EVALS & OBSERVABILITY • 238 / 397
Measure whether the AI is good and trace where it fails.

Recall

Recall asks: of all actual positive items, how many did the system successfully find?

Think of it like

Think of Recall like testing and telemetry for any production service: quality must be measured, not guessed.

Real life

Teams use Recall to decide whether a new release is safer, faster or more useful.

SRE lens

High recall matters when missing a severe incident is costly.

Remember thisRecall asks: of all actual positive items, how many did the system successfully find?
AIForSREJump to a concept
Search is optional. The main journey is simply ↓ Next.