Sampling calculator
See how many traces and how much data you keep at a sampling rate, with or without keeping every error.
How it works
Head sampling keeps a fixed share of traces decided at the start. Keeping every error trace as well, through tail sampling, costs little because errors are rare.
kept = traces × (errors + (1 − errors) × rate)Questions
What rate should I start with?
10% is a common start for busy services. Keep 100% for low-traffic or critical paths.
Does sampling break metrics?
Request metrics should come from metrics, not counted spans, so they stay exact.