Encyclopedia Evalica / Observability / Sampling rate
Sampling rate
/'sa.mplihng rayt/The fraction of production requests captured for logging or scoring. Sampling trades off cost and storage against visibility. (noun)
“We lowered the sampling rate to 10% to cut costs while still spotting regressions.”
Customer example
Dropbox runs online LLM-as-a-judge at a controlled sampling rate on production queries, using the same rubric as offline tests to catch regressions while keeping eval cost manageable. Read more
Related Observability terms
- AI observability •
- Alert / threshold •
- Dashboard •
- Data flywheel •
- Deep search •
- Drift •
- Error rate •
- Feedback loop •
- Logs •
- Model drift •
- Online evaluation (production scoring) •
- P50 / P95 / P99 (Percentiles) •
- Service Level Indicator (SLI) •
- Service Level Objective (SLO) •
- Time-to-first-token (TTFT) •
- Token usage / cost tracking •
- Topics
From the docs
Get started with Evals
Braintrust is the AI observability and eval platform for production AI. By connecting evals and observability in one workflow, teams at Notion, Stripe, Zapier, Vercel, and Ramp ship quality AI products at scale.
Start building