Why sampling 2% of calls is costing your lending team more than you think
Manual QA looks cheap until you price the 98% of calls nobody hears: soft PTPs, conduct slips and coaching that arrives a week late.
Most collection floors audit somewhere between 1% and 3% of calls. An auditor listens to 15–20 recordings a day, fills a sheet, and the agent hears about it at the next weekly huddle. On paper the QA line item is small. In practice it is the most expensive blind spot in the business.
The hidden cost is not the auditor
A 200-seat floor making 60,000 calls a month needs over 150 auditors to hear every call manually. Nobody staffs that, so the floor samples instead. The problem is that sampling is designed for consistent processes. Collection calls vary with bucket, borrower mood and agent fatigue.
The calls that leak the most money, such as soft promises, unrecorded amounts and agents who skip the identity check, are exactly the ones a random 2% sample misses.
Three leaks we see in every pilot
Soft PTPs: the agent logs a promise, but the call never pins down an amount or a date. These rarely convert. Conduct risk: a single threatening or misleading line becomes a grievance, or worse, a regulator complaint. Bucket blindness: overdue and due-date calls need different scripts, but sampled QA cannot show performance by bucket with any confidence.
What 100% coverage changes
When every call is scored on the same scorecard, the conversation shifts from "was this call good?" to "which parameter is costing us money this week?". Team leads stop listening at random. They verify the flagged calls and coach the one parameter each agent misses most.
In pilots this typically lowers QA cost by around 80% and lifts PTP conversion in the high teens. These figures are illustrative; we size them against your own data in the first two weeks.