Containment Rate Audit (QA)

Containment Rate Audit (QA)

1. 01

   Pull a 30-day interaction sample stratified by channel and intent

   Voice, SMS, chat, and email behave differently, so a blended containment number hides the channel that is actually failing. Sample at least 200 conversations per channel or the full population when volume is lower.

2. 02

   Define containment as resolved without human touch, then verify it against transcripts

   Platform dashboards often count a hang-up or a bot-closed session as contained. Read 25 transcripts per channel and mark each one resolved, escalated, or abandoned before you trust the reported rate.

3. 03

   Reconcile the vendor's containment claim against your own count

   A platform such as Nectar Desk may report 40 to 65 percent autonomous resolution, but that range collapses once abandoned sessions and repeat contacts inside 24 hours are removed from the numerator.

4. 04

   Segment failures by intent and route the worst three to retraining

   Group every non-contained conversation by intent label, rank by volume times handling cost, and hand the top three intents to whoever owns prompt and knowledge-base updates.

5. 05

   Check escalation handoffs for context loss

   Listen to the first 30 seconds after a warm transfer and confirm the human agent receives transcript, account history, and stated intent. Missing context turns a contained call into a frustrated repeat contact.

6. 06

   Price the gap in labor hours, not percentages

   Multiply non-contained volume by average handle time and the client's loaded hourly rate. A 12-point containment miss on 8,000 monthly calls is a retainer conversation, not a dashboard footnote.

7. 07

   Log the audit result and the retraining date in the client's service record

   Record sample size, measured containment, top failing intents, and the next review date so the next audit starts from a documented baseline rather than a fresh argument.