8.6Metrics to Track Over Time

Metric	What It Tells You	Target Trend
Gate pass rate on first attempt	Quality of AI generation + prompts	Increasing
Tests deleted during review	How much waste AI produces	Decreasing
Time from generation to merge	Efficiency of the review process	Decreasing
Mutation score of AI tests vs hand-written	Relative quality	Converging
Flaky test introduction rate	Determinism of AI tests	Decreasing

These per-gate metrics roll up into the suite-level view -- mutation score, coverage deltas per generation session, and flake rate -- covered in LLM Evals and AI Test-Suite Quality, which also extends the same evaluation discipline to LLM-backed features themselves.