The real cost of manual QA
in a contact center.

Every contact-center leader knows quality has to be monitored. Very few have actually calculated what doing it by hand costs — or why, at any budget, coverage stalls at low single digits.

A formula worth running once

The cost of manual QA comes down to three variables: how many interactions your line generates, how long it takes to review one interaction, and what an hour of the specialist doing it costs.

Worked example: a 300-agent line

375 supervisors for 300 agents is absurd, which is why no contact center actually staffs that way. A real QA team of 5–7 people gets through 1.5–2% of interactions. The other 98% is a blind spot — and that is where the worst calls, the compliance breaches and the churn drivers live.

Why a 2% sample is not "small but sufficient"

• The sample is biased. Long calls and complaints that reached escalation get reviewed most often. Quiet negatives — the customer who says a polite goodbye and defects to a competitor — almost never make it into the sample.
• Scoring is subjective. Two supervisors can differ by up to ±2 points on the same call. Agents know it, so they distrust the reviews — and appeals eat up hours.
• Feedback arrives late. The manual loop of call → sample → review → feedback takes a week or more. A systemic script error has time to repeat itself thousands of times.

What automated scoring changes

AI scoring against your own checklist covers 100% of interactions on the day they happen. The marginal cost of that line item trends toward zero, and your people stop being a listening post and focus on the one thing that genuinely requires a human — coaching.

More on the methodology in the NUMU QA FAQ: calibration, appeals, and recognition accuracy on telephony audio.