Call Sample Review Board (QA)

Call Sample Review Board (QA)

1. 01

   Pull 30 real call recordings per client before scoring anything

   Sample across inbound and outbound, peak and after-hours, and at least five calls that ended in a transfer or hangup. Twenty calls hides escalation failures; thirty surfaces them.

2. 02

   Score each call on four axes: task completion, escalation accuracy, latency, and disclosure

   Escalation accuracy means the agent routed to a human at the right moment, not merely that it routed. Latency gets a threshold in seconds, agreed with the client in writing, because a two-second pause reads as a dropped line on mobile.

3. 03

   Listen for the handoff seam, not the happy path

   The failure that costs retainers sits in the ten seconds around a transfer: context lost, caller asked to repeat themselves, or a warm transfer that lands cold. Log every seam failure with a timestamp.

4. 04

   Verify consent language and recording disclosure on every sampled call

   Confirm the disclosure plays before any data capture and that outbound calls respect the client's jurisdiction rules. A missing disclosure is a legal finding, not a quality score.

5. 05

   Reconcile usage cost against the quoted retainer line

   Total minutes, per-minute rate, telephony fees, and any LLM pass-through. If a client's average call runs longer than the pricing model assumed, the margin gap shows up here first.

6. 06

   Document the labor that remains after deployment

   Count hours per week the client's staff still spends on callbacks, reschedules, and transcript cleanup. That number belongs in the QA report because it is the honest counterweight to automation claims.

7. 07

   Issue a two-page findings memo with three prioritized fixes and a re-test date

   Rank fixes by caller impact, not engineering convenience. Set the re-test date before the memo goes out so the next review is scheduled rather than negotiated.