QA monitoring sample size
“Five contacts an agent a month” is a habit, not a method. If you want a quality score you can defend — one that means roughly the same thing for a busy agent as a quiet one — the sample has to be sized to a confidence level and a margin of error. This tool does that, then totals the monitoring load and the analyst hours it costs, so you can plan QA capacity instead of guessing it.
Your numbers
per agent / period
across the team
total volume
/ period
Sample per agent vs the margin of error you ask for — precision gets expensive fast.
How the number is built
How it works
This is the standard sample-size formula for estimating a proportion. Start with n₀ = z² × p(1−p) ÷ E², where z is the confidence multiplier (1.64 at 90%, 1.96 at 95%, 2.58 at 99%), p is the expected pass rate and E is the margin of error. Because each agent only handles so many contacts in a period, we apply the finite-population correction n = n₀ ÷ (1 + (n₀−1)÷N), which pulls the sample down when the population N is small. The pass rate matters because variance is highest near 50%: if you genuinely don’t know the rate, leave it at 50% for the safest (largest) sample. The headline is per agent, because a fair score has to be defensible at the level you act on it — the individual. Total load, analyst hours and the share of volume follow from there. These are planning-grade figures to size the QA function, not a substitute for a statistician on a regulated programme.
Pair this with the quality scorecard & calibration tracker, read designing a meaningful QA programme, or learn the craft in the quality Academy track.