctly identified — a measure of how much the model misses.
In machine learning, recall answers a simple question: of all the actual positive cases, how many did the model catch? To calculate it, you divide the true positives by all actual positives. A recall of eighty-five percent means the model found eighty-five percent of the genuine positives but missed fifteen percent.
You should prioritize recall when missing a positive case is costly. For instance, a search tool needs high recall to ensure it flags every query that requires a live database search. If recall is too low, the system will miss queries, leading to outdated or fabricated answers.
However, maximizing recall comes with a trade-off. If you lower the threshold to catch more true positives, you will inevitably let in more false positives, which lowers your precision. Finding the right balance depends entirely on your goals. When both errors matter equally, you can use the F1 score, which averages the two metrics.
Recall is especially critical when dealing with highly imbalanced data. A model can easily look accurate just by guessing the majority class every time, but tracking recall for the minority class will reveal whether the model is actually finding those rare, important cases.
What recall is
Recall answers the question: of all the actual positive cases, how many did the model catch? Formally, it is the number of true positives divided by all actual positives (true positives plus false negatives). A recall of 0.85 means the model found 85% of the genuine positives and missed 15%.
When recall matters most
Recall is the metric to prioritise when missing a positive case is costly. A well-formedness classifier used to flag queries for expansion should have high recall — it's better to over-flag slightly ambiguous queries for review than to miss genuinely ambiguous ones and leave them unhandled. A grounding classifier with low recall would fail to trigger live search for queries that genuinely needed it, producing outdated or hallucinated answers.
The precision–recall trade-off
Recall and precision pull in opposite directions. Lowering the decision threshold catches more actual positives (higher recall) but also catches more false positives (lower precision). The right balance depends on what the model is used for. When neither error type dominates, the F1 score — the harmonic mean of the two — provides a single summary number.
For models with severe class imbalance, recall for the minority class is often the number that matters most — it's easy to score high overall accuracy by always predicting the majority class, but recall will reveal that the model never finds the rare cases it actually needs to catch.