Contact center quality monitoring

Sampling was never a methodology decision

The standard program reviews a handful of interactions per agent per month. That number did not come from statistics. It came from arithmetic: auditing one interaction takes about 2.5 times its duration, an analyst has roughly 121 productive hours a month, and the rest is whatever fits.

A 40-seat operation handling around 36,000 interactions a month, with four analysts, audits about 4% of them. The other 96% happen and disappear. Nobody listened, nobody scored, nobody coached. When one of those becomes a dispute, a complaint or a regulatory inquiry, the answer is that no record exists.

How to size a quality team

The formula has three variables: analysts = (monthly volume × target coverage) ÷ audits per analyst. The first term is how many audits you need. The second is how many each person delivers. Most sizing errors live in the second.

Three discounts have to be applied. Productive hours: of a 173-hour monthly schedule, roughly 121 are actually spent auditing, with the rest going to calibration, feedback sessions, quality meetings and reporting. Time per audit: about 2.5× the interaction length, because reviewing means replaying segments, completing the scorecard, writing the justification and logging the result. Availability: PTO, sick days and absence take around 10% of team capacity across the year.

Run those and one analyst delivers roughly 415 audits a month at a 7-minute average handle time, closer to 375 once availability is applied. Anything above 400 per person is sizing for a month that never happens.

Why hiring never closes the gap

Coverage grows in a straight line with headcount, and the unit cost never moves. Running the same operation with 4 analysts and with 90 produces an identical cost per audited interaction while payroll multiplies more than twentyfold. Manual auditing has no economies of scale: every additional audit demands an additional person-hour, in the same ratio, forever.

That is why reaching 99% coverage by hiring takes 97 analysts and around $751,750 a month in a 40-seat operation. It is not a budget problem. It is a model that stops working before the budget does.

What coverage target makes sense

No US regulation sets a minimum sampling rate. What exists is sector obligation around conduct and evidence: CFPB and FDCPA in collections and lending, HIPAA in healthcare, TCPA in outbound contact. None of them prescribes a percentage, and all of them assume you can account for a specific interaction when asked.

A more useful way to set the target: ask what share of problem interactions you are willing to never discover. At 5% coverage the answer is 95%. Most quality leaders have never said that number out loud, but it is the decision already made by the size of the team.

What changes at full coverage

When every interaction is scored, three things change. Findings become representative rather than anecdotal, so a pattern across the operation is visible instead of inferred from forty calls. Coaching gets specific, because the exact moment is retrievable. And the quality team stops being a listening operation: the analysts move to calibrating criteria, reviewing edge cases and running coaching, which is the work the listening hours were crowding out.

One operation running BRIGGS went from 11 quality monitors to 4 while coverage went from 5% to 99%. The seven who moved were not let go. They went to the work that scoring interactions by hand had been preventing.

The condition that makes this real is agreement. Full coverage is worthless if the score does not match the judgment of the evaluators your operation already trusts, which is why the model is calibrated against your own scorecard until it agrees with them 99% of the time.