Definition
The ratio of review workload generated by automated change production (model versions, prompt variants, fine-tunes, config changes) to the human review capacity actually available for it.
The Measurement Gap
There is no metric that tells you whether generated change volume exceeds the human review budget, which means review silently becomes sampling with no declared sampling policy.
Formula
CRAR = sum(changes_r * review_hours_r) / available_review_hours, reported with the age of the unreviewed backlog
Proposed formula, pending validation pilots.
At a Glance
Decision Supported
Whether review is a real gate or silent sampling, and when to throttle generation or expand review capacity.